Release Notes

Microchip TCP/IP Stack Version Log:

Please file bug-reports/bug-fixes at http://support.microchip.com/ or post to the Microchip TCP/IP -> Ethernet Forum at http://forum.microchip.com/ Look for stack updates at http://www.microchip.com/mal/ 

v5.42.08 June 2013

Changes:

1. Clarify item#4 in v5.42.04 Oct 2012 release. Change is applicable to MRF24WB/G only. (MRF24WB/G) SSL client with RSA 2048bits is enabled by default. Applies only for PIC24/32. Not supported for PIC18.
2. (WiFi G Demo Board) Disable option WF_EASY_CONFIG_DEMO in WF_Config.h to resolve compilation error with IperfAppInit().

Fixes:

1. (MRF24WB) Resolve firmware 0x120C host scan bug : Host scan asserts when it is conducted after WPA-PSK fails due to key mismatch. Refer to WFScan.c and WFMgmtMsg.c.
2. (MRF24WB) Resolve reset problem bug. Modified WF_SetCE_N and WF_SetRST_N to first set the level, than configure pin as output. Refer to WFDriverPrv.h. This change is already available in WFDriverPrv_24G.h for MRF24WG in MLA v5.42.02 Aug 2012 release.
3. (MRF24W) Resolve WiFi Console demo IPERF issue bug fix : Enlarge generic tcp rx/tx buffer size. Refer to TCPIP MRF24W.h.
4. (MRF24W) Resolve EasyConfig bug fix for retry count when DISABLE_MODULE_FW_CONNECT_MANAGER_IN_INFRASTRUCTURE is enabled. Refer to function WFEasyConfigProcess() in WFEasyconfig.c
5. (MRF24W) Resolve handling of max length SSID of AP. Refer to WFEasyConfig.c and CustomHTTPApp.c

v5.42.06 Feb 2013

Changes:

1. (MRF24WG) Due to memory constraints, all future RF module FW releases will follow this roll-out order (Odd number eg 0x3109, 0x310b) Differences: Supports Wi-Fi Direct (P2P). But no EAP and no multi-client support. (Even number eg 0x3108, 0x310a) Differences: Supports EAP (with MCHP approval) and multi-clients. But no Wi-Fi Direct support.
2. (MRF24WG) Added new project Wifi G demo. Wifi G demo currently only support softAP function,
3. (MRF24WG) In Wifi Console, for XC32-EX116-EAP_MRF24WG, supports security mode WF_SECURITY_WPA2_ENTERPRISE (EAP-PEAP/MSCHAPv2 and EAP-TTLS/MSCHAPv2) . Requires RF module FW version 0x3108 and even number releases. This requires special approval submitted to marketing.
4. (MRF24WG) In EZConfig, added mobile application demo to Explorer16+PIC32. Refer to MobileTCPServer.c Supported devices : Android (version 4.0 or later), iOS6 or later Current functions supported are (i) Temperature reading (ii) Potentiometer reading and (iii) LEDs ON/OFF toggling App name is "OpenWifiFog"
5. (MRF24WG) In WiFi Console, added cloud demo to Explorer16+PIC32. Refer to CloudTCPClient.c Current functions supported are (i) Temperature reading (ii) Potentiometer reading and (iii) LEDs ON/OFF toggling Reference web server is http://www.openwificloud.com
6. (MRF24WG) Supports multi-client DHCP server (max 4 clients). SoftAP supports up to max 4 client Requires RF module FW version 0x3108 and even number releases.
7. (MRF24WG) softAP supports active scan. Requires RF module FW version 0x3108 and future releases.
8. (MRF24WG) RF module FW version 0x3107 and future releases will NO longer support programming of regional domain via function WF_SetRegionalDomain() due to changes in FCC requirements, which does not allow programming of the regional domain.
9. (MRF24WB/G) Only WEP key index 0 is valid for security mode WEP.

10.(MRF24WG) Created new event WF_EVENT_SOFT_AP_EVENT_SUBTYPE / WF_EVENT_SOFT_AP_EVENT, for softAP mode, to indicate client's status (connected or disconnected to softAP). Requires RF module FW version 0x3108 and later releases. If prior MLA releases are used with this new RF module FW version release, you need to port over this event handling in WFEVentHandler.c and WF_ProcessEvent() (WiFi EZConfig).

11. Added full versions of RSA.c, ARCFOUR.c, and AES_PIC32MX.a to the TCP/IP Stack Distribution to support SSL and SNMPv3. Previously these files were distributed in a separate cryptographic code distribution. These files are subject to the U.S. Export Administration Regulations and other U.S. law, and may not be exported or re-exported to certain countries or to persons or entities prohibited from receiving U.S. exports (including Denied Parties, entities on the Bureau of Export Administration Entity List, and Specially Designated Nationals).

Fixes:

1. (MRF24WG) In EZConfig, SoftAP now supports Zeroconf & mDNS. Resolves wrong port issue. Search for keyword SOFTAP_ZEROCONF_SUPPORT.

v5.42.04 Oct 2012

Changes:

1. (MRF24WB/G) Patch update features are added into MRF24WB0MA/B and MRF24WG0MA/MB. MRF24WB0MA/B (no RF module FW version update is necessary) Serial Port (Xmodem) MRF24WG0MA/B (Requires RF module FW version 3107 or later.) Serial Port (Xmodem), Web Client, Web Server
2. (MRF24WB/G) FTP client is added into WiFi console demo app.
3. (MRF24WB/G) Flexible scratch memory is used to allocate TCB.
4. SSL client with RSA 2048bits is enabled by default. Applies only for PIC24/32. Not supported for PIC18.
5. 16-bit mode feature is available on PIC32.
6. (MRF24WG) Enhance SoftAP with EZConfig features with pre-scan and redirection features. Change SoftAP address from 192.168.1.1 to 192.168.1.3 Added WEP security, which requires RF module FW version 3107 or later. Refer to WF_Config.h for more information.
7. Added support for PIC32MX6XX external PHY's: RTL 8201FL. Only RMII configuration is supported with RTLPHY8201FL PHY. This PHY driver works in Default Ethernet IO mode with PIC32MX675F512H . If Alternate Ethernet IO mode is used for other PIC devices, then configuration bit need to be changed w.r.t Alternate Ethernet IO. To get the status of MAC link with this PHY, PHY Link Status need to be read twice. EthPhyGetLinkStatus(int refresh) API is used to read the PHY link status and for this PHY to read link status twice, the parameter refresh should be 1.

Fixes:

1. (MRF24WG) SoftAP under EZConfig has pre-scan and redirection features. Able to be redirected to infrastructure mode AP.
2. If SNMP_TRAP_DISABLED macro is enabled from TCPIP XXX.h file, there will be no TRAP table information for SNMP manager. That is SNMP agent won't send information related to trap. By default SNMP_TRAP_DISABLED macro is disabled.
3. Fix typo error in v5.42.02 (MRF24WG) SoftAP default address is changed from 169.254.1.1 to 192.168.1.1
4. Fixed SNMP tabular issue if SNMP instances starts other than 0.
5. Fixed ARP initialization bug that resulted in TCP packets sent to ETH MAC broadcast address.

v5.42.02 Aug 2012

Fixes:

1. (MRF24WB/G) TCPIPConfig.exe is modified to support changes in macro definitions (CFG_WF_INFRASTRUCTURE, CFG_WF_ADHOC) used in WF_Config.h.
2. (MRF24WB/G) Ad-hoc mode was left out in the TCPIP-Demo app. This mode is reinstated back into TCPIP-Demo app. 2 files are changed. .MicrochipIncludeTCPIP StackWFDebugStrings.h .TCPIPDemo AppWF_Config.h
3. (MRF24WB/G) SoftAP default address is changed from 192.168.1.1 to 169.254.1.1
4. (MRF24WG) Apply SSL fix. Root cause is traced to be due to scratch pad memory to memory copy functions in WFMac_24G.c
5. (MRF24WG) Reset fix. a. In file WFDriverCom_24G.c, MRF24WG reset sequence is modified.
   • WFHardwareInit() does hard reset (putting I/O lines in correct state)
   • ChipReset() does soft reset

b. In file WFDriverPrv_24G.h, macros WF_SetCE_N() and WF_SetRST_N() modified to first set the output level, then set the I/O direction

6. (MRF24WG) Applies to using PICDEMNet2 board and PIC18. Related to SPI interface settings. a. In file WF_Spi.c, ConfigureSpiMRF24W() modified, for PIC18 only, to change the SPI clock idle state to low.
7. (MRF24WB/G) When host attempts to read scan results while the module FW is in reconnecting state, the module FW returns WF_ERROR_NO_STORED_BSS_DESCRIPTOR error. Reading scan results is only allowed in A) connected state; B) idle state. Fix is in WaitForMgmtResponseAndReadData().

void WaitForMgmtResponseAndReadData(UINT8 expectedSubtype, UINT8 numDataBytes, UINT8 startIndex, UINT8 *p_data) { .................................... /* check header result and subtype fields */ WF_ASSERT(hdr.result == WF_SUCCESS || hdr.result == WF_ERROR_NO_STORED_BSS_DESCRIPTOR); WF_ASSERT(hdr.subtype == expectedSubtype); .................................... } 

v5.42 Jul 2012

Changes:

1. All MPLAB 8 and MPLAB X projects have been modified to use the XC16 and XC32 compilers.
2. The colon character ":" has been added as a valid terminator for an IP address in the StringToIPAddress function.
3. (MRF24WB/G) PIC18 is only supported with "TCPIP Demo App". MRF24WB0MA/MB works on PICDEM PIC18 Explorer. MRF24WG0MA/MB works on PICDemNet2 PIC 18 but has issues with PICDEM PIC18 Explorer.
4. (MRF24WB/G) Enhanced debug messages providing more details such as authentication or association failues, etc.
5. (MRF24WB/G) Added option to disable module FW connection manager by adding "#define DISABLE_MODULE _FW_CONNECT_MANAGER_IN_INFRASTRUCTURE".
6. (MRF24WB/G) Added option to derive real key from pass-phrase in host side by adding "#define DERIVE _KEY_FROM_PASSPHRASE_IN_HOST".
7. (MRF24WB/G) Enabled WEP security for EasyConfig Ad-hoc connection
8. (MRF24WB/G) Added support for MRF24WG0MA/MB, a superset of MRF24WB0MA/MB (MRF24WB). Required to add to project definition files (#define MRF24WG). Ranges are different from MRF24WB0MA/MB. i. Valid RSSI Range : 43 ~ 128 (max) (WF_ScanGetResult()) ii. Max transmit power range : 0 ~ 18 dbm (WF_TxPowerSetMax())

MRF24WG0MA/MB enhancements are above and beyond on those features listed above. a. Added WPS(1.0) security method (#define WF_SECURITY_WPS_PUSH_BUTTON & WF_SECURITY_WPS_PIN ) (TCPIP demo + console demo). b. Added Wi-Fi Direct (P2P) function in the GC (group client) role (#define MY_DEFAULT_NETWORK_TYPE CFG_WF_P2P) (TCPIP demo + console demo) c. Added simplified basic SoftAP functionality (WF_SOFT_AP). Current features are 1 client STA, open security and no routing. (Easy config demo) d. Selection of support for 16 (max) software multicast filters (ENABLE_SOFTWARE_MULTICAST_FILTER ) or 2 hardware multicast filters. RTS or ES release requires software patch. 

Fixes:

1. The BigInt_helper.S file will now correctly include the processor include files for the PIC24E and dsPIC33E architectures.
2. Fixed bug in TCP state machine that corrupted the sequence number when client socket is closed.
3. Fixed bug in BerkeleiAPI.c::sendto() that checked for the wrong UDP socket to be opened.
4. (All parts including RTS or ES release) MRF24WG0MA/MB requires PLL work-around initialization code. Please refer to July/2012 release.
5. (All parts including RTS or ES release) MRF24WG0MA/MB needs to update stack from July/2012 release to cater for SSL.
6. (MRF24WB/G) Fixed stack issues encountered in Stack version v2012-04-03 and older. Code fix is needed in WF_CPSetElements() API. Modification is as below.

void WF_CPSetElements(UINT8 CpId, tWFCPElements *p_elements) { UINT8 elements[sizeof(tWFCPElements) + 2]; WF_ASSERT(p_elements->ssidLength <= WF_MAX_SSID_LENGTH); memset(elements, 0, sizeof(elements)); memcpy(elements, p_elements, sizeof(*p_elements)); LowLevel_CPSetElement(CpId, /* CP ID */ WF_CP_ELEMENT_ALL, /* Element ID */ (UINT8 *)elements, /* pointer to element data */ sizeof(elements)); /* number of element data bytes */ } 

Code fix is also needed in WaitForMgmtResponse() API (WFMgmtMsg.c). Refer to description of this fix is in the Supplementary TCPIP Help folder. 

Known Problems:

1. Telnet server module does not implement a lot of Telnet functions. As a result, it will likely not display correctly or work at all with some Telnet clients. The server was tested with the Microsoft telnet.exe utility which is provided with Microsoft Windows.
2. If the DHCP client and DHCP server are used at the same time and you connect two similar boards to each other (ex: two PICDEM.net 2 boards connected via a crossover cable), a race condition can occur where both nodes will disable their DHCP server and neither board will get a successful DHCP lease. If this unlikely scenario occurs, as a work around, simply reset one of the boards to reenable it's DHCP server.
3. LCDBlocking.c timing for the LCD_E_IO enable signal is too fast to meet published data sheet limits for various LCD controllers when compiling for a PIC32 running at > 43MHz. Despite this potential timing violation, the LCD does normally work correctly on the precompiled PIC32 demos for the Explorer 16 at 80MHz.
4. For PIC32 implementations, depending on the configuration, it is possible that the MACGetFreeRxSize() returns a value greater than 64 KB. For backward compatibility reasons the stack uses a 16 bit value to check the returned value and it won't work corretly.
5. Limiting the number of UDP sockets to 8 in the stack demos may prevent SNMP trap functionality. If this occurs, you can increase the MAX_UDP_SOCKETS definition in TCPIPConfig.h to 10 (if your system will support the increased data memory usage) to fix this issue.
6. The SNMP mib file's date and version parameter does not match the date/version of the current stack release.
7. (MRF24WG0MA/MB) RTS or ES release requires software patch for multicast filters.
8. (MRF24WG0MA/MB) Under softap option with easy config demo, unable to connect to infrastructure mode AP.
9. Certain APs have power save issue with reconnect. Please refer to AP compatibility list.

v5.41.02 Apr 2012

Fixes:

1. Added the Bigint helper libraries for C30 into the stack source folder to prevent linker errors in 16-bit projects that include those libraries.

Known Problems:

1. Telnet server module does not implement a lot of Telnet functions. As a result, it will likely not display correctly or work at all with some Telnet clients. The server was tested with the Microsoft telnet.exe utility which is provided with Microsoft Windows.
2. If the DHCP client and DHCP server are used at the same time and you connect two similar boards to each other (ex: two PICDEM.net 2 boards connected via a crossover cable), a race condition can occur where both nodes will disable their DHCP server and neither board will get a successful DHCP lease. If this unlikely scenario occurs, as a work around, simply reset one of the boards to reenable it's DHCP server.
3. LCDBlocking.c timing for the LCD_E_IO enable signal is too fast to meet published data sheet limits for various LCD controllers when compiling for a PIC32 running at > 43MHz. Despite this potential timing violation, the LCD does normally work correctly on the precompiled PIC32 demos for the Explorer 16 at 80MHz.
4. For PIC32 implementations, depending on the configuration, it is possible that the MACGetFreeRxSize() returns a value greater than 64 KB. For backward compatibility reasons the stack uses a 16 bit value to check the returned value and it won't work corretly.
5. Limiting the number of UDP sockets to 8 in the stack demos may prevent SNMP trap functionality. If this occurs, you can increase the MAX_UDP_SOCKETS definition in TCPIPConfig.h to 10 (if your system will support the increased data memory usage) to fix this issue.
6. The SNMP mib file's date and version parameter does not match the date/version of the current stack release.

v5.41 Feb 2012

Changes:

1. Added socket number validation to TCP functions.
2. Removed the WiFi projects from the Energy Monitoring and WebVend demos.
3. Added the "-mperipheral-libs" linker option to C32 projects.
4. Moved AES library and header files used by SNMPv3 from the "Crypto" source and include folders to the to the "TCPIP Stack" source and include folders.
5. Changed SNMP MIB trap number from 7 to 0.
6. Changed SNMP MIB SYOBJECTID to 43.6.1.4.1.17095.1 (added .1).
7. Updated MPFS2 Utility to make path settings persistent.
8. Changed default channel list from {1,6,11} to {1,2,3,4,5,6,7,8,9,10,11} for FCC domain
9. Added support for WEP with Shared Key
10. Added new command "iwconfig scan" & "iwconfig scanresults"
11. Added gRFModuleVer1209orLater" flag to identify version 1209 and later specific API's.
12. Added the following 1209 and later specific API: a) WF_CPSetWepKeyType() b) WF_CMGetConnectContext() c) WFEnableBroadcastProbeResponse() d) WFEnableAggressivePowerSave() e) WF_CPSetSsidType() f) WFEnableDeferredPowerSave() g) WF_FixTxRateWithMaxPower()

Fixes:

1. Updated MPFS2 Utility path strings for Mac/Linux compatibility.
2. Iperf - Resolved application crash due to network disconnection.
3. Host Scan - More robust and prevent system hang up. #define WF_HOST_SCAN.
4. Power Save - Improvements with Aggressive PS mode, and better handling of PS in host code.
5. Host connect & disconnect - Prevent system hang.
6. DHCP Refresh - Better handling of DHCP session in PS mode, and issue DHCP renewal any STA reconnect with AP.

Known Problems:

1. Telnet server module does not implement a lot of Telnet functions. As a result, it will likely not display correctly or work at all with some Telnet clients. The server was tested with the Microsoft telnet.exe utility which is provided with Microsoft Windows.
2. If the DHCP client and DHCP server are used at the same time and you connect two similar boards to each other (ex: two PICDEM.net 2 boards connected via a crossover cable), a race condition can occur where both nodes will disable their DHCP server and neither board will get a successful DHCP lease. If this unlikely scenario occurs, as a work around, simply reset one of the boards to reenable it's DHCP server.
3. LCDBlocking.c timing for the LCD_E_IO enable signal is too fast to meet published data sheet limits for various LCD controllers when compiling for a PIC32 running at > 43MHz. Despite this potential timing violation, the LCD does normally work correctly on the precompiled PIC32 demos for the Explorer 16 at 80MHz.
4. For PIC32 implementations, depending on the configuration, it is possible that the MACGetFreeRxSize() returns a value greater than 64 KB. For backward compatibility reasons the stack uses a 16 bit value to check the returned value and it won't work corretly.
5. Limiting the number of UDP sockets to 8 in the stack demos may prevent SNMP trap functionality. If this occurs, you can increase the MAX_UDP_SOCKETS definition in TCPIPConfig.h to 10 (if your system will support the increased data memory usage) to fix this issue.
6. The SNMP mib file's date and version parameter does not match the date/version of the current stack release.

v5.36.4 Oct 2011

Changes:

1. SNMPNotify() is updated to support ASCII string varible type for both TRAPv1 and TRAPv2.ASCII string address pointer is assigned to argument val(SNMP_VAL) of SNMPNotify().
2. The SSL module has been updated to support 1024-bit RSA key lengths for server and client on all architectures. PIC32 microcontrollers now support client/server RSA key lengths of 2048 bits. NOTE: To support these changes, you must manually modify your copy of RSA.c. A description of the required changes ("Required SSL Changes.pdf") can be found in your Microchip Applications Libraries installation directory in the "MicrochipHelpSupplementary TCPIP Help" subdirectory.

Fixes:

1. SNMP local variable community length increased with plus one.SNMP warnings has been removed for the compiler version C32 2.01 for zero optimisation.
2. Updated MPFS2.jar and mib2bib.jar to support Java version 1.7.
3. Fixed MPFS2.jar offset issue for fileRcrd.bin and dynRcrd.bin file and it was due to the file which has zero dynamic variable.Fixed Crimson editor problem with webPage2 folder where user couldn't save files using Crimson Editor if the WebPages2 folder that contained those files was selected in the MPFS2 utility.
4. MPFS2.jar file was getting hanged for the zero file size access.Now Zero file size also is the part of the respective generated files.

Known Problems:

1. Telnet server module does not implement a lot of Telnet functions. As a result, it will likely not display correctly or work at all with some Telnet clients. The server was tested with the Microsoft telnet.exe utility which is provided with Microsoft Windows.
2. If the DHCP client and DHCP server are used at the same time and you connect two similar boards to each other (ex: two PICDEM.net 2 boards connected via a crossover cable), a race condition can occur where both nodes will disable their DHCP server and neither board will get a successful DHCP lease. If this unlikely scenario occurs, as a work around, simply reset one of the boards to reenable it's DHCP server.
3. LCDBlocking.c timing for the LCD_E_IO enable signal is too fast to meet published data sheet limits for various LCD controllers when compiling for a PIC32 running at > 43MHz. Despite this potential timing violation, the LCD does normally work correctly on the precompiled PIC32 demos for the Explorer 16 at 80MHz.
4. For PIC32 implementations, depending on the configuration, it is possible that the MACGetFreeRxSize() returns a value greater than 64 KB. For backward compatibility reasons the stack uses a 16 bit value to check the returned value and it won't work corretly.
5. Limiting the number of UDP sockets to 8 in the stack demos may prevent SNMP trap functionality. If this occurs, you can increase the MAX_UDP_SOCKETS definition in TCPIPConfig.h to 10 (if your system will support the increased data memory usage) to fix this issue.
6. The SNMP mib file's date and version parameter does not match the date/version of the current stack release.

v5.36.2 July 2011

Changes:

1. Removed the Google PowerMeter and Google PowerMeter EZConfig demos. Google, Inc. has deprecated Google PowerMeter and has expressed the intent to remove access to it on September 16, 2011. To obtain Microchip Technology's Google PowerMeter reference implementation, you can download the June 2011 Microchip Application Libraries archived release from www.microchip.com/mal.
2. Modified the Energy Monitoring demo to remove Google PowerMeter functionality. The demo will still display measured power data on its internal web page.
3. Updated the TCP/IP Stack Performance table to use the testing methodology from previous releases. More information is available in the TCP/IP Stack Help file.
4. gSnmpNonMibRecInfo[] has been moved from snmp.c file to CustomSNMPApp.c file and SNMP_MAX_NON_REC_ID_OID macro has been moved from snmp.h file to CustomSNMPApp.c file. gSnmpNonMibRecInfo[] is used to list the static variables parent OID strings which are not part of mib.h file. This structure is used to restrict the access to the SNMPv3 objects from SNMPv2c and SNMPv1 version requests. Macro STACK_USE_SMIV2 is used to support gSnmpNonMibRecInfo[] with MODULE-IDENTITY number. For V5.31 STACK_USE_SMIV2 need to commented.
5. Removed the SPI2CON register freeze-on-halt bit macro from the SPIFlash, RAM, and EEPROM driver files to provide compatibility with C32 v2.00.

Fixes:

1. Removed the MPFSImg2 files from the MPLAB X C18/C30 projects so that the projects will compile. Disabled MPFSImg2.c for PIC32 Explorer 16 projects.
2. Added a heap and minimum stack size for the PIC32 Ethernet Starter Kit MPLAB X project.
3. The TCP/IP Stack Help File's performance table has been updated using the same test procedure used in previous releases.

Known Problems:

1. Telnet server module does not implement a lot of Telnet functions. As a result, it will likely not display correctly or work at all with some Telnet clients. The server was tested with the Microsoft telnet.exe utility which is provided with Microsoft Windows.
2. If the DHCP client and DHCP server are used at the same time and you connect two similar boards to each other (ex: two PICDEM.net 2 boards connected via a crossover cable), a race condition can occur where both nodes will disable their DHCP server and neither board will get a successful DHCP lease. If this unlikely scenario occurs, as a work around, simply reset one of the boards to reenable it's DHCP server.
3. LCDBlocking.c timing for the LCD_E_IO enable signal is too fast to meet published data sheet limits for various LCD controllers when compiling for a PIC32 running at > 43MHz. Despite this potential timing violation, the LCD does normally work correctly on the precompiled PIC32 demos for the Explorer 16 at 80MHz.
4. For PIC32 implementations, depending on the configuration, it is possible that the MACGetFreeRxSize() returns a value greater than 64 KB. For backward compatibility reasons the stack uses a 16 bit value to check the returned value and it won't work corretly.
5. Limiting the number of UDP sockets to 8 in the stack demos may prevent SNMP trap functionality. If this occurs, you can increase the MAX_UDP_SOCKETS definition in TCPIPConfig.h to 10 (if your system will support the increased data memory usage) to fix this issue.
6. The SNMP mib file's date and version parameter does not match the date/version of the current stack release.

v5.36 2 June 2011

Changes:

1. Because of changes to the SHOUTcast server configuration, the Internet Radio demo is no longer functional. This demo has been retained in the stack distribution to provide a TCP/IP code example.
2. The UDP module will now perform address resolution and DNS querying automatically. As a result, the UDP module APIs have changed. The UDPOpenEx function provides this additional functionality. Please consult the TCP/IP Stack Help File's "Stack API > UDP" topic for more information.
3. The UDPOpen macro has been added to conform to the legacy UDPOpen interface.
4. The Announce, BerkeleyAPI, DHCP client/server, DNS client/server, NBNS, Reboot, SNMP, SNTP, TFTPc, UDPPerformanceTest, and ZeroConf modules have been updated to use the new UDP API. The iPerf demo has also been updated.
5. The MPFS Classic and HTTP(1) modules have been removed from the stack. Function- ality to support these modules has also been removed from the TCP/IP Stack software tools. MPFS2 and HTTP2 are still supported.
6. The UARTConfig demo module has been updated to upload MPFS2 images to the demo board in place of MPFS Classic images.
7. To facilitate linking on PIC18 platforms, the number of UDP sockets in demo projects has been reduced from 10 to 8.
8. The SNMP Stack application and mib2bib.jar PC utility now both support 1024 dynamic IDs.
9. SNMP_DEMO_TRAP is a new dynamic variable added to the snmp.mib file to support SMIv2 with TRAPv2. This will correct a previously existing issue viewing traps with the iReasoning MIB browser. As per those changes, the mchp.mib file has been modified to support the SMIv2 standard. This mib includes MODULE-IDENTITY which will provide MICROCHIP and MIB information. snmp.mib also includes MODULE-IDENTIY(1), a new number (1) to the existing OID after ENTERPRISE-ID(17095).
10. Added a preprocessor check that will include the ultoa function if a version of the C32 compiler earlier than 1.12 is used.
11. Modified the WiFi module to use separate retry counters for AdHoc and Infrastructure modes.
12. Modified Berkeley API module to accept IPPROTO_IP as a valid protocol for the socket function. The code will determine the protocol type from the socket type (datagram or stream).
13. Created MPLAB X projects corresponding to most MPLAB 8 projects and configurations. These projects are located in the MPLAB.X subfolder in the associated demo project directory. The MPLAB X import wizard can be used to create MPLAB X projects from MPLAB 8 projects that don't have an analogue in the new demo project folders.
14. Added project support for the dsPIC33E and PIC24E architectures.
15. All TCP/IP Stack demo projects have been moved to the "TCPIP" subdirectory in the stack installation directory.
16. Created Java versions of several TCP/IP tools to provide cross-platform support. The TCP/IP Configuration Wizard has not been ported to Java; thus, it is only available for Windows users.
17. To prevent issues with path length, MPLAB 8 project names have been changed. A list of the abbreviations used in the project names is available in the MAL help folder (Microchip Solutions/Microchip/Help/Abbreviations.htm). The names of the HardwareProfile and TCPIPConfig configuration files have been abbreviated as well.
18. Changed the configuration inclusion macros used by the TCP/IP Stack demo projects to match the terms used in the project/configuration names.
19. The "Alternative Configurations" subfolders in most demo projects has been renamed to "Configs."
20. Added a Google PowerMeter demo for use with the PIC18F87J72 Energy Monitoring PICtail.
21. The Web Preview tool is no longer included with the stack.

Fixes:

1. Fixed a DHCP Server (DHCPs.c) lease leak problem that would occur when STACK_USE_ZEROCONF_LINK_LOCAL was defined. This problem would have resulted in the DHCP server stop giving out any leases until being rebooted.
2. Updated the PIC32MX6XX/7XX external PHY SMSC 8720LAN reference design.
3. Fixed bug with window expecting MACGetFreeRxSize() to return values < 32KB.
4. Fixed a type casting bug with the CalcIPChecksum function that would cause an incorrect TX checksum if the checksum value overflowed again after adding the carry bits to the checksum value.
5. Fixed a bug in the AutoIP module that may have prevented the module from correctly defending its own address.
6. Added a check to the Announce module to ensure the MAC layer is linked before attemping to transmit an Announce message.
7. Fixed a bug in the ETH97J60 MACPut function.
8. Added an additional preprocessor check in a debug menu setting in WF_Spi.c to prevent a build error.
9. Added a fix to the Google PowerMeter demo code to restore SNTP timestamp sourcing if SNTP is enabled. Previously, it would be overwritten by a possibly invalid HTTP timestamp.

Known Problems:

1. Telnet server module does not implement a lot of Telnet functions. As a result, it will likely not display correctly or work at all with some Telnet clients. The server was tested with the Microsoft telnet.exe utility which is provided with Microsoft Windows.
2. If the DHCP client and DHCP server are used at the same time and you connect two similar boards to each other (ex: two PICDEM.net 2 boards connected via a crossover cable), a race condition can occur where both nodes will disable their DHCP server and neither board will get a successful DHCP lease. If this unlikely scenario occurs, as a work around, simply reset one of the boards to reenable it's DHCP server.
3. LCDBlocking.c timing for the LCD_E_IO enable signal is too fast to meet published data sheet limits for various LCD controllers when compiling for a PIC32 running at > 43MHz. Despite this potential timing violation, the LCD does normally work correctly on the precompiled PIC32 demos for the Explorer 16 at 80MHz.
4. For PIC32 implementations, depending on the configuration, it is possible that the MACGetFreeRxSize() returns a value greater than 64 KB. For backward compatibility reasons the stack uses a 16 bit value to check the returned value and it won't work corretly.
5. Limiting the number of UDP sockets to 8 in the stack demos may prevent SNMP trap functionality. If this occurs, you can increase the MAX_UDP_SOCKETS definition in TCPIPConfig.h to 10 (if your system will support the increased data memory usage) to fix this issue.
6. The SNMP mib file's date and version parameter does not match the date/version of the current stack release.

v5.31 19 October 2010

Changes:

1. Reorganized demo projects. The TCPIP ENCX24J600 Demo App, TCPIP PIC32 ETH Demo App, and TCPIP WiFi Demo App projects in stack versions 5.25 and prior have all been merged into the TCPIP Demo App folder. All four of these projects were almost identical to each other, with the primary difference being the network interface controller the projects were preconfigured to support. In this 5.31 TCP/IP Stack release, each hardware combination now has its own MPLAB IDE project in the TCPIP Demo App folder.
2. Reorganized HardwareProfile.h and TCPIPConfig.h structure for the TCPIP Demo App, TCPIP Google PowerMeter Demo App, and TCPIP WebVend App projects. Now, instead of having one massive HardwareProfile.h file that supports a multitude of hardware platforms simultaneously, simple individual hardware profiles have been created for specific hardware combinations and placed in the "Alternative Configurations" sub-folder. The correct hardware profile or TCPIP config file is selected by #including "HardwareProfile.h" or "TCPIPConfig.h" as in previous stack versions. However, the active hardware profile or config file from the Alternative Configurations folder is selected by a preprocessor macro defined in the MPLAB project settings (passed on the command line to the compiler).
3. Added HTTP_SAVE_CONTEXT_IN_PIC_RAM option to TCPIPConfig.h (HTTP2 server module). This option, when enabled, will increase HTTP2 server performance when servicing multiple simultaneous connections at the expense of using more PIC RAM.
4. Added automatic TPIN+/- polarity swapping logic to ETH97J60.c driver for PIC18F97J60 family devices. Some 3rd party Ethernet switches, routers, and end devices have their TX+/- pins wired backwards such that the remote TX+ signal connects to the PIC TPIN- pin and the remote TX- signal connects to the PIC TPIN+ pin. Because 10BaseT Ethernet signaling is polarized and the PIC18F97J60 family does not implement an auto-polarity feature, this normally prevents all RX communications with these non-IEEE 802.3 compliant link partners. To work around this incompatibility, it is possible to implement circuitry to swap the RX+ and RX- signals before reaching the TPIN+ and TPIN- pins. The PICDEM.net 2 Rev 6 reference design includes this necessary circuitry (U6, U7, R54, and RX_SWAP GPIO output pin from PIC). This stack version automatically controls the RX_SWAP signal based on the ETH_RX_POLARITY_SWAP_TRIS and ETH_RX_POLARITY_SWAP_IO definitions in HardwareProfile.h. If these macros are undefined, then the automatic polarity swapping feature is disabled in the ETH97J60.c driver.
5. Added portable LFSRRand() and LFSRSeedRand() public functions to Helpers.c and removed all references to C rand() and srand() functions. The C rand() function returns a 15 bit integer on 8 and 16 bit PICs, but a 31 bit integer on PIC32s. The LFSRRand() function returns a 16-bit integer, regardless of which PIC you are using.
6. Added support for various SST/Microchip brand SST25xxxxx SPI Flash chips to SPIFlash.c driver. Previously only devices supporting the Auto Address Increment (AAI) Word Program opcode (0xAD) would work. Now, devices such as the SST2525VF010A should work, which require the AAI Byte program opcode (0xAF) instead.
7. Removed support for Spansion brand SPI Flash chips in the SPIFlash.c driver. If your application is already using one of these devices, continue to use the SPIFlash.c/.h files from TCP/IP Stack 5.25. These older files are API compatible with the current version, so can be dropped in by simply overwriting the SPIFlash.c and SPIFlash.h files.
8. Removed a -4 offset from the advirtised TCP Maximum Segment Size option (MSS) and TCP_MAX_SEG_SIZE_RX configuration value in TCP.c. The default TCP MSS advertised is now 536 instead of 532.
9. For Wi-Fi projects in the TCPIP Demo App folder, changed MY_DEFAULT_LIST_RETRY_COUNT to WF_RETRY_FOREVER instead of 3. This changes default connection behavior to keep trying to connect instead of just trying 3 times which makes more sense for demonstration.
10. Changed WF_Connect() beacon timeout to 40.
11. IFConfig command in TCPIP WiFi Console Demo App modified to return application-perspective MAC address from the AppConfig structure, and not the Wi-Fi serialized MAC address (they may not match if user desired custom MAC).
12. Updated the TCP/IP Configuration Wizard. The user can now configure wireless settings and stack settings seperately. Because of the changes to the TCPIPConfig.h file, the user must now select the specific copy of TCPIPConfig.h (or any of its variants) instead of selecting a project directory. Added the ability to select WF_RETRY_FOREVER in the Wi-Fi configuration settings. Added a selection parameter for BSD socket count. Added validation to check for the proper number of Berkeley sockets and TCP performance test sockets in the socket configuration screen (Advanced Settings) if either of these features are enabled. Added the ability to create sockets of the same type with different TX/RX buffer sizes in the socket configuration screen.
13. Updated the TCPIP WebVend Demo App to support Wi-Fi in several configurations.
14. Modified the Google PowerMeter demo to automatically determine the date/time from the HTTP module if the date/time cannot be obtained from the SNTP module.
15. Added a new Google Map project example to the Combo Demos folder. This example runs on a PIC24FJ256DA210 Development Board + Fast 100Mbps Ethernet PICtail Plus (or Ethernet PICtail Plus) + Truly 3.2" 240x320 display, TFT_G240320LTSW_118W_E (or Powertip 4.3" 480x272 display, PH480272T_005_I11Q). It also can run on the PIC32 Multimedia Expansion Board + PIC32 Ethernet Starter Kit. This demo connects to the Internet, sends an HTTP query for a specific map tile to the Google Static Maps API, and then displays the compressed tile to the graphics display. For more information, see the "Getting Started - Running the Graphics Google Map Demo.htm" file in the Combo DemosGoogle Map folder.
16. Added preliminary SNMPv3 module. This module, enabled with the STACK_USE_SNMPV3_SERVER option in TCPIPConfig.h, implements the Simple Network Management Protocol, version 3. Among other things, SNMPv3 adds secure authentication and cryptographic privacy as compared to SNMPv2C. This implementation currently only supports AES encryption (no DES support). It also has only been tested with the PIC32 Ethernet Starter Kit (TCPIP Demo App - C32 - PIC32_ENET_SK_DM320004_INTERNAL_ETHERNET.mcp MPLAB IDE project). SNMPv3 on PIC18, PIC24, and dsPIC platforms are not supported at this time. Because AES encryption has specialized United States export requirements, this TCP/IP Stack release does not include the required AES library to enable SNMPv3. To obtain the needed AES library, you must purchase SW300052 v2.6 or later. Older v2.5 and previous versions include AES related files on them, but do not include the new AES files required by SNMPv3. For more information on using SNMP, refer to the TCP/IP Stack Help (Demo Information -> Available Demos -> TCPIP Demo App -> Demo Modules -> Network Management (SNMP) Server).
17. Altered the SaveAppConfig() function in MainDemo.c to store a more robust signature to EEPROM/SPI Flash when saving the AppConfig structure. In v5.25 and prior stack versions, when EEPROM or SPI Flash memory was available, the stack would automatically write a one byte marker character to address 0x000000 in the EEPROM/Flash indicating if a valid AppConfig structure was stored in the non-volatile memory. This resulted in the EEPROM/Flash contents being organized like the following: Address Data Contents =================== ============================================== 0x000000: Marker Byte 0x000001: AppConfig structure MPFS_RESERVE_BLOCK: Start of MPFS/MPFS2 image containing web pages In this stack version, EEPOM/Flash contents will now contain: Address Data Contents =================== ============================================== 0x000000: Length of AppConfig structure (16-bit integer) 0x000002: IP checksum of the AppConfig default values, as defined in TCPIPConfig.h and WF_Config.h (16-bit integer). 0x000004: IP checksum of the subsequent EEPROM/Flash bytes of the AppConfig values. 0x000006: AppConfig structure MPFS_RESERVE_BLOCK: Start of MPFS/MPFS2 image containing web pages

The additional checkums allow automatic detection to occur if you change one of the values in TCPIPConfig.h or WF_Config.h that affects AppConfig. If you change one of the values in code, then upon boot up, the application will automatically detect this change and start using the values that you selected in code. If, at run time, you decide to change the AppConfig values and commit the changes to EEPROM/Flash, then the stack will subsequently use the run-time saved values on future reboots. The checksum at offset 0x000004 ensures that if any corrupted AppConfig contents are found in EEPROM/Flash (ex: power is lost between writing the signature structure and actual AppConfig structure, or code unintentionally overwrites something in the AppConfig memory area), then the original defaults defined in TCPIPConfig.h and WF_Config.h will be used instead of the corrupted values. This EEPROM/SPI Flash change affects all projects except TCPIP Internet Radio App, TCPIP Internet Bootloader App, and all PIC32 Starter Kit projects since these projects do not have or use external EEPROM or SPI Flash memory. 

Fixes:

1. Fixed a UDP bug in which a transmitted packet would have been addressed to the wrong destination node if the UDP socket received a broadcast packet from a different remote node from the last received packet, but using the same source port number as the last received packet. The FindMatchingSocket() function in UDP.c will now always change the local socket parameters to send to the most recent remote node's unicast IP address, regardless of if the last received packet was addressed to a multicast or broadcast destination. Thanks go to Billy Walton for reporting this erroneous behavior. If you wish to change the destination IP/MAC addresses or port number for a UDP packet that you are ready to send, write the new parameters to the UDPSocketInfo[SocketHandle] global structure before calling UDPFlush(). This structure contains remoteNode and remotePort parameters for the remote IP address/MAC address and remote UDP port, respectively. You can also read these values to obtain the remote addressing parameters for the last received packet on the given UDP socket. Note that "SocketHandle" refers to the UDP socket handle returned by the UDPOpen() API, not the literal string "SocketHandle".
2. Fixed ADC state save/restore bug in GenerateRandomDWORD() function in Helpers.c. PIC24, dsPIC, and PIC32 platforms require the ADC ON/ADON bit to be cleared before modifying certain other ADC register contents.
3. Fixed an ENC28J60 MAC/MII register write timing violation when using a PIC24H or dsPIC at over 33MIPS. There was inadequate Chip Select hold time provided, violating the 210ns minimum specified in the ENC28J60 data sheet. This violation may have resulted in certain devices losing the ability to receive packets (due to the MARXEN bit, MACON1<0>, getting cleared unintentionally).
4. Fixed an ENCX24J600.c driver bug in which operating at 100Mbps with the ENC424J600/624J600 Ethernet controller, it would be possible for the MACGetHeader() function to issue a Reset() operation under rare circumstances. The PIC would reset whenever the PHY detected an illegal symbol during 4B5B decoding but guessed the correct 4B symbol such that no data corruption or CRC error occurred. This condition results in a valid packet being received but with the Received Ok Receive Status Vector bit being clear (RSV<23> == 0). This issue would become more probable when using very long Ethernet cables (ex: 100 meters) and receiving a lot of data.
5. Fixed a TCP bug in which calling TCPDisconnect() to close a connection when the remote node's RX window was 0 bytes would have caused the stack to enter an infinite loop sending duplicate ACK packets.
6. Fixed Wi-Fi bug that caused assert condition if too many management messages were being received during data traffic.
7. Fixed Wi-Fi bug that caused WF_EVENT_CONNECTION_RESTABLISHED event case to send the wrong notification to the app.
8. Fixed Wi-Fi bug that caused assert failure with Scratch move failure.
9. Fixed Wi-Fi bug in WF_CAGetChannelList() and WF_CAGetSecurity that caused failure.
10. Fixed Wi-Fi EasyConfig bug that required development boards to be manually reset even after new network was selected.
11. Fixed MRF24WB0 bug that caused assert if invalid WPA/WPA2 key was entered.
12. Fixed Wi-Fi power management bit behavior in PS-Poll frame that was causing some AP’s to never send data or disconnect when in power save mode.
13. Fixed a TCP bug in which attempting to open a client TCP socket to a remote server, specified by IP address (not DNS address), that was offline, but who's MAC address was already cached by the ARP client, would result in endless back-offs. For example, when attempting to contact the remote node (that was not responding), the TCP module would have transmitted a SYN at time T=0 seconds, T=1s, T=3s, T=7s, T=15s, T=31s, T=63s, T=127s, T=255s, etc. The exponential back-off between retransmissions would grow indefinitely until the retransmission interval would have grown so large that effectively no-retransmissions would be occurring. Assuming the application wasn't written with its own timeout to prevent endless waiting, this would prevent the socket from automatically establishing the connection to the remote server once the server came back online. With this TCP fix, the exponential back off now saturates after TCP_MAX_RETRIES (5) back offs and continues to retransmit using the same interval. By default, this means SYN transmissions will occur at T=0 seconds, T=1s, T=3s, T=7s, T=15s, T=31s, T=63s, T=95s, T=127s, etc. After 5 back-offs the retransmission interval stops growing and stays constant at 32 seconds.
14. Fixed an RSA computation bug that would cause the RSA module to never complete if you attempted to compute y = x^e % n where e = 3 (or similar number < 256 with only 0, 1, or 2 bits set). Thanks go to Kevin Maidment for pointing this error out and suggesting a solution. Note, that this fix to RSA.c is not distributed with the ordinary TCP/IP Stack due to United States export restrictions. To get this fix, you must repurchase SW300052. This fix is included in SW300052 v2.6 or later. If you don't have CD media to identify the SW300052 version that you have, you can test the RSA.c file that you have. RSA.c in SW300052 v2.6 has a CRC32 checksum of 0x91F66711. RSA.c in v2.5 and prior had a checksum of 0xB1E8B0CC.

Known Problems:

1. Telnet server module does not implement a lot of Telnet functions. As a result, it will likely not display correctly or work at all with some Telnet clients. The server was tested with the Microsoft telnet.exe utility which is provided with Microsoft Windows.
2. If the DHCP client and DHCP server are used at the same time and you connect two similar boards to each other (ex: two PICDEM.net 2 boards connected via a crossover cable), a race condition can occur where both nodes will disable their DHCP server and neither board will get a successful DHCP lease. If this unlikely scenario occurs, as a work around, simply reset one of the boards to reenable it's DHCP server.
3. LCDBlocking.c timing for the LCD_E_IO enable signal is too fast to meet published data sheet limits for various LCD controllers when compiling for a PIC32 running at > 43MHz. Despite this potential timing violation, the LCD does normally work correctly on the precompiled PIC32 demos for the Explorer 16 at 80MHz.

v5.25 07 May 2010

Changes:

1. Added support for the Microchip MRF24WB0 802.11 WiFi controller (module part number MRF24WB0MA). This product is intended as a backwards compatible, drop in replacement to the ZG2100. The MRF24WB0MA should work with previous TCP/IP Stack versions as if it were a ZG2100M, but when the MRF24WB0MA is used with this (5.25) and future TCP/IP Stack versions, feature improvements inside the MRF24WB0 allow the TCP/IP Stack code/RAM size to be smaller and run faster.
2. Dropped support for the ZeroG ZG2100 802.11 WiFi controller. Applications that must stay with this device should continue to use TCP/IP Stack version 5.20b or earlier. All new projects or preexisting projects undergoing updates should be developed with the MRF24WB0 instead.
3. The WiFi connection management state machines now run on the MRF24WB0 instead of the PIC host, freeing up code and data space. Connection profiles can be created and the connection algorithm fine-tuned by the PIC application. In the WiFi demos see the WF_Connect function in MainDemo.c for an example of how to configure and then establish a WiFi connection. The programming model has changed to an API model which is documented in 'TCPIP Stack Help.chm'.
4. Changed "VERSION" macro definition in TCPIP.h to "TCPIP_STACK_VERSION". "VERSION" is overly generic and will likely conflict with other identical tokens one may use in their application code or source libraries.
5. Added support for the PIC24FJ256GA110, PIC24FJ256GB110 and PIC24FJ256GB210 PIMs for the Explorer 16. Note that when using the PIC24FJ256GA110 general purpose PIM, the Ethernet PICtail Plus, Fast 100Mbps Ethernet PICtail Plus, MRF24WB0MA Wi-Fi PICtail Plus, or other SPI PICtail daughter board should be installed in the middle SPI2 slot of the Explorer 16, not the ordinary topmost SPI1 slot used by other PIMs, including the PIC24FJ256GB110 and PIC24FJ256GB210 ones. The software is set up to use SPI2 for the PIC24FJ256GA110 PIM to avoid incompatibilities with silicon revision A3, which does not allow the SCK1 pin to be configured as a PPS output.
6. Added support for the PIC24FJ256DA210 Development Board.
7. Added TFTPUploadRAMFileToHost(), TFTPUploadFragmentedRAMFileToHost() and TFTPGetUploadStatus() APIs to the TFTPc.c file. These APIs provide a very simple means of uploading a whole file to a remote TFTP server without requiring a great deal of application state machine logic. These APIs require the DNS client module to be enabled (STACK_USE_DNS must be defined, in addition to STACK_USE_TFTP_CLIENT).
8. Added a dummy DNS Server module. This server always sends the local IP address in response to all queries received. When using the PIC DHCP server, its purpose is to allow a user to type anything into a web browser (ex: http://asdf/) and still receive the web page provided by the PIC, much as a hotel or airport WiFi router will serve to you before you've paid or agreed to the network's terms of service. This DNS server module is implemented in DNSs.c, requires one UDP socket, and is enabled via the STACK_USE_DNS_SERVER option in TCPIPConfig.h.
9. Changed SPIFlash.h file defaults to target an SST SPI Flash with 4096 byte sectors instead of a Spansion Flash with 65536 byte sectors. These new defaults are, among other reasons, in support of the PIC24FJ256DA210 Development Board, which has an SST SST25VF016B on it.
10. Made TCP Keep-Alive packets consistently get sent TCP_KEEP_ALIVE_TIMEOUT (default 10 seconds) after the last socket TX or RX activity. In earlier stack versions, if the local node transmitted some data and then let the socket go idle, the first Keep-Alive packet sent would use the TCP_START_TIMEOUT_VAL (default 1 second) timer value before getting sent. While benign in terms of application behavior, these faster than normal keep-alive transmissions were distracting when viewed in Wireshark or other packet capture tools.
11. Disabled STACK_USE_DYNAMICDNS_CLIENT option in TCPIPConfig.h by default for the TCPIP Demo App and TCPIP ENCX24J600 Demo App projects. This option was enabled by default in earlier stack releases. This was done to save code size and allow out-of-box compilation on devices with 128KB of Flash when not using compiler optimizations. The TCPIP PIC32 ETH Demo App project continutes to have this option enabled by default.
12. Added SNMP v2 TRAP PDU format. Macro SNMP_STACK_USE_V2_TRAP is used to enable the SNMP v2 trap format. New API function SNMPV2TrapDemo() is included to support more than one variable binding to the SNMPv2 TRAP. This API can be used for a single SNMPv2 TRAP variable varbind and is part of CustomSNMPApp.c. A multiple variable binding demo can be enabled MainDemo.c. One should not enable both SNMPTrapDemo and SNMPV2TrapDemo simultaneously. Global flag "gSetTrapSendFlag" is used to indicate the start and end of SNMPv2 trap varbinds. If gSetTrapSendFlag is FALSE, then very next variable varbind for the SNMPv2 TRAP, is the last or only one variable varbind. If gSetTrapSendFlag is TRUE, then there is another variable varbind available to be part of the SNMPv2 TRAP PDU.
13. Added support for PIC32MX6XX/7XX external PHY's: SMSC 8700LAN and National DP83640.
14. Added schematics and BOM for the PIC32 Ethernet Starter Kit.
15. Added the Google PowerMeter demo project. Consult the "Reference Implementation for Google PowerMeter.chm" help file for more information.
16. Modified the SSL and TCP modules to create the TCPStartSSLClientEx function. This function will enable the SSL module to store supplementary data (currently only SSL Certificate Public Keys) in a structure.
17. Moved the HTTP_PORT, HTTPS_PORT, HTTP_MAX_DATA_LEN, and HTTP_MIN_CALLBACK_FREE macros from HTTP2.c to TCPIPConfig.h.

Fixes:

1. The SPIFlashEraseSector() function in the SPIFlash.c file incorrectly erased the sector specified by the current write pointer (set by calling SPIFlashBeginWrite()) instead of the specified dwAddr parameter address. This error had no impact on any TCP/IP Stack code as these parameters always matched. However, application code using the API would have been affected. Thanks go to Marc Boon for reporting this issue on the Microchip Ethernet forum.
2. Fixed ENC424J600/624J600 driver for PSP modes 2, 4, 6, and 10. The PIC's PMP PMMODE<9:8> bits were not set correctly.
3. Removed from lingering references to TickGetDiff() in FTP.c, TFTPc.c and UARTConfig.c.
4. Fixed DNS client module from returning the DNS server IP address if the DNS query failed due to a server error (i.e. DNS did respond, but did not return any records, such as when attempting to resolve a name that isn't in the DNS). DNSIsResolved() will now return 0.0.0.0 on any non-recoverable DNS error or timeout.
5. Fixed HTTP2 MPFS upload page being accessible from URLs that weren't an exact match. For example, in 5.20 and earlier, accessing http://mchpboard/mpfsuploadASDF would still have opened the mpfsupload page. Thanks go to Andrea Rivolta on the Microchip Ethernet Forum for identifying this error.
6. Improved UDP TX checksum code for the special case when the computed checksum was 0x0000. According to the UDP RFC, for this corner case, the checksum should be converted to 0xFFFF before transmission to differentiate from the checksum disabled case, improving error detection by a minuscule amount.
7. Fixed GetCLKOUT() function in ENCX24J600.c driver file. Previously, 0x00 would always be returned, regardless of the value in the COCON bits of ECON2. The function documentation for SetCLKOUT() and GetCLKOUT() was also corrected (had obsolete information ported over from ENC28J60 driver file).
8. Fixed DHCP client rebinding bug in which the DHCP client would request the wrong IP address if an unrelated DHCP OFFER or ACK message were received after we transmitted a DHCP REQUEST but before we received our DHCP ACK. Under rare conditions, this would have resulted in the TCP/IP stack reverting to the static or AutoIP assigned address for a few seconds between DHCP lease renewals.
9. Fixed TFTP Internet Bootloader bug in which uncommon .hex files containing a certain data pattern could not be uploaded correctly to the PIC18F97J60 family device. For these problem .hex files, a block of 32 program words (64 bytes) would remain unprogrammed (left as 0xFFFF) due to a parsing error in the bootloader's DecodeHex() function. The TFTP upload operation would succeed without reporting a programming error. The problem can be detected by using an ICD3 or similar ICSP programmer and reading the program Flash out of a device that is programmed with the bootloader and application .hex files. Compare the resulting memory dump to a device programmed only with the application .hex file. If you have devices deployed in the field with the previous bootloader and happen to generate a problem application .hex file, you can potentially work around the bootloader bug by opening the application .hex file with Notepad and appending dummy address records to the beginning to move the data around in the file. For example, at the very top of the .hex file, add lines containing ":020000040000FA" until the bootload process works correctly. You may alternatively try adding spaces at the end of any line, although this may make the .hex file incompatible with some programming utilities. Thanks go to Jonathan Seidmann for identifying and reporting this bug.
10. Fixed SNMPv2 TRAP format issue where SNMP browser was displaying all the SNMPv2 traps as SNMP version 1. SNMP v2 TRAP pdu format is rectified. Macro SNMP_STACK_USE_V2_TRAP is used to form and send a SNMPv2 TRAP PDU. SNMPTrapDemo API is used for both SNMPv1 and SNMPv2 single variable varbind trap.
11. Fixed an HTTP2.c server module initialization bug when using the PIC32MX7XX/6XX series internal Ethernet module. During initialization the HTTPLoadConn() function would overwrite over 100 bytes of PIC RAM past the end of the reserved memory allocated for the HTTP2 module. This problem would manifest itself by locking up the TCPIP PIC32 ETH Demo App-C32 demo shortly after power up if you compiled TCP/IP Stack version 5.20 with the MPLAB C Compiler for PIC32 MCUs (C32) version 1.11.
12. Fixed SSL client from incorrectly parsing for the server's public key in rare cases where the RSA Public Key Algorithm identifier was received, but the key hadn't been received by TCP yet. Thanks go to Kevin Maimdnet for identifing this error in SSL.c and reporting it via http://support.microchip.com/.
13. Fixed Tick.c TickGet(), TickGetDiv256() and TickGetDiv64K() APIs sometimes returning the wrong value on PIC32 platforms. On the PIC32MX3XX/4XX family devices a wrong return result would sometimes occur if using -O3 compiler optimizations (maximum speed) with the Microchip MPLAB C Compiler for PIC32 MCUs (C32). On the PIC32MX5XX/6XX/7XX family devices, such as the PIC32MX795F512L device used on the PIC32 Ethernet Starter Kit, wrong values could be returned, regardless of the compiler optimization level.

Known Problems:

1. Telnet server module does not implement a lot of Telnet functions. As a result, it will likely not display correctly or work at all with some Telnet clients. The server was tested with the Microsoft telnet.exe utility which is provided with Microsoft Windows.
2. If the DHCP client and DHCP server are used at the same time and you connect two similar boards to each other (ex: two PICDEM.net 2 boards connected via a crossover cable), a race condition can occur where both nodes will disable their DHCP server and neither board will get a successful DHCP lease. If this unlikely scenario occurs, as a work around, simply reset one of the boards to reenable it's DHCP server.
3. HI-TECH PICC-18 compilers are not supported in this release. The supplied HI-TECH PICC-18 MPLAB projects usually will not compile and/or link.
4. LCDBlocking.c timing for the LCD_E_IO enable signal is too fast to meet published data sheet limits for various LCD controllers when compiling for a PIC32 running at > 43MHz. Despite this potential timing violation, the LCD does normally work correctly on the precompiled PIC32 demos for the Explorer 16 at 80MHz.

v5.20 18 November 2009

Changes:

1. Added PIC32MX7XX/6XX Family integrated Ethernet controller support. The "TCPIP PIC32 ETH Demo App" folder was added to compile for the PIC32 Ethernet Starter Kit. Ethernet driver files "ETHPIC32ExtPhy.c" and "ETHPIC32IntMac.c" were added, in addition to the "ETHPIC32ExtPhyDP83848.c" file, which is a specific driver file for the National DP83848 10/100 PHY.
2. Added RFC 3927 Auto IP module. This module will automatically assign a local IP address to the node in the 169.254.xxx.xxx private address range (subnet mask 255.255.0.0) if a DHCP server is not present on the network or the DHCP client is disable. The exact IP address chosen will be pseudo-random, but as required by the protocol, it will perform gratuatous ARPs to avoid clobbering anyone else's IP address. Also, unless there is an address collision with a preexisting node on the network, the IP address generated by the Auto IP module will not change between power cycle events (random number generator is seeded by local MAC address). To enable this module, STACK_USE_AUTO_IP must be defined in TCPIPConfig.h. When compiled in, the module defaults to enabled, but will automatically yield to the DHCP client module, which has higher priority.
3. Added "TCPIP MDD Demo App" beta application projects. Projects in this folder store the HTTP2 web pages in external FAT16/FAT32 formatted SD card or USB Mass Storage media instead of an MPFS2 formatted EEPROM or SPI Flash. For more information on these projects, see the "Running the TCPIP MDD Demo App (Beta Release).pdf" file in the MicrochipHelp folder.
4. Expanded XEEReadArray() API's third length parameter from a BYTE to a WORD.
5. Converted all variable declarations and type casts of TICK data type to DWORD. The TICK typedef is now deprecated and will be removed in a future stack release. This data type conflicts with the TICK structure used in certain other Microchip software libraries.
6. Added TCP_WINDOW_UPDATE_TIMEOUT_VAL option to the TCP.c file (default 200ms). This timeout controls the time after calling TCPGet() or TCPGetArray() before the stack will transmit a RX window update to the remote node. Historically, the TCP_AUTO_TRANSMIT_TIMEOUT_VAL value was used for this purpose (default 40ms). This change decreases the net window update transmission overhead. If this adversely affects your application RX performance (unlikely, but possible for certain communications patterns), set TCP_WINDOW_UPDATE_TIMEOUT_VAL equal to or shorter than TCP_AUTO_TRANSMIT_TIMEOUT_VAL to get the same or better behavior relative to previous stack versions.
7. Split TCP_MAX_SEG_SIZE configuration constant in TCP.c into separate TCP_MAX_SEG_SIZE_TX and TCP_MAX_SEG_SIZE_RX configuration constants. Previously, TCP_MAX_SEG_SIZE was used to limit both the maximum size of transmit and receive packets. In cases where large TX FIFOs are allocated, and the remote node advirtises a large Maximum Segment Size TCP option, this change improves TCP transmit performance by roughly 10%.
8. Renamed "Internet Radio App", "Internet Bootloader App" and "WiFi Iperf App" folders to "TCPIP Internet Radio App", "TCPIP Internet Bootloader App" and "TCPIP WiFi Iperf App" respectively. These new names ensure consistent folder placement when viewing the Microchip Solutions folder with other Microchip Application Libraries installed.

Fixes:

1. Fixed SSL functionality (ex: HTTPS server) from failing when using the ENC424J600 and ENC624J600 Ethernet controllers. In stack versions 5.00 and 5.10, the BFSReg() and BFCReg() functions were being incorrectly used to set and clear CRYPTEN (EIR<15>). ENC424J600/624J600 silicon errata #6 on production silicon revision A2 prevents BFSReg() and BFCReg() from being able to modify CRYPTEN. This resulted in the SSL RSA encrypt/decrypt operations from ever finishing. The ENC424J600/624J600 errata #6 workaround is now implemented in the ToggleCRYPTEN() function in ENCX24J600.c.
2. Fixed an RSA padding error in the ENCX24J600.c's version of RSASetData() and RSASetE() functions. This fixes the Bad Record MAC problem when using SSL client APIs with the ENC424J600 and ENC624J600, as mentioned in the 5.10 stack release notes' Known Problems section. Although unknown at the time of release this problem also occurred in stack version 5.00.
3. Fixed DNS client from mishandling DNS responses that did not use name compression. Thanks go to Will Stone on the Microchip Ethernet forum for identifying this bug.
4. Fixed an ExtractURLFields() API bug which would incorrectly parse the URLs containing other URLs. Ex: "http://www.google.com/search?q=http://www.microchip.com/"
5. Fixed TickGet(), TickGetDiv256(), and TickGetDiv64K() APIs from potentially returning an incorrect time value (0x10000 ticks less than it should have) on rare occasions when using a PIC32 and with compiler optimizations turned on. The Tick.c module was also revised so that the IEC0 register does not get written to via a load-modify-store operation on PIC32s so that it is now possible for other application ISR functions to write to IEC0 without risking state corruption.
6. Fixed PIC32 Starter Kit Debugger losing access to the PIC32 target when the project was run. JTAG was being disabled at run time, but the PIC32 Starter Kit Debugger requires JTAG to communicate with the debug executive. JTAG is now conditionally disabled on PIC32s when the __MPLAB_DEBUGGER_PIC32MXSK macro is undefined.
7. Fixed a Berkeley sockets API bug in which calling closesocket() on a SOCK_STREAM type socket (TCP) did not actually close the socket if the remote node did not first send a FIN to the local node. This would leak a TCP socket each time the affected API calling sequence occurred and result in no FIN getting transmitted to the remote node.
8. Fixed an HTTP2 filename parsing bug that would occur when a web browser submitted a request for a file with hex encoded characters in it. For example, with stack version 5.10 and Firefox 3.5.3, typing "http://mchpboard/%70rotect" into the URL field would have resulted in an HTTP 404 not found error when "http://mchpboard/protect/index.htm" should have been returned instead. Thanks go to Steve Tuttle for reporting this issue and suggesting a solution.
9. Fixed a Berkeley sockets API bug in which calling recvfrom() on a datagram type socket (UDP) would return an incorrect remote IP address and port number when the from pointer was non-NULL.
10. Fixed HTTP2 server bug in which the HTTPReadPostName() function was failing to convert the field name from URL encoding to plain-text. If the browser posted, for example, a field named "Stock Remaining", it would have been incorrectly returned from HTTPReadPostName() as "Stock+Remaining".
11. In Stack 5.10, any new values you saved into AppConfig via the Network Configuration demo web page would have been mishandled for WiFi projects. HTTPPostConfig() in CustomHTTPApp.c of the TCPIP WiFi Demo App and TCPIP Iperf Demo App projects were corrected so that they now write a magic 0x61 marker into EEPROM/SPI Flash address 0x0000 to inidicate that the AppConfig structure is valid in EEPROM/SPI Flash. This prevents the InitAppConfig() function in MainDemo.c from restoring the default settings when changing the values through the Network Configuration page.
12. For WiFi projects, a Gratuitous ARP Work-around was implemented to work around cases where access points send broadcast messages at data rates that the ZG2100 cannot listen to. The define USE_GRATUITOUS_ARP (in TCPIPConfig.h) turns this feature on or off.

Known Problems:

1. Telnet server module does not implement a lot of Telnet functions. As a result, it will likely not display correctly or work at all with some Telnet clients. The server was tested with the Microsoft telnet.exe utility which is provided with Microsoft Windows.
2. If the DHCP client and DHCP server are used at the same time and you connect two similar boards to each other (ex: two PICDEM.net 2 boards connected via a crossover cable), a race condition can occur where both nodes will disable their DHCP server and neither board will get a successful DHCP lease. If this unlikely scenario occurs, as a work around, simply reset one of the boards to reenable it's DHCP server.
3. HI-TECH PICC-18 compilers are not supported in this release. The supplied HI-TECH PICC-18 MPLAB projects usually will not compile and/or link.
4. ENC624J600 PSP modes 2, 4, 6, and 10 do not work at this time. Some Parallel Bit Bang modes may not work either. Some minor firmware changes are needed.
5. TCPIP ENCX24J600 Demo App-C18.mcp project does not compile by default using MPLAB C Compiler for PIC18 MCUs (C18) version 3.34. There is not quite enough program memory available on the PIC18F97J60 for the large number of selected stack features to allow linking. To get this project to compile, turn on compiler optimizations or disable one of the modules in TCPIPConfig.h (ex: comment out STACK_USE_DYNAMICDNS_CLIENT).

v5.10 29 July 2009

Changes:

1. Added SSL capability to the Telnet server. If STACK_USE_SSL_SERVER is defined, the Telnet server will now listen on port 992 for SSL secured connections. If you do not have a telnets client, you can use an SSL proxy, such as stunnel (http://www.stunnel.org/) to add SSL/TLS support to any telnet client.
2. Moved a number of string pointer tables in the HTTP.c, HTTP2.c, FTP.c, and DynDNS.c files to allocate in ROM instead of RAM. This reduces around 120 bytes of RAM usage in the HTTP2 server when compiled for the PIC18 or PIC24/dsPIC platforms. The gains are even greater on PIC32 platforms.
3. Added redefinition of SPIRAM*(), SPIFlash*(), and XEEPROM*() functions so that when compiled and used without proper HardwareProfile.h definitions, a more descript linker error will be generated instead of a mysterious symbol not found error.
4. Added several new APIs:
   • ExtractURLFields() in Helpers.c. This function provides an easy means of parsing an URL string and extracting the protocol, hostname, port, file path, etc. Currently, this function is commented out to save code space as no stack modules require it. However, it should work correctly if you simply uncomment it (remove the #if 0...#endif around it).
   • strnchr() in Helpers.c. Finds the first occurrence of a character within a string but limited to a maximum length before giving up.
   • TCPPeek() and TCPPeekArray() in TCP.c. Reads from a TCP socket's RX FIFO buffer without removing the data from the stream.
   • TCPClose() in TCP.c. Disconnects a socket from the remote node (if connected) and closes the socket handle, including for server type sockets. This function is identical to the TCPDisconnect() API except for the handling of server sockets. TCPDisconnect() returns server sockets to the listning state and maintains the socket handle. TCPClose() closes the socket and frees all associated resources, invalidating the socket handle.
5. Updated the DHCP client module:
   • Modified so that it wouldn't attempt to transmit DHCP Discover packets when the MAC layer reports no link (MACIsLinked() == FALSE). This avoids main() while(1) loop performance degredation when you unplug the Ethernet cable or lose association to your access point.
   • Added capability of performing DHCP discovers and requests without setting the BOOTP broadcast flag. Now, the DHCP client module will start up and attempt to obtain an IP address with the broadcast flag set, but if it fails the next DHCP retry will attempt to obtain the IP address with the broadcast flag cleared. The flag will toggle back and fourth between unicast mode and broadcast mode if no DHCP server responds. This feature improves compatibility with certain DHCP servers and WiFi access points.
   • Added several new APIs including DHCPInit(), DHCPIsEnabled(), DHCPStateChanged(), DHCPIsBound(), and DHCPIsServerDetected().
   • Removed the DHCPFlags DHCP_CLIENT_FLAGS global variable. Use the above named APIs now to get equivalent functionality.
   • Removed the DHCPBindCount global variable. To detect if the DHCP state has changed, poll the new DHCPStateChanged() function.
   • Removed the DHCPReset() API. To perform this operation, now call the DHCPInit() API. Use 0x00 for the vInterface parameter.
6. Removed deprecated TickGetDiff() macro. To get a tick difference, just subtract the two values in-line. This macro was removed because it promoted confusing code. Ex: a-b is different from b-a. However, it was not contextually obvious which of the two was returned when TickGetDiff(a, b) was called.
7. Added PIC32MX460F512L USB and dsPIC33FJ256GP710 PIM support to the Explorer 16 hardware profile for the TCPIP WiFi Demo App and WiFi IPerf App projects.
8. Added all files needed for SSL (assuming the crypto libraries are present) to the TCPIP WiFi Demo App-C30 and TCPIP WiFi Demo App-C32 projects.
9. Converted TCPIP Demo App, TCPIP WebVend App, Internet Radio App, and Internet Bootloader App MPLAB Build Directory Policy to compile in the project folder instead of the source folder. This reduces the depedancies on the MPLAB project include path and allows new projects to be created by copying one of the pre-existing folders (ex: copy "TCPIP Demo App" to "My App") without having problems including the wrong HardwareProfile.h and TCPIPConfig.h files.
10. Changed EEPROM/SPI Flash AppConfig record valid flag from 0x60 to 0x61 in the TCPIP WiFi Demo App and WiFi Iperf App projects. This will force the various EEPROM settings to get erased when switching between Ethernet and WiFi projects. This is done since the AppConfig structure changes when using WiFi (SSID string is added).
11. The Wifi Iperf App and TCPIP WiFi Demo App projects have been optimized for better performance.

Fixes:

1. Fixed a TCPDisconnect() API bug in which the last few bytes of data (up to the TCP socket's TX FIFO size less 532 bytes) was not transmitted and no FIN was sent out if the TX FIFO was full of data when TCPDisconnect() was called. This problem could have only occurred for TCP sockets with a large TX FIFO (>=532 bytes). This problem could have been observed in stack version 5.00's "TCPIP Demo App-C32 EXPLORER_16 32MX360F512L ENC624J600 PSP 9.hex" precompiled application, among others, if you connected to the TCPPerformanceTest.c module and then attempted to simultaneously access the web server. The web server was returning data very slowly and failing to send the last parts of each file requested by the browser.
2. Eliminted a potential buffer overflow vulnerability from the HTTPHeaderParseContentLength() function in HTTP2.c. If an oversized or malformed Content-Length header is sent from the web client, the function will now gracefully fail by returning an HTTP 400 Bad Request error page. Thanks go to Mark Philipp for identifying this error and suggesting a solution.
3. Fixed a TCPOpen() problem in which the stack would continuously flood the network with nearly back-to-back ARP query packets if a client socket was created that specified a non-reachable remote IP address (ex: local gateway was offline, or for destinations on the same subnet, the actual remote node was offline). This problem would occur only after a few minutes (<10) had passed since the PIC was last reset. Thanks go to Sergey of DPS TELECOM for reporting this problem.
4. Fixed linking problem with BigInt_helpers.S (PIC24/dsPIC only) when targeting a PIC with more than 8KB of RAM. The interface registers (_iA, _xA, _iB, _xB, _iR, and _wC) are now forced into near RAM.
5. Cleaned up some uninitialized variable warnings in SNMP.c.
6. Fixed a sequence variable traversal bug in SNMP.c.
7. Cleaned up a large number of unsigned integer to signed integer comparison warnings produced by the MPLAB C Compiler for PIC18 MCUs (C18) version 3.32. With earlier versions of this compiler, these warnings would only be generated as messages, so they did not get displayed by default.
8. Some ENCX24J600 parallel bit bang modes work now. PSP Mode 5 indirect has been tested.
9. SSL client and server capabilities now work when using the ZeroG ZG2100M WiFi interface. In the 5.00 stack release, attempting to enable the STACK_USE_SSL_CLIENT or STACK_USE_SSL_SERVER TCPIPConfig.h options with this network controller would have resulted in an error trap. If an LCD was present, the LCD would display “encRdPtrRAWId = encWrPtrRAWId” when the error occurred.
10. The WiFi Iperf App demo locked up when an invalid command was entered at the serial port console. This is now fixed.
11. The WiFi Iperf App demo locked up when running with a PIC32 if iwconfig was typed at the serial port console. This is now fixed.
12. The Wifi Iperf App demo, when running on the PIC24 and PIC32, and compiled with the –Os option (min code size optimization), did not work. This is now fixed.
13. Change a lot of BerkeleyAPI.c internals. This may fix a number of BSD API problems.
14. Fix a problem with SNMP variables being inaccessible with certain unique PEN numbers.

Known Problems:

1. Telnet server module does not implement a lot of Telnet functions. As a result, it will likely not display correctly or work at all with some Telnet clients. The server was tested with the Microsoft telnet.exe utility which is provided with Microsoft Windows.
2. If the DHCP client and DHCP server are used at the same time and you connect two similar boards to each other (ex: two PICDEM.net 2 boards connected via a crossover cable), a race condition can occur where both nodes will disable their DHCP server and neither board will get a successful DHCP lease. If this unlikely scenario occurs, as a work around, simply reset one of the boards to reenable it's DHCP server.
3. HI-TECH PICC-18 compilers are not supported in this release. The supplied HI-TECH PICC-18 MPLAB projects usually will not compile and/or link.
4. ENC624J600 PSP modes 2, 4, 6, and 10 do not work at this time. Some Parallel Bit Bang modes may not work either. Some minor firmware changes are needed.
5. SSL client code doesn't work with ENC424J600/624J600 devices. The remote server terminates the connection reporting a bad record MAC (Message Authentication Code). The SSL client does work with other controllers.

v5.00 27 April 2009

Changes:

1. Added ZeroG ZG2100 802.11 WiFi controller support. The new TCPIP WiFi Demo App and WiFi Iperf App projects have been added, which default to using this controller.
2. Added Microchip ENC424J600/624J600 10/100 Ethernet controller support. Support for this controller is provided by the new ENCX24J600.c/.h files which perform the same role as the ENC28J60.c/.h or ETH97J60.c/.h files. Precompiled .hex files for the ENC624J600 controller require the use of the new Fast 100Mbps Ethernet PICtail Plus daughter card (AC164132). This product is not available at the time of the 5.00 TCP/IP stack release. However, it is anticipated to be available for purchase on www.microchipdirect.com in CQ3 2009.
3. Significantly updated the Internet Radio App project. Previously, radio stations were hard coded into program memory at compile time. Now, a dynamic Shoutcast directory client has been implemented which allows retrieval of radio stations at run time, offering endless stations you can tune into. The web pages for the radio have also been updated to allow control and status reporting of the board from a web browser.
4. Update SNMP Server (Agent) module to support SNMPv2C. The default Demo App web pages now include an SNMP reconfiguration capability to set the read and write community strings.
5. Added ICMPSendPingToHost() and ICMPSendPingToHostROM() APIs to ICMP (ping) client module. These two APIs are available only when STACK_USE_ICMP_CLIENT and STACK_USE_DNS is defined in TCPIPConfig.h. These functions allow pinging of DNS hostnames directly without the need for the application to convert the hostname to an IP address first by manually calling the DNS client APIs. With this addition, the PingDemo.c file was updated to ping the hostname "ww1.microchip.com" instead of a static IP address. Previously, the PingDemo would stop working a couple of months after the stack was released, due to the IP address of the www.microchip.com server dynamically changing. If the DNS module is not enabled, the ping demo will instead ping the local gateway IP address instead of ww1.microchip.com.
6. Updated TCPPerformanceTest.c code. The previous version would generate incorrect speed calculations at high data rates (ex: >1Mbyte/sec).
7. Added multiple connection support to Telnet server example module. To allow multiple connections, define MAX_SIMULTANEOUS_CONNECTIONS in Telnet.c greater than 1 and create an equal number of TCP_PURPOSE_TELNET type TCP sockets in the TCPSocketInitializer[] definition in TCPIPConfig.h.
8. Added more randomness to the local port selection when opening a client-mode TCP socket. This reduces the risk of reusing a previously used port number if the user power cycles the device.
9. Updated XEE* SPI EEPROM API functions. Writes are no longer required to start on an EEPROM page boundary, and writes can now be arbitrarily long without having to call XEEEndWrite() at each page boundary. Additionally, the XEEWriteArray() API has been added, which performs a similar operation to the SPIFlashWriteArray() API (but with no special erase cases to worry about).
10. Decoupled AppConfig storage in external SPI EEPROM or SPI Flash option from MPFS_USE_EEPROM and MPFS_USE_SPI_FLASH options. MainDemo.c will now save the AppConfig structure in external non-volatile memory, even if MPFS is unused (no HTTP or SNMP server modules enabled) or MPFS is using internal Flash program memory to store web pages/bib information. This change also allows the XEE*() and SPIFlash*() non-volatile read/write functions to be available at all times (even if MPFS is unused), as long as the appropriate hardware pinout definitions are present in HardwareProfile.h. SPI Flash and SPI EEPROM are no longer mutually exclusive with each other. However, if both are enabled simultaneously, AppConfig will be stored in the EEPROM, not the SPI Flash.
11. Added required SSL files to TCPIP Demo App MPLAB projects. SSL capabilities can now be turned on directly via the STACK_USE_SSL_SERVER and STACK_USE_SSL_CLIENT options in TCPIPConfig.h for these projects, assuming appropriate crypto libraries are installed (SW300052 available from https://www.microchipdirect.com/). With this change, the historical "SSL Demo App" folder has been removed.
13. Updated HardwareProfile.h files. This includes the addition of PIC18 Explorer board support, removal of the PICDEM Z profile, changes to the HI-TECH PICC-18 profiles for newer compilers, among other changes.
14. Added a TCP and UDP performance test measurements table to TCPIP Stack Help (TCPIP Stack Help.chm). Access this from the "Microchip TCP/IP Stack" book, "Stack Performance" page.
15. Updated MPFSlib project (Microchip.MPFS.dll file) so that C18 and C32 output from the MPFS2.exe utility is now identical for MPFS2 images. The generated .c file is now compatible with both C18 and C32 compilers simultaneously. Previously, the images generated for C18 would compile successfully for C32 projects, but would potentially operate incorrectly when compiler optimizations were turned on. Images generated for C32 would compile successfully and work on C18 projects, but the C18 compiler would take a very long time to process the file each time you rebuilt your MPLAB project. Now, the image generated for C18 matches the image generated for C32 and it will compile fast and work correctly on both platforms, even with compiler optimizations turned on.
16. Added schematics and BOMs for the Ethernet PICtail, Ethernet PICtail Plus, Fast 100Mbps Ethernet PICtail Plus, Internet Radio, PICDEM.net 2, and ZeroG ZG2100M PICtail development boards to the "MicrochipTCPIP StackDemo Board Files" folder.

Fixes:

1. Fixed a denial of service vulnerability in the NBNSGetName() function of the NBNS.c file. Previously, if a deliberately malformed packet was received, the PIC RAM could have become corrupted. Thanks go to David Talmage for finding this vulnerability.
2. Fixed Timer1 interrupt flag clearing code on PIC32 products. Previously, the Tick.c module was clearing the interrupt flag in an unsafe manner which could have corrupted other interrupt flags in the IFS0 register. Thanks go to Leon van Snippenberg working on the AVIX-RT RTOS for pointing this error out on the Microchip forums.
3. Fixed SNMP up-time variable. Previously the CustomSNMPApp.c module would respond with the number of Tick API ticks that elapsed, not the number of 10ms time slices that elapsed. The SNMP standard uses 10ms as its time base.
4. Fixed BigInt_helper.asm's _masBI() and _masBIROM() functions when the Br parameter's length modulo 4 was equal to 1 or 2. This bug previously caused the BigIntMod() function to sometimes go into an endless calculation loop on PIC18 products when using the SSL libraries and certain combinations of modulus data and length were used. Thanks go to Vasil Stoianov on the Microchip Ethernet forum for running into this defect and reporting it.
5. Fixed SSLSessionNew() so that it wouldn't "lose" SSL sessions after waiting a few hours. This would previously make it impossible to make new SSL connections after a while, but then after a few more hours, the sessions would become free again. Thanks go to Jim Stephens for identifying this issue and finding the solution.
6. Fixed an SSL 2.0 antique client hello record length calculation bug occurring when a received record was > 255 bytes.
7. Added retransmission capability to SendNotification() function in CustomSNMPApp.c. Previously, if an SNMP trap were sent, but the initial ARP query or response was lost on the network, the SendNotification() code would have deadlocked, and suppressed all future transmission of SNMP traps.
8. Fixed DNS client timeout if the DNS server is unable to be ARPed. Previously, the DNS client would retry ARPing the DNS server indefinitely if it was offline. Now, the DNS client will correctly abort if too many attempts to ARP the DNS server fail. Thanks go to Phil "andersop" on the Microchip Ethernet forum for identifying this error.
9. Suppressed transmission of a TCP RST packet to an unknown IP or MAC address if the TCPDisconnect() function was called on a client mode socket that was not finished with ARP or DNS resolution yet. Thanks go to Phil "andersop" on the Microchip Ethernet forum for pointing this behavior out.
10. Fixed TCP socket from disconnecting if the remote receive window was zero and TCPFlush() was still called. Thanks go to Bob Topper for identifying this issue and suggesting a solution.
11. Fixed Tick.c module returning incorrect values when TickGet() or other API was used with compiler optimizations turned on. Wrong values were observed when using MPLAB C Compiler for PIC24 MCUs and dsPIC DSCs version 3.12.
12. Fixed a number of SPI communications problems that could occur when compiler optimizations were turned on. The ENC28J60 was observed to not work correctly on the dsPIC33FJ256GP710 processor when compiled with MPLAB C Compiler for PIC24 MCUs and dsPIC DSCs version 3.12.
13. Fixed possible MPFS2 error when using an ASM30 .s image where MPFS_Start would be read using the wrong PSVPAG setting. You must rebuild your MPFS2 image file (ex: MPFSImg2.s) with this stack version's MPFS2.exe utility to get this correction applied.

Known Problems:

1. Telnet server module does not implement a lot of Telnet functions. As a result, it will likely not display correctly or work at all with some Telnet clients. The server was tested with the Microsoft telnet.exe utility which is provided with Microsoft Windows.
2. If the DHCP client and DHCP server are used at the same time and you connect two similar boards to each other (ex: two PICDEM.net 2 boards connected via a crossover cable), a race condition can occur where both nodes will disable their DHCP server and neither board will get a successful DHCP lease. If this unlikely scenario occurs, as a work around, simply reset one of the boards to reenable it's DHCP server.
3. HI-TECH PICC-18 compilers are not supported in this release. The supplied HI-TECH PICC-18 MPLAB projects usually will not compile and/or link.
4. ENC624J600 PSP modes 2, 4, 6, and 10 do not work at this time. Parallel Bit Bang mode does not work either. Some minor firmware changes are needed.

v4.55 10 November 2008

SSL Note: RSA.c and ARCFOUR.c have not changed between the 4.50 and 4.55 releases. Although the precompiled SSL Demo App .hex files will differ, you can continue to use the previous TCP/IP Stack v4.50 Encryption Add-on with this 4.55 stack version. 

Changes:

1. Added DNS client support for a secondary DNS server address. Previously, the AppConfig.SecondaryDNSServer setting was unused. Now, the DNS client module will automatically swap the AppConfig.PrimaryDNSServer and AppConfig.SecondaryDNSServer values after any DNS query timeout (or ARP timeout for the DNS server) and attempt the query with the alternative server. If AppConfig.SecondaryDNSServer is disabled by setting it to the IP address 0.0.0.0, the DNS client will only use the AppConfig.PrimaryDNSServer value and never swap the values. With this change, the DHCP client was also updated. If the DHCP server does not specify a secondary DNS server, then the DHCP client will now set the AppConfig.SecondaryDNSServer value to 0.0.0.0. Previously, it would change the AppConfig.SecondaryDNSServer setting only if the remote DHCP server offered a secondary DNS server.

Fixes:

1. Updated Internet Bootloader App project to correctly detect if the configuration bits are being changed or not. Previously, the bootloader always thought the configuration bits were being changed and thus had to always erase the last Flash page (largest memory address) twice for each firmware update. This did not cause any functional problems or specification violations, but it would decrease the effective Flash endurance of the last page.
2. Fixed a TCP socket memory corruption bug that would occur if TCPGetRemoteInfo() API was called twice with different socket handles without an intermediate call to any other TCP API that takes a TCP_SOCKET input. Thanks go to Bob Topper for identifying this problem and suggesting a solution.
3. Fixed the UDPIsGetReady() function so that it returns the number of bytes remaining in the packet based on the current read location. This is the same behavior as stack versions 4.18 and earlier. In stack versions 4.50 and 4.51, the UDPIsGetReady() function would always return the total number of bytes in the current packet, regardless of how many bytes the read pointer had been advanced through the UDPGet() and UDPGetArray() functions. Thanks go to Bob Topper for identifying this problem and suggesting a solution.
4. Fixed demo admin web page in TCPIP Demo App project so that the last byte of the MAC address can be changed, independent of the format it was entered by the user.
5. Fixed a buffer overflow bug that would occur when using the SSL server during hashing of the server certificate for the initial handshake. This error previously caused several bytes of random variables elsewhere in the project to get overwritten for each SSL connection.
6. BSD sockets API was updated to fix some issues.
7. LCDBlocking.c was updated to relax start up timing. This timing fix is specifically needed to support Explorer 16 boards with a Truly TSB1G7000 display (Novatek NT7603H controller).
8. Removed four uses of Arial Black font in MPFS2.exe utility. On some rare PC configurations, the use of this font caused the executable to not run.

Known Problems:

1. Telnet server module does not implement a lot of Telnet functions. As a result, it will likely not display correctly or work at all with some Telnet clients. The server was tested with the Microsoft telnet.exe utility which is provided with Microsoft Windows.
2. If the DHCP client and DHCP server are used at the same time and you connect two similar boards to each other (ex: two PICDEM.net 2 boards connected via a crossover cable), a race condition can occur where both nodes will disable their DHCP server and neither board will get a successful DHCP lease. If this unlikely scenario occurs, as a work around, simply reset one of the boards to reenable it's DHCP server.
3. HI-TECH PICC-18 projects may not compile when targeting the external ENC28J60 chip on the PICDEM.net 2 development board (instead of the internal Ethernet controller). This problem only applies when a PIC18F97J60 family part is the target. I.e. it compiles correctly for the HPC_EXPLORER + Ethernet PICtail.
4. MAC.h RXSIZE precompiler test for proper range doesn't work. This is not a functional problem, just a compile-time configuration test. Ensure that you don't over allocate TCP_ETH_RAM_SIZE or MAX_HTTP_CONNECTIONS.
5. HI-TECH PICC-18 STD 9.51PL1 cannot compile DynDNS.c. It raises an "inconsistent type" error while trying to perform a ROM pointer to integer cast. The older 9.50PL3 compiler release is required to compile this file.
6. HI-TECH PICC-18 STD 9.50PL3 does not initialize several static variables correctly on reset. This behavior breaks many stack modules in the TCPIP Demo App and TCPIP WebVend App projects. Additionally, string printing functions do not work correctly, so the supplied "TCPIP Demo App-HITECHPICC18 PICDEMNET2 18F97J60.hex" and "TCPIP WebVend App-HITECHPICC18 PICDEMNET2 18F97J60.hex" files may not correctly print the board's DHCP assigned IP address on the board's LCD (if present) and UART. To avoid these severe problems, use the Microchip MPLAB C Compiler for PIC18 MCUs. A free student edition can be downloaded from http://www.microchip.com/c18.

v4.51 24 July 2008

IMPORTANT NOTE: You must use MPLAB 8.10 or higher to successfully open the MPLAB projects. 

SSL Note: RSA.c and ARCFOUR.c have not changed between the 4.50 and 4.51 releases. Although the precompiled SSL Demo App .hex files will differ, you can continue to use the previous TCP/IP Stack v4.50 Encryption Add-on with this 4.51 stack version. 

Changes: None. This release includes bug fixes only. It is very important that applications using the ENC28J60 get fix item 7, below. 

Fixes:

1. TCPOpen() was previously failing if you used it to start a connection with a remote hostname, but the DNS module failed to resolve the remote address on the first try. This, for example, would occur if you powered up your board and tried to connect to a remote server before the Ethernet cable was attached. Once the Ethernet cable was attached, the socket would attempt to resolve and connect to a garbage address. The Internet Radio application would sometimes not begin playing the default station upon power up because of this problem.
2. Set SEQ.ACK = 0 for outbound TCP SYN packets. This fixes a connection compatibility problem with certain paranoid TCP/IP stacks that would validate this field even though the ACK flag was clear. This problem would previously cause the Microchip TCP/IP stack to be unable to connect client-mode TCP sockets to certain rare servers/services. Thanks go to Jean LE TUTOUR for finding one of these problem servers.
3. MPFSOpen() and MPFSOpenROM() for MPFS2 could leak a file handle if a name hash matched but no complete file name did. This has been corrected to prevent potential DOS attacks on the HTTP2 web server. Thanks to David Tan on the Microchip Ethernet formus for identifying this issue.
4. Fixed a bug in MPFS2.1 that caused compile errors when MPFS Classic images were generated for ASM30 containing files whose length was either zero or a multiple of 12.
5. Fixed an issue in HTTPPostConfig() that caused it to ignore the flag that was set when invalid IP address input was detected. This issue only affects the example configuration page and only exists in v4.50 (prior versions functioned correctly). Also corrected an issue where user input could potentially overflow into part of the shadow AppConfig in the same function. Thanks to prinz3nroll3 on the Microchip Ethernet forums for identifying both of these issues.
6. Implemented Explorer 16 development board 5V LCD errata workaround to LCDBlocking.c. This corrects the A/D converter from returning erratic readings on certain Explorer 16 boards. LCD I/O pins are now continuously driven by the microcontroller instead of going high impedance when idle.
7. Fixed a critical ENC28J60 revision B7 errata workaround problem in the ENC28J60.c, MACFlush() function. Previously, the code was checking for an EREVID register value of 0x07 for silicon revision B7. This was incorrect. Silicon revision B7 actually has an EREVID value of 0x06. Note that this problem was caused by an incorrect EREVID value published in DS80349A, the B7 silicon errata documentation. Make sure to use DS80349B or later.

Known Problems:

1. Telnet server module does not implement a lot of Telnet functions. As a result, it will likely not display correctly or work at all with some Telnet clients. The server was tested with the Microsoft telnet.exe utility which is provided with Microsoft Windows.
2. If the DHCP client and DHCP server are used at the same time and you connect two similar boards to each other (ex: two PICDEM.net 2 boards connected via a crossover cable), a race condition can occur where both nodes will disable their DHCP server and neither board will get a successful DHCP lease. If this unlikely scenario occurs, as a work around, simply reset one of the boards to reenable it's DHCP server.
3. HI-TECH PICC-18 projects may not compile when targeting the external ENC28J60 chip on the PICDEM.net 2 development board (instead of the internal Ethernet controller). This problem only applies when a PIC18F97J60 family part is the target. I.e. it compiles correctly for the HPC_EXPLORER + Ethernet PICtail.
4. MAC.h RXSIZE precompiler test for proper range doesn't work. This is not a functional problem, just a compile-time configuration test. Ensure that you don't over allocate TCP_ETH_RAM_SIZE or MAX_HTTP_CONNECTIONS.
5. HI-TECH PICC-18 STD 9.51PL1 cannot compile DynDNS.c. It raises an "inconsistent type" error while trying to perform a ROM pointer to integer cast. The older 9.50PL3 compiler release is required to compile this file.
6. HI-TECH PICC-18 STD 9.50PL3 does not initialize several static variables correctly on reset. This behavior breaks many stack modules in the TCPIP Demo App and TCPIP WebVend App projects. Additionally, string printing functions do not work correctly, so the supplied "TCPIP Demo App-HITECHPICC18 PICDEMNET2 18F97J60.hex" and "TCPIP WebVend App-HITECHPICC18 PICDEMNET2 18F97J60.hex" files may not correctly print the board's DHCP assigned IP address on the board's LCD (if present) and UART. To avoid these severe problems, use the Microchip MPLAB C Compiler for PIC18 MCUs. A free student edition can be downloaded from http://www.microchip.com/c18.

v4.50 02 June 2008

IMPORTANT NOTE: You must use MPLAB 8.10 or higher to successfully open the MPLAB projects. Also, ensure that the latest C compiler is used. This release was tested against MPLAB C Compiler for PIC18 MCUs version 3.20, MPLAB C Compiler for PIC24 MCUs and dsPIC DSCs version 3.10, MPLAB C Compiler for PIC32 MCUs version 1.01, and HI-TECH PICC-18 version 9.50PL3 (STD). Earlier compilers may not be able to compile this TCP/IP stack release. 

Changes:

1. Added SSL 3.0 client capabilities, including SMTP over SSL. The SSL modules supports up to 1024-bit RSA handshakes and 128-bit ARCFOUR bulk encryption. This can be demonstrated using the SMTP client. SSL server support is functional, but a key generation utility is not yet provided and support over HTTPS is not yet reliable with all browsers. IMPORTANT: Encryption software is covered by US Export Control law, so it is not directly downloadable from the Microchip website. To use the encryption modules, you must order SW300052 from microchipDIRECT [ https://www.microchipdirect.com/ ] and install the required libraries.
2. Added Berkeley Sockets Distribution (BSD) API translation layer. You can now call the well know Berkeley APIs instead of or in addition to the Microchip specific APIs. To use this new functionality, define STACK_USE_BERKELEY_API and configure BSD_SOCKET_COUNT in TCPIPConfig.h. Three new source code demos are provided to demonstrate this API: BerkeleyTCPClientDemo.c, BerkeleyTCPServerDemo.c, and BerkeleyUDPClientDemo.c. The TCP client demo is identical to the GenericTCPClient.c demo, but implemented using Berkeley Sockets. The UDP client demo is similarly identical to the SNTP.c client. The TCP server demo listens on TCP port 9764 and will echo any traffic received back to the sender. It allows up to 3 simultaneous connections when there are an adequate number of sockets defined in the TCPSocketInitializer[] array in TCPIPConfig.h.
3. Added support for Dynamic DNS services. See the Dynamic DNS Client module in the TCP/IP Stack Help for details. Presently, dyndns.org, dyndns.com, no-ip.com, and dns-o-matic.com are supported.
4. Added the Microchip TCP/IP Configuration Wizard to the Utilities folder, facilitating easier configuration of the TCP/IP Stack through a graphical application.
5. Restructured TCPIPConfig.h to remove rule-enforcement logic, placing the removed sections in TCPIP.h. Many other project structure changes were also made to clean up the general distribution appearance.
6. Increased DHCP Server default lease duration to 60 seconds instead of 15 seconds. Some computers were losing their IP lease before performing a renew operation with only a 15 second lease.
7. Removed CLOCK_FREQ, INSTR_FREQ, and PERIPHERAL_FREQ macro definitions. GetSystemClock(), GetInstructionClock(), and GetPeripheralClock() now return these respective values. This change was made for compatibility with other Microchip software libraries.
8. Added TCP Fast Retransmission capability. Whenever three duplicate ACK packets arrive, the stack will now immediately perform a retransmit operation. This greatly improves recovery latency whenever the network loses a packet for applications that stream TX data using TCP.
9. Improved TCP Keep Alive mechanism to automatically close TCP sockets which do not receive any keep-alive responses for TCP_MAX_UNACKED_KEEP_ALIVES (default 6) times. This means that, by default, any connection that catastrophically breaks without notifying us (ex: user unplugs cable, Internet connection goes down, etc.) will time out and automatically close after 60 seconds (TCP_MAX_UNACKED_KEEP_ALIVES * TCP_KEEP_ALIVE_TIMEOUT). Server oriented sockets will return to the listening state. Client oriented sockets will close, but the TCP_SOCKET handle will continue to remain valid until the application calls TCPDisconnect(). Applications can check if the socket became disconnected and reset by calling TCPWasReset() or TCPIsConnected(). Note that this keep alive implementation will only close sockets that are broken (remote node is not responding to TCP requests). It will not close or otherwise interfere with idle connections in which the application is not transmitting or receiving data and wishes to keep the connection open.

10.Added a TCP RX SYN queue of depth TCP_SYN_QUEUE_MAX_ENTRIES (default 3). This queue automatically saves incoming SYN packets destined for a local server port which is already connected to a different client. When the client disconnects, the SYN data is pulled out of the queue and the socket immediately attempts to connect to the next client. This improves connect time performance since the remote client no longer has to retransmit the SYN request if it was unserviceable the first time around. This is most apparent with the HTTP/HTTP2 servers which previously performed poorly with certain modern web browsers which attempt to open many simultaneous connections to the web server, such as Mozilla Firefox 3 beta 5 and Apple Safari 3.1. Entries in the queue automatically time out after TCP_SYN_QUEUE_TIMEOUT (default 3 seconds) so as to prevent the queue from filling up permanently if several connection requests arrive for a service that is in use and will not be available for an extended period. 11.Modified the structure of the MPFS2 FAT (now known as MPFS2.1) to include name hashes first. This speeds up opening files by 25%, and makes opening index files nearly instant. 12.Updated the MPFS2 Utility. MPFS2.1 now supports the new FAT structure and provides a cleaner interface. It also writes images to disk as they are created, which eliminates the IndexOutOfBounds exceptions some users had reported. Finally, uploads are now truly multi-threaded. 13.Source code to the MPFS2.exe PC utility is now released. Find it in the Microchip SolutionsMicrochipTCPIP StackUtilitiesSourceMPFS21 folder. This project is designed to compile with Microsoft Visual C# 2008 Express Edition. 14.Added support for SST25VFxxxB serial flash parts in 2, 4, 8, 16, and 32Mbit densities. These parts can be used to replace EEPROMs for storing MPFS images (both versions) and custom data. 15.Added HTTPReadPostName, HTTPReadPostValue, and HTTPReadPostPair functions to facilitate easier processing of data arriving via POST. 16.Split HTTPAuthenticate API into separate functions: HTTPNeedsAuth and HTTPCheckAuth. This function was already split internally, and didn't make sense as a single API. 17.Updated DHCP client to close its UDP socket when idle (bound state) to save a small amount of resources. 18.Removed LED_IO macro from all hardware profiles because it is not suitable for use on certain hardware platforms that have non-contiguous LEDs or reversed bit ordering. Use the new LED_GET() and LED_PUT(val) macros to read and write to all of the LEDs at once. 19.Added Ethernet Hash Table Calculator.exe to the Utilities folder and start menu. This tool will calculate the correct bit that you must set in the EHT0-EHT7 registers on the ENC28J60 and PIC18F97J60 family devices for using the Hash Table RX filter. This is useful only for fixed MAC addresses known at design time. For addresses that are known at run time, use the SetRXHashTableEntry() function in the ENC28J60.c or ETH97J60.c files to set the correct EHT0-EHT7 bit. 

Fixes:

1. Fixed a buffer overflow data corruption issue in the FTP module that arises when too many parameters were passed on the command line.
2. Moved TCPWasReset checking in HTTP2 to execute for every socket on every loop. Previously, it would only execute when a socket reconnected, which caused the RX buffer to not resize until after data was received. Some platforms (notably FF2 on Ubuntu) would stall if the initial advertised RX window was too small, and this change corrects that issue.
3. Updated SendSystemReset() and MACInit() initialization routine in ENC28J60.c. Previously, if the ENC28J60 was placed into sleep mode by calling MACPowerDown(), the SendSystemReset() command would not work anymore. This would leave the ENC28J60 in power down if the host PIC was ever reset. SendSystemReset() should work for all conditions with this update. Thanks go to Rob Haverkort on the Microchip Ethernet forum for identifying this problem.
4. Fixed an alignment bug in HTTP2 that caused redirects to fail when the MPFS2 image was stored in Flash program memory. Thanks to Todd Boaz on the Microchip Ethernet forum for identifying this bug, and Chen Qu for posting a solution.
5. Fixed SNTP client from losing accuracy if you called SNTPGetUTCSeconds() 10s of thousands of times since the last server synchronization. Thanks go to "pic123" on the Microchip Ethernet forum for noticing this error.
6. Fixed a TickGet*() API problem where the returned tick value could be off by 64K ticks occasionally on PIC24, dsPIC30/33, and PIC32 processors. This bug was previously fixed in stack versions 4.13 and 4.16, but it was unintentionally recreated in 4.18 due to PIC32 changes.
7. Fixed UART2TCPBridge module from failing to connect to a remote server when USE_REMOTE_TCP_SERVER was defined.
8. Fixed an issue that prevented SNMP SETs on 16 and 32 bit parts when using MPFS2. Thanks go to Milena K on the Microchip Ethernet forum for identifying this problem.
9. Fixed a rare buffer corruption issue that could occur with UDP if TCP was also enabled.

10.Fixed a Tick rollover error in HTTP2. Thanks go to Paul Bixel on the Microchip Ethernet forum for identifying this problem. 11.Fixed an MPFS2 bug in which an excessive value to MPFS_SEEK_REWIND may have failed to return an error. Thanks go to Paul Bixel on the Microchip Ethernet forum for identifying this problem as well. 12.SMTP Client now sends EHLO when using authentication. Previously, the HELO command was used, even with authentication enabled. Using HELO with authentication creates incompatibilities with certain SMTP servers. 13.Improved Internet Bootloader robustness by retransmitting ACKs in response to data retransmissions by the remote sending node. Previously, if an ACK packet was lost before reaching the sending node, the TFTP upload would fail and need to be restarted. Thanks go to "coolvibe" Dave Collier on the Microchip Ethernet forum for identifying this behavior. 14.Fixed TFTP Internet Bootloader from not being accessible from Linux TFTP clients which were setting the IP header "Don't Fragment" flag bit. 15.Changed TCP so that unsent data that is automatically flushed by the TCP_AUTO_TRANSMIT_TIMEOUT_VAL timer includes the PSH flag. This improves GUI responsiveness for certain applications which rely on this automatic flush feature, such as the UART2TCPBridge module. 16.Fixed TCP socket loss issue which could occur if the TCP TX FIFO size was greater than 536 bytes (TCP_MAX_SEG_SIZE). Before the fix, the socket would have gotten tied up indefinitely performing retransmissions every 1.0 seconds without detecting that the remote node was disconnected. 17.Fixed TCP socket hang issue that would occur if the PIC sent out a FIN and the remote node never responded with a corresponding FIN. The socket would have gotten stuck indefinitely in the TCP_FIN_WAIT_2 state. Thanks go to Mr. Kyle Strickland with AW North Carolina for identifying this bug. 18.Fixed UDPSetRxBuffer() function from not working if it was called before having called UDPGet() or UDPGetArray() at least once. 19.Fixed an offset error of +2 milliseconds being returned from TickConvertToMilliseconds(). Thanks go to Andrés ("saturn") on the Microchip Ethernet forum for finding this error. Note that due to integer truncation during division, this function can be off by 0.2% or so, depending on the value returned by GetPeripheralClock(). 20.Updated DelayMs() macro for MPLAB C Compiler for PIC18s to work correctly when a large parameter was given. You should now be able to delay between 0 and 65535 milliseconds across all supported compilers without ending up with an unexpectedly short delay. 

Known Problems:

1. Telnet server module does not implement a lot of Telnet functions. As a result, it will likely not display correctly or work at all with some Telnet clients. The server was tested with the Microsoft telnet.exe utility which is provided with Microsoft Windows.
2. If the DHCP client and DHCP server are used at the same time and you connect two similar boards to each other (ex: two PICDEM.net 2 boards connected via a crossover cable), a race condition can occur where both nodes will disable their DHCP server and neither board will get a successful DHCP lease. If this unlikely scenario occurs, as a work around, simply reset one of the boards to reenable it's DHCP server.
3. HI-TECH PICC-18 projects may not compile when targeting the external ENC28J60 chip on the PICDEM.net 2 development board (instead of the internal Ethernet controller). This problem only applies when a PIC18F97J60 family part is the target. I.e. it compiles correctly for the HPC_EXPLORER + Ethernet PICtail.
4. MAC.h RXSIZE precompiler test for proper range doesn't work. This is not a functional problem, just a compile-time configuration test. Ensure that you don't over allocate TCP_ETH_RAM_SIZE or MAX_HTTP_CONNECTIONS.
5. HI-TECH PICC-18 STD 9.51PL1 cannot compile DynDNS.c. It raises an "inconsistent type" error while trying to perform a ROM pointer to integer cast. The older 9.50PL3 compiler release is required to compile this file.
6. HI-TECH PICC-18 STD 9.50PL3 does not initialize several static variables correctly on reset. This behavior breaks many stack modules in the TCPIP Demo App and TCPIP WebVend App projects. Additionally, string printing functions do not work correctly, so the supplied "TCPIP Demo App-HITECHPICC18 PICDEMNET2 18F97J60.hex" and "TCPIP WebVend App-HITECHPICC18 PICDEMNET2 18F97J60.hex" files may not correctly print the board's DHCP assigned IP address on the board's LCD (if present) and UART. To avoid these severe problems, use the Microchip MPLAB C Compiler for PIC18 MCUs. A free student edition can be downloaded from http://www.microchip.com/c18.

v4.18 28 November 2007

Changes:

1. Added C32 and PIC32MX support. Some things were cleaned up in the process.
2. Removed linker scripts from C30 MPLAB projects. MPLAB IDE 8.00 can automatically select the correct linker script for 16-bit and 32-bit products.
3. Updated TCPPerformanceTest.c module. Now it automatically calculates the TX throughput and displays it for you. Also, there is now an RX throughput testing mode, which listens on a separate TCP socket (port 9763) when a TCP socket of type TCP_PURPOSE_TCP_PERFORMANCE_RX is allocated in TCPIPConfig.h. The RX socket is by default not enabled to save memory, so you must create a TCP_PURPOSE_TCP_PERFORMANCE_RX socket in TCPIPConfig.h and ensure that enough memory is allocated to accommodate it to test the RX performance test. When connected to port 9763, send a large amount of data and the PIC microcontroller will send back a count of how many bytes were received per second.
4. UDPPerformanceTest.c module now transmits 1024 packets on start up and then stops to prevent continually broadcast flooding your network. To transmit more packets after 1024 is reached, hold down BUTTON3 (left-most button on most boards).
5. Significantly improved the speed of the MD5 and SHA-1 functions. Gains for the 8-bit compilers were 50-75%, while 16-bit parts saw more modest improvements (~10%).
6. Reimplemented TCP_CLOSE_WAIT TCP state ("CLOSE WAIT" in RFC793). Now, TCP sockets that receive a FIN from the remote node will hold off transmitting a FIN back to the remote node until the TCP_CLOSE_WAIT_TIMEOUT (defined at the top of TCP.c) elapses or immediately when the application calls the TCPDisconnect() function. This makes it possible for the application to transmit a response back to the remote node before the socket becomes closed on our end. Similarly, it simplifies application usage of the last RX bytes received as these bytes are now assured to still be in the RX FIFO for at least TCP_CLOSE_WAIT_TIMEOUT seconds. TCP_CLOSE_WAIT_TIMEOUT defaults to 200ms in this stack version.
7. Pushed the SNTP requery on failure timeout up some. It was ~14 seconds and is now ~20 seconds.
8. Added TFTPOpenROMFile() API to complement TFTPOpenFile() when using PIC18 products.
9. Added a fourth parameter to newAJAXCommand() in mchp.js, allowing data to be POSTed along with the AJAX request.

10.Deprecated the TCP Loopback functions, which includes TCPOpenLoopback, TCPCloseLoopback, TCPIsLoopback, TCPInject, and TCPSteal. These functions were added in 4.10 for future SSL support, but have since become unnecessary. They are of limited usefulness, and so are being removed to save code space. The functions are still available in this version, but will be removed in the next release. 11.Added SMTPClient.ServerPort field to the SMTP API. This allows the remote server port number to be specified dynamically at run time instead of being hard coded to the SMTP_PORT value defined at the top of SMTP.c. SMTP_PORT is now only a default. 12.Added web interface to the SMTP module in the TCPIP Demo App applications. You can now configure the SMTP module and send emails directly from within your web browser. The HTTPPostEmail() function in CustomHTTPApp.c also demonstrates how to send MIME encoded attachments in emails. The default demo will send button states, LED states, and the current potentiometer reading as a CSV file attached to the email. 13.Changed SMTPDemo() in MainDemo.c to trigger on BUTTON2 and BUTTON3 simultaneously held down instead of BUTTON0 only. 

Fixes:

1. Fixed an ENC28J60.c MACGetArray() bug which would overwrite one byte of memory at address 0xFFFFFFFF if you provided NULL for the destination address pointer.
2. Fixed an MPFS2.c MPFSGet() bug which would overwrite memory address 0x00000000 if a NULL pointer was provided as the destination.
3. Fixed a bug in the HTTP2 server accessing incorrect sockets if an inadequate number of sockets were available on POR.
4. Fixed Internet Bootloader project from failing with a timeout if an ARP packet arrived during the Erase/Write operation.
5. Fixed DHCP client RFC non-compliance where it would send the ciaddr field in the initial SELECTING state. Also, in the RENEWING state, the Requested IP Address option was being sent, which is illegal. These changes may fix compatibility problems with certain DHCP servers.
6. Fixed TFTP Client's TFTPCloseFile() function from sending data using a wrong UDP socket if StackTsk() was called after TFTPIsFileOpened() was last called.
7. Added two zero bytes to the ICMP echo request payload to improve compatibility with some buggy NAT routers.

Known Problems:

1. Telnet server module does not implement a lot of Telnet functions. As a result, it will likely not display correctly or work at all with some Telnet clients. The server was tested with the Microsoft telnet.exe utility which is provided with Microsoft Windows.
2. If the DHCP client and DHCP server are used at the same time and you connect two similar boards to each other (ex: two PICDEM.net 2 boards connected via a crossover cable), a race condition can occur where both nodes will disable their DHCP server and neither board will get a successful DHCP lease. If this unlikely scenario occurs, as a work around, simply reset one of the boards to reenable it's DHCP server.
3. HI-TECH PICC-18 projects may not compile when targeting the external ENC28J60 chip on the PICDEM.net 2 development board (instead of the internal Ethernet controller). This problem only applies when a PIC18F97J60 family part is the target. I.e. it compiles correctly for the HPC_EXPLORER + Ethernet PICtail.
4. MAC.h RXSIZE precompiler test for proper range doesn't work. This is not a functional problem, just a compile-time configuration test. Ensure that you don't over allocate TCP_ETH_RAM_SIZE or MAX_HTTP_CONNECTIONS.

v4.16 06 November 2007

Changes:

1. Added Internet Radio application. This is a TCP client application which downloads streaming MP3 audio from a Shoutcast server and then plays it back to stereo earphones via a VLSI VS1011 audio decoder.
2. Added SPIRAM.c module. This module is intended for interfacing to an AMI Semiconductor N256S0830HDA SPI RAM chip. The TCP module can now interface directly to this SPIRAM module to store TCP socket FIFO buffers and other TCB data in the external RAM.
3. Added TCP_OPTIMIZE_FOR_SIZE compile time configuration macro to TCP.c file. When optimizing for small code size, the TCP module ROM footprint shrinks up to 6KB, but performance may slow down on some processors (namely PIC18s, where the penalty is approximately 15%).
4. Added USE_EEPROM_25LC1024 compile time configuration macro to TCPIPConfig.h. Enable this definition if you are storing your MPFS[2] on a 1Mbit 25LC1024 or similar EEPROM device that uses 24-bit addressing and a 256 byte write page size.
5. Changed LCDBlocking.c module initialization code. It should now be possible to use 4-bit mode on certain "unusual" LCD controllers, like the Samsung S6A0032. Most PICDEM.net 2 and Explorer 16 boards use an LCD with this controller.
6. SNTP client now attempts to requery the SNTP server about every 14 seconds if the last query attempt fails. This allows the internal time value to become valid quickly should the board be powered up before an Ethernet cable is attached or if the DHCP client doesn't obtain an IP address quickly enough. Previously, it would take up to 10 minutes after plugging the Ethernet cable in to get a correct time value from the SNTP server.
7. Added UDP_USE_TX_CHECKSUM compile time configuration macro to TCPIPConfig.h. When enabled, all UDP packets will have a correct UDP checksum computed and inserted into the UDP header of outbound packets. If you do not define this macro, the UDP checksum will be disabled (left as 0x0000), which is how previous stack versions operated. Note that enabling checksum generation cuts your maximum UDP TX throughput by nearly half due to the required computations.
8. Substantially changed TCP socket RX and TX FIFO allocation. Now, sockets can be stored either in Ethernet RAM, PIC RAM, or external (SPI) RAM. Previously, sockets could only be allocated in Ethernet RAM, which was not scalable.
9. Added TCPOpen() API function. This replaces TCPListen() and TCPConnect(). TCPOpen() supports a large number of options that will make the creation of client mode sockets much easier. You can specify the remote node as a hostname that needs DNS and ARP resolution, an IP address that only needs ARP resolution, or legacy NODE_INFO pointer for direct compatibility with the previous TCPListen() and TCPConnect() APIs. TCPOpen() also supports a socket type parameter which will allow you to use the new TCP socket RAM allocation system.

10.Added TCP Keep Alive mechanism defined by RFC 1122 section 4.2.3.6 to the TCP module. This helps automatically detect lost connections. If the remote node sends back an RST, this immediately closes the lost connection on our end. Currently, no action is taken if the keep alive gets no response. Note that this feature deviates from the standard by defaulting to only 10 seconds instead of over 2 hours. Also deviating from the standard, this feature is enabled by default. To disable it, undefine TCP_KEEP_ALIVE_TIMEOUT at the top of TCP.c. 11.Moved TCPPerformanceTest.c module from default port 12345 to 9762. 12.Moved UDPPerformanceTest.c module from default port 12345 to 9, the "discard" protocol port. 

Fixes:

1. The DHCP client now specifically requests the previous IP address when a DHCP renewal occurs.
2. The SNTP client now correctly maintains time when repetitively calling SNTPGetUTCSeconds() between an NTP requery event. Thanks go to Rob Haverkort on the Microchip Ethernet forum for noticing the time value incrementing far faster than it should have.
3. TCP module will not transmit a bunch of unnecessary duplicate ACK packets when data is ready to be transmitted but the remote RX window is zero. This previously didn't cause anything to break, but would waste CPU time and bandwidth sometimes.
4. TCP sockets will no longer automatically close if the remote RX window stays zero for several seconds.
5. Fixed TFTP Internet Bootloader project from corrupting the configuration fuses. Previously, this would result in the Watchdog timer being enabled and causing an unintentional reboot every few minutes with the demo TCP/IP stack.

Known Problems:

1. Telnet server module does not implement a lot of Telnet functions. As a result, it will likely not display correctly or work at all with some Telnet clients. The server was tested with the Microsoft telnet.exe utility which is provided with Microsoft Windows.
2. TFTPc module has not been tested with this version.
3. If the DHCP client and DHCP server are used at the same time and you connect two similar boards to each other (ex: two PICDEM.net 2 boards connected via a crossover cable), a race condition can occur where both nodes will disable their DHCP server and neither board will get a successful DHCP lease. If this unlikely scenario occurs, as a work around, simply reset one of the boards to reenable it's DHCP server.
4. HI-TECH PICC-18 projects may not compile when targeting the external ENC28J60 chip on the PICDEM.net 2 development board (instead of the internal Ethernet controller). This problem only applies when a PIC18F97J60 family part is the target. I.e. it compiles correctly for the HPC_EXPLORER + Ethernet PICtail.
5. MAC.h RXSIZE precompiler test for proper range doesn't work. This is not a functional problem, just a compile-time configuration test. Ensure that you don't over allocate MAX_TCP_SOCKETS, TCP_TX_FIFO_SIZE, TCP_RX_FIFO_SIZE, or MAX_HTTP_CONNECTIONS.

v4.13 02 October 2007

Changes:

1. Added command line support to the MPFS2.exe tool. You can now generate MPFS output files using batch scripts or other console applications.
2. Added dynamic variable parameter capabilities to the MPFS2 utility. To use, add the parameters you wish to pass to the end of the dynamic variable. All parameters are passed as WORD values. (ex: ~myArray(2,5)~ )
3. Added TCPWasReset() API to allow the application layer to be notified if an underlying socket reset has occurred (ex: remote node disconnects, cable is disconnected and times out, user calls TCPDisconnect(), etc.). The reset state is latching, which allows the application layer to detect if a remote node disconnects and a new connection occurs on the same socket before the application can detect the original disconnection through the TCPIsConnected() API.
4. Added a counter to the UDPPerformanceTest module and made it supress transmission if an Ethernet link is not present.
5. Added TCPIP WebVend App example application to the main stack distribution. This corresponds to three new Microchip Webinars being published on the HTTP2 server usage topic.

Fixes:

1. Fixed MPFS2.exe PC utility from crashing if you attempt to generate an MPFS classic .bin/.c/.s output file.
2. Fixed RCONbits definition for HPC_EXPLORER hardware profile when using the HI TECH PICC-18 compiler.
3. Fixed a MPFSGetFilename() bug when using C30 and MPFS2 images stored in program memory. Thanks to Billy Walton on the Microchip Ethernet forum for identifying this issue.
4. Fixed a TCP RX FIFO corruption problem which would occur if the remote node sent more data than could fit in our RX FIFO in a single packet. The GeneticTCPClient.c module was subject to experiencing this problem when connected to www.google.com's servers.
5. Fixed a DHCP client UDP socket leak if you called DHCPDisable() after the DHCP client had already obtained a UDP socket. Thanks go to Matthew Kendall on the Microchip Ethernet forum for identifying this problem.
6. Fixed a SNMP Server module bug testing a string length (with respect to SNMP_COMMUNITY_MAX_LEN) being off by one, resulting in possible memory corruption. Thanks go to Matthew Kendall on the Microchip Ethernet forum for identifying this problem.
7. Cleaned up some C30 compiler warnings related to macro definitions with inadequate parenthesis in them.
8. Fixed HTTP2 module sometimes returning a 501 error instead of a correct web page when being bombarded with new connection requests.
9. Fixed a TickGet*() API problem where the returned tick value could be off by 64K ticks occasionally on PIC24 and dsPIC processors.

10.Fixed SMTP client module failing to send email when attempting to send an email with a `CC' or `BCC' field that was in ROM while the `To' field was in RAM or visa versa. 11.Fixed TCP module sending an incorrect sequence number in RST packets sent when in the TCP_SYN_SENT state and an invalid segment arrives. In prior stack versions, some TCP client applications might take a very long time to recover in the event of a power failure, reset, and subsequent reconnect to a remote server that still thinks the old connection is still active. With this fix, reconnections should be possible almost immediately after a power failure because the correct RST packet will cause the old connection to get closed right away. 12.Fixed a TCP socket leak problem that would occur over if the local PIC called TCPDisconnect() and the remote node didn't send us a correct FIN response. Sockets could previously get lost in the TCP_FIN_WAIT_2 state and wouldn't recover unless the application called TCPDisconnect() a second time with the same socket handle. 

Known Problems:

1. Telnet server module does not implement a lot of Telnet functions. As a result, it will likely not display correctly or work at all with some Telnet clients. The server was tested with the Microsoft telnet.exe utility which is provided with Microsoft Windows.
2. TFTPc module has not been tested with this version.
3. If the DHCP client and DHCP server are used at the same time and you connect two similar boards to each other (ex: two PICDEM.net 2 boards connected via a crossover cable), a race condition can occur where both nodes will disable their DHCP server and neither board will get a successful DHCP lease. If this unlikely scenario occurs, as a work around, simply reset one of the boards to reenable it's DHCP server.
4. HI-TECH PICC-18 projects may not compile when targeting the external ENC28J60 chip on the PICDEM.net 2 development board (instead of the internal Ethernet controller). This problem only applies when a PIC18F97J60 family part is the target. I.e. it compiles correctly for the HPC_EXPLORER + Ethernet PICtail.
5. HI-TECH PICC-18 projects will not correctly set the processor configuration fuses through code using the __CONFIG() macro. Ensure that the configuration fuses are manually set correctly via the MPLAB IDE Configuration Bits dialog. This problem has been observed with compiler version 9.50PL3.
6. MAC.h RXSIZE precompiler test for proper range doesn't work. This is not a functional problem, just a compile-time configuration test. Ensure that you don't over allocate MAX_TCP_SOCKETS, TCP_TX_FIFO_SIZE, TCP_RX_FIFO_SIZE, or MAX_HTTP_CONNECTIONS.
7. GenericTCPClient example of downloading a web page from www.google.com is extremely slow. The default TCP socket has too little RX space to accept a full packet sent from Google's servers, so the remote server must retransmit a lot of data, slowing the transfer down a lot. Making TCP_RX_FIFO_SIZE 536 bytes or bigger and correspondingly shrinking MAX_TCP_SOCKETS will correct this problem.

v4.11 27 August 2007

IMPORTANT NOTE: You must use MPLAB 7.62 or higher to successfully open the MPLAB projects. 

Changes:

1. Added a Microchip TCP/IP Stack Users' Guide to document the stack features/modules/and APIs and address the stale AN833 documentation. Note that this is a work in progress. Many modules have yet to be documented in the Users' Guide.
2. Added HTTP2 module. This HTTP module includes a whole new API and supreme new features, such as POST support, cookies support, browser authentication support, and more.
3. Added MPFS2 module. This module is required for the new HTTP2 module and performs better while having fewer limitations. Long filenames and folders are now supported.
4. Added a new GUI based MPFS2.exe PC utility. The older MPFSv2.exe GUI application and MPFS.exe command line tool has been retired. The new utility has advanced features, such as MPFS2 file format support, GZIP compress, etc.
5. Added a TFTP bootloader. This is a stand alone project and currently only supports the PIC18F97J60 family of PIC processors with internal Ethernet.
6. Added UART2TCPBridge.c file and STACK_USE_UART2TCP_BRIDGE option to TCPIPConfig.h. This new module acts as a TCP and UART bridge, with a high priority UART interrupt and dedicated UART TX and RX FIFOs for minimum UART latency and maximum performance. By default, the bridge acts as a TCP server and listens on port 9761. The UART baud rate defaults to 19200. The bridge can be reconfigured to act as a TCP client.
7. Added Simple Network Time Protocol (SNTP) client. This module automatically obtains the current time (date) from the Internet. Enable this module by defining STACK_USE_SNTP_CLIENT in TCPIPConfig.h. Obtain the current time (in seconds since 00:00:00 1970) by calling the SNTPGetUTCSeconds() API.
8. Added support functions Base64Encode() and Base64Decode() in Helpers.c. Base 64 is required for the new HTTP2 module, but of general use to many applications.
9. Added SMTP Authentication support to the SMTP Client. To use this, set the SMTPClient.Username and SMTPClient.Password string pointers to a non-NULL value before calling SMTPSendMail(). Applications implementing email transmission capabilities should expose these options to the end-user for configuration. To use SMTP servers that do not support the AUTH LOGIN authentication command, simply leave the SMTPClient.Username and SMTPClient.Password pointers as their default NULL value.

10.Converted DHCPDisable() from a macro to a real function and added the complementary DHCPEnable() function. These two functions can be used at run time to dynamically switch between using a static IP address and configuration and DHCP assigned IP address and configuration. 11.Updated StringToIPAddress() to work more robustly, including the ability to decode host name strings and determine if they contain a valid IP address or not. Also, the complementary ROMStringToIPAddress() function was added. 12.Updated the DNS module. Now, if you give it an IP address string to resolve, it will convert the string to an IP address and immediately return without querying the DNS. 13.Shrunk the advertised TCP Maximum Segment Size from 576 bytes to 528 bytes. This might improve compatibility if your TCP data has to propagate over nodes with small MTUs and you have a correspondingly large TCP RX FIFO defined. 14.Performed some maintenance on the FTP.c file. No significant functionality has been changed, but some potential problems were corrected. 15.Altered Tick.c file and API. Now, the Tick module can operate maximum precision, returning the value of the actual Timer as it is counting, without disturbing the timer count by writing to it or disabling it. Three new APIs were added, TickGetDiv256(), TickGetDiv64K(), and TickConvertToMilliseconds(). Internally the tick counter is now 48-bits wide and as accurate as your Timer clock source, allowing you to use it as a Real Time Clock. 16.Added PIC24FJ64GA004_PIM hardware profile. This hardware profile is intended for use with the PIC24FJ64GA004 PIM on the Explorer 16 development board. In this mode, BUTTON2 and BUTTON3 and several of the LEDs do not work correctly due to lack of I/O pins on this device. Also, you cannot have the POT and TEMP jumpers on the PIM bridged because these signals are multiplexed with the SDO1/SDI1 pins needed for the Ethernet PICtail Plus. 17.Removed most ROM APIs when using a 16-bit compiler (C30). PIC24s and dsPICs usually don't need separate ROM functions since the Program Space Visibility feature maps ROM into RAM space. All ROM APIs are still supported, but they are now macros to base RAM APIs. This change saves a couple of kilobytes of code space on PIC24 and dsPICs. 18.Improved MyTCB structure caching. This should reduce TCP packet processing overhead with the ENC28J60 where TCBs are stored in the Ethernet RAM. 19.MAX_RETRY_COUNTS TCP configuration option has been renamed to TCP_MAX_RETRIES. 20.FTP server is no longer enabled by default. HTTP2 now supports POST, so you can upload new webpages through the /mpfsupload page now. FTP required two precious TCP sockets. 21.Began adding hooks for an SSL/TLS transport for secure HTTPS and other future stack modules. Note that these cryptographic modules are not available at this time. Configuration options such as MAX_SSL_CONNECTIONS do nothing and should not be modified. 22.Username has changed for all of the modules. Now all modules have a default username of "admin" and password of "microchip". Previously, the FTP and Telnet modules used "ftp" and "telnet" respectively for the usernames. 

Fixes:

1. Fixed a SendFile() bug in HTTP.c where parsing dynamic cgi files could send garbage back to the web browser sometimes. Thanks go to Matt Watkins on the Microchip Ethernet forum for identifying this issue.
2. Fixed an off by one error in the calculation of RESERVED_TCP_MEMORY. Previously, the last TCP socket's RX FIFO would incorrectly overlap with the Ethernet RX buffer, causing incoming packets to occasionally be corrupted or the incoming data on the last socket to get corrupted.
3. Fixed the QWORD_VAL's dword struct element types. dword.LD and dword.HD were incorrectly defined as WORDs instead of DWORDs. Thanks go to Iñaki Esparza on the Microchip Ethernet forum for identifying this issue.
4. Fixed the incorrect processing of received IP fragments with a non-zero offset. This stack does not support IP packet reconstruction due to the limited amount of available RAM. Thanks go to Iñaki Esparza on the Microchip Ethernet forum for noticing this behavior.
5. Board now only responds to ping requests to our IP address, the directed subnet broadcast address, or the broadcast address of 255.255.255.255. Previously, it would respond to any ping request to any IP address, assuming the MAC address was correct.
6. Fixed a memory corruption/UDP packet loss problem when handling incoming UDP packets. Previously, StackTask() would incorrectly continue processing more packets if it came upon a UDP packet. Thanks go to Iñaki Esparza on the Microchip Ethernet forum for identifying this issue.
7. Fixed the SMTPClient.ROMPointers.Server flag having an inverted meaning. Previously, the SMTP client module would treat the SMTPClient.Server pointer as a ROM pointer if this bit was cleared. In most cases, this would cause the SMTP client to return an error code of 0x8000 when the SMTPClient.SMTPServer address pointer was set.
8. Fixed the DHCP Server module from incorrectly parsing received packets which had a DHCP_PARAM_REQUEST_IP_ADDRESS option followed by more options. Previously due to the length miscalculation, the parser would enter a random state, depending on the packet's contents. Thanks go to Iñaki Esparza on the Microchip Ethernet forum for identifying this issue.
9. Fixed potential incorrect results when UDPIsGetReady() was called and a previous application did not call UDPDiscard() on an RX packet. Now, StackTsk() calls UDPDiscard() as appropriate to let it know when it's old RX data is being thrown away. This fixes a potential bug in the DHCP Server module and makes the UDP API more robust. Thanks go to Iñaki Esparza on the Microchip Ethernet forum for identifying the potential DHCP server issue.

10.Fixed a potential ARP bug where the Gateway's MAC address would be returned for an IP address on the local subnet. This unusual case would occur when two application tasks were using the ARP module at the same time and the second application was trying to resolve an IP address off of our subnet. Thanks go to Iñaki Esparza on the Microchip Ethernet forum for pointing this issue out. 11.Fixed an PIC18F97J60 family MAC layer bug where MACGetArray() might not correctly increment the Ethernet read pointer if a NULL pointer was given for the destination. The C compiler might have optimized the function so that it would increment the read pointer one less than it was supposed to. 12.The TCP module now acknowledges TCP Keep-Alive packets which will help prevent connection loss if the remote node fills up our RX FIFO and then our window-update packet gets lost on the network/Internet. In stack version 4.02, a zero-window probe would have been required to restore the communications. 13.Fixed a TCP RX FIFO corruption issue that would occur in (uncommon) circumstances when too many out-of-order segments arrived such that a second "hole" would have been required to accommodate the data. Thanks go to Iñaki Esparza and his eagle eyes on the Microchip Ethernet forum for finding this corner case bug. 14.Inline assembly in the ETH97J60.c file has been modified to accommodate the C18 Extended mode and C18 Auto default storage class. Previously, the Ethernet module would transmit garbage packets when using the C18 parameter stack. 15.Fixed potential buffer overflow in NBNS.c's NBNSGetName() function where an unexpected string length retrieved from the packet could cause random memory corruption. 16.Fixed some potential PIC18F97J60 family Ethernet module transmit lockup conditions that occur on some networks. Previously blocking while() loops would wait indefinitely for the ECON1bit to become clear by hardware, which the hardware might never have done. 17.In MainDemo.c, a call to DelayMs() was being made using a value of 100ms. This was too long for the underlying Delay1KTCYx() C18 function and would result in a shorter than expected delay when compiled with C18. This has been fixed with a loop. Thanks go to Andy123 on the Microchip Ethernet forum for pointing this problem out. 18.Fixed a potential C18 memory overlaying problem in the TickUpdate() function. Previously, the local variable used in this function might have been overlayed on other memory, resulting in random memory corruption as the ISR occurred. 19.The demo AJAX web pages in the TCPIP Demo AppWebPages folder now correctly display and self-refresh in Firefox 2. Previously, it would work in Firefox 1.5 and Microsoft Internet Explorer, but not Firefox 2. Thanks go to "gohsthb" on the Microchip Ethernet forum for identifying this correction. 20.Rewrote the GenericTCPServer.c example to not use an application RAM FIFO for buffering. Since the TCP module implements its own FIFOing, the application has limited need for its own FIFO too. This fixes a previous bug where the GenericTCPServer was not checking the number of incoming bytes with the remaining size available of the App FIFO. This would have previously resulted in a buffer overflow, corrupting the RX data if too much arrived all at once. 21.Fixed a potential MPFS classic inline ASM30 assembly code problem where web pages stored in internal Flash and C30 with optimizations enabled could result in data corruption. 22.Fixed a UDPPut() tracking problem that would result in extra bytes being appended to the end of a packet if the UDPSetTxBuffer() function was used. This previously caused the SNMP module to send some junk data at the end of its packets. 23.Fixed a potential TCP problem where transmitted FIN packets might not get retransmitted properly if the remote node never acknowledged the data that was transmitted just before the FIN was sent. 24.Fixed a NetBIOS Name Service bug where the response packet would sometimes get sent to an incorrect address. It now consistently responds to the unicast MAC/IP address of the NBNS query packet. 25.Added padding to all transmitted DHCP messages to make the minimum UDP payload at least 300 bytes. This fixes compatibility with some older BOOTP relay devices which discard smaller packets. Thanks go to Dave Collier on the Microchip Ethernet forum for pointing this problem out. 26.Substantially shrunk the number of retransmission attempts made in the TCP_SYN_RECEIVED state. This improves recovery time when attacked by a SYN flood Denial of Service event. The recovery time is now 7 seconds (3 total packets) instead of 31 seconds (6 total packets) 27.Fixed the possibility of the NetBIOS Name Service module giving out the board's static IP address before a DHCP lease could be obtained. NBNS requests are now only serviced when originating from nodes on the same subnet. 28.Fixed storage of MPFS classic in internal program memory when using the HI-TECH PICC-18 compiler. 29.Substantially revised TCP.c, fixing many TCP bugs and possibly adding new ones. Thanks go to Michael Rubinstein for finding several of these TCP problems. 30.The DNS client module will now time out and return failure if the DNS server cannot be ARPed or does not respond to the DNS query. Each timeout is set to 1 second and 3 total ARP and 3 total DNS query attempts are possible. Previously, it would retry indefinitely, causing the calling application to deadlock. 

Known Problems:

1. Telnet server module does not implement a lot of Telnet functions. As a result, it will likely not display correctly or work at all with some Telnet clients. The server was tested with the Microsoft telnet.exe utility which is provided with Microsoft Windows.
2. TFTPc module has not been tested with this version.
3. If the DHCP client and DHCP server are used at the same time and you connect two similar boards to each other (ex: two PICDEM.net 2 boards connected via a crossover cable), a race condition can occur where both nodes will disable their DHCP server and neither board will get a successful DHCP lease. If this unlikely scenario occurs, as a work around, simply reset one of the boards to reenable it's DHCP server.
4. HI-TECH PICC-18 projects may not compile when targeting the external ENC28J60 chip on the PICDEM.net 2 development board (instead of the internal Ethernet controller). This problem only applies when a PIC18F97J60 family part is the target. I.e. it compiles correctly for the HPC_EXPLORER + Ethernet PICtail.
5. HI-TECH PICC-18 projects will not correctly set the processor configuration fuses through code using the __CONFIG() macro. Ensure that the configuration fuses are manually set correctly via the MPLAB IDE Configuration Bits dialog. This problem has been observed with compiler version 9.50PL3.

Testing and Performance Notes:

1. Make sure to use MPLAB IDE 7.62 or higher with this version. Versions below 7.61 will not work. Version 7.62 has cool new features like C auto-word complete and function parameter tooltips that can be enabled (disabled by default).
2. Testing was done using MPLAB C18 version 3.12, MPLAB C30 version 3.01, and HI-TECH PICC-18 version 9.50PL3. Make sure to upgrade your tools to at least these versions.

v4.02 10 April 2007

IMPORTANT NOTE: You must use MPLAB 7.41 or higher to successfully open the MPLAB projects. IMPORTANT NOTE2:If an external serial EEPROM memory is used to store AppConfig, it's contents will be invalidated the first time you run this version, restoring the AppConfig defaults. The AppConfig structure has been optimized. IMPORTANT NOTE3:If an external serial EEPROM memory for MPFS, you will need to recreate the MPFS image and program your EEPROM. A 32 bit addressing format is now used. 

Changes:

1. Implemented TCP RX packet order correction logic. The stack can now accept TCP frames that arrive out-of-order without requiring the remote node to go through a retransmit cycle. This dramatically improves RX performance when communicating over the Internet.
2. UDPOpen() now can handle a NULL pointer for remoteNode. In this case, the broadcast IP/MAC addresses will be used for the remoteNode (destination address of outbound packets).
3. Recreated MPLAB projects for the HI-TECH PICC-18 compiler. These were temporarily absent from 4.00RC. This project works with the PIC18F97J60 with internal Ethernet module, assuming the correct compiler version is present.
4. Moved all the headers around. Most of them are in "Microchip SolutionsMicrochipIncludeTCPIP Stack" now. This change was made to again be more compatible with other (future) Microchip software libraries.
5. New UDPPut() behavior. Now, if space in the Ethernet TX buffer runs out, the packet will not automatically be transmitted. You must call UDPFlush() to cause the packet to be transmitted.
6. Added UDPGetArray(), UDPPutArray(), UDPPutROMArray(), UDPPutString() and UDPPutROMString() user API functions. These functions perform substantially better than calling UDPPut() successively and allow greater application programming flexibility.
7. Changed TCPPutString() and TCPPutROMString() APIs to now return an updated string pointer instead of a count of bytes successfully placed in the TX buffer.
8. Added UDPPerformanceTest.c. By default this module causes UDP packets containing 1024 bytes of application data to be broadcasted on UDP port
   12345. Use a packet sniffer, such as Wireshark (http://www.wireshark.com/)

to capture and derive stack overhead/UDP TX performance characteristics with this module. Note that this test uses the UDPPutROMArray() function. Applications which use successive calls to UDPPut() will be slower. To enable this module, #define STACK_USE_UDP_PERFORMANCE_TEST in TCPIPConfig.h.

9. Added TCPPerformanceTest.c. By default this module listens on TCP port
   12345. When a remote client connects, this server module will being

transmitting the maximum possible amount of application data that it can, given your TCP TX FIFO size. Use a packet sniffer, such as Wireshark (http://www.wireshark.com/) to capture and derive stack overhead/TCP TX performance characteristics with this module. Any TCP client can be used, including readily available utilities such as the telnet.exe utility available on Microsoft Windows XP. To use it to connect to the test module, run: "telnet.exe xxx.xxx.xxx.xxx 12345" where xxx.xxx.xxx.xxx is the board's IP address. Note that this test uses the TCPPutROMArray() function. Applications which use successive calls to TCPPut() will be slower. To enable this module, #define STACK_USE_TCP_PERFORMANCE_TEST in TCPIPConfig.h. 10.Added Reboot.c module. By default, this module listens on UDP port 30304. If the application byte 0x00 arrives on this port, the PIC will reset. This is primarily useful for remote Bootloader entry. #define STACK_USE_REBOOT_SERVER in TCPIPConfig.h to enable this module. Note that since no encrypted challenge/response algorithm is currently implemented, this module is a Denial of Service vulnerability, so it should not be enabled unless there is a specific need for it. 11.Made the TickUpdate() ISR routine execute in the low priority ISR instead of the default high priority ISR. The Microchip TCP/IP stack does not need any interrupts except this low priority timer. 12.Renamed STACK_USE_DHCP macro to STACK_USE_DHCP_CLIENT 13.Added STACK_USE_MPFS macro. 14.Changed UDPIsPutReady() to return a WORD instead of a BOOL. The WORD is the number of bytes that can be put into the buffer. 15.Changed MACGetArray() to accept a NULL pointer. If NULL, the retrieved data will simply be discarded. This also changes the behavior of UDPGetArray() and TCPGetArray() to match, throwing bytes away if a NULL pointer is given. 16.Added a very simple DHCP Server module. This module has limitations and is useful for a single client only. Its purpose is to allow you to directly connect the board to a standard PC through a crossover cable (no other network nodes attached). The server is coded to automatically disable itself if the DHCP client is also enabled and another DHCP server is detected on the network. This allows both the DHCP server and DHCP client to coexist without any manual reconfiguration. 17.Added DNSResolveROM() function for resolving host names that are stored in program memory, ex: literal strings. 18.Added a TCP automatic transmit/window update timer. It defaults to TCP_AUTO_TRANSMIT_TIMEOUT_VAL (40ms) after the first get or put operation following the last automatic transmit/window update. This timer enhances performance, especially when streaming data over the Internet where round trip times can be several tens to low hundreds of milliseconds. This also improves application coding flexibility as TCPFlush() need not be called anymore. 19.Added TCP delayed ACKnowledgement timer. This conserves bandwidth by transmitting fewer ACKs and prevents inadvertently influencing remote slow start/collision avoidance and fast retransmit algorithms. 20.Completely rewrote ICMP (ping) server module. It is now much smaller (ROM and RAM), faster, and can handle packets of 576 bytes or larger, if no IP fragmentation occurs. 21.Rewrote StackTsk() stack manager. It is much simpler now. 22.Added TCPFind(), TCPFindArray(), and TCPFindROMArray() user API functions. These functions peek inside a given TCP socket's RX FIFO (without removing anything) and looks for a particular byte or array of bytes. This should greatly simplify the creation of application code whenever variable length fields are used (ex: text strings terminated by rn). It supports case insensitive text searching or binary searching, as well as an offset to start searching at. 23.Added TCPGetRxFIFOFree() user API. It returns the number of bytes of free space in the TCP's RX FIFO. 24.Changed default TICK resolution to 1ms (from 10ms) and improved accuracy. 25.Added outbound ping capabilities (i.e. board can now ping another board or a PC). To enable these features, define STACK_USE_ICMP_CLIENT. This will enable several new APIs, including ICMPBeginUsage(), ICMPSendPing(), ICMPGetReply(), and ICMPEndUsage(). The functions should be called in this order. See the PingDemo() function in MainDemo.c for an example of how to use them. By default, pushing BUTTON3 (left-most one) will cause a ping to be sent to 4.78.194.159 (ww1.microchip.com). The response time will be displayed on the LCD (assuming your development board has an LCD). 26.Cleaned up C30 3.00 signed/unsigned warnings. 27.Removed PIC18F97J60_TEST_BOARD hardware profile support. This stack no longer supports it due to the old beta silicon (with errata) mounted on these boards. 28.Added support for ROM pointers for all of the SMTP strings (To, From, CC, Subject, etc.). If you use a ROM string, you must also set the corresponding SMTPClient.ROMPointers.xxx bit to let the SMTP module know which type of pointer was provided. See the SMTPDemo() code in MainDemo.c for and example calling sequence using both ROM and RAM strings for the various fields. 

Fixes:

1. Fixed a critical TCP buffer corruption issue where the start of a TCB header overlapped with the last byte of the RX FIFO from the previous socket. This bug affected version 4.00RC only.
2. ETH97J60.c, TCPIP.h, and TCPIP Stack Version.txt were correctly readded to the TCPIP Demo App-C18 project using relative paths instead of absolute paths.
3. UDPOpen() now dynamically assigns a local port number if you call it and give it a 0x0000 port number. This should fix some UDP applications from not working (ex: DNS Client module) with some computers/routers/networks which throw away traffic originating from the invalid port 0x0000 value.
4. Fixed a ENC28J60 bank selection error that would occur if an application called GetCLKOUT() in ENC28J60. By default, this function is not called.
5. UnencodeURL() function in Helpers.c is now tested and working.
6. Fixed a TCP Window Update problem when TCPGetArray() was used. Before the problem was fixed, performance could have been terrible on reception.
7. Fixed a unintended TCP connection close if the socket was idle for about a minute. Now, TCP sockets will remain open indefinitely if there is no traffic going on.
8. Serial numbers >32K are now displayed correctly on the serial port as a positive value when C18 is used and the board is placed in configuration mode (BUTTON0 is depressed on power up).
9. HI-TECH PICC-18 compiler would previously incorrectly initialize the AppConfig structure.

10.Previously a processor reset was possible when accessing items in the AppConfig strucutre on 16 bit MCUs (PIC24, dsPIC) due to unaligned word accesses. This was fixed by reordering the Flags byte in the APP_CONFIG structure. 11.Rewrote DHCP client state machine, fixing the previously known problem where it would not perform a new discovery if it was trying to renew a lease with an offline DHCP server. 12.Fixed a critical deadlock problem in the ETH97J60.c MAC layer driver for the PIC18F97J60 family Ethernet controller. Previously, it was possible (although rare) that the DMAST or TXRTS bits would get stuck set if too much Ethernet traffic was received within a short interval. Previously, the MACFlush() function was unnecessarily setting TXRST, which it should not do while the Ethernet interface or DMA is being used. 13.Fixed an HTTP server state machine problem where a new connection occurring too soon on a previously used socket could cause the HTTP server to no longer respond. 14.Fixed a potential memory corruption error in the HTTPGetVar() callback which would exceed the bounds of the VarString array when returning the VAR_STACK_DATE variable. 15.Fixed a TCP transmission sequence tracking problem whenever data is retransmitted and new unflushed data is also in the TX FIFO. Thanks go to Matt Watkins on the Microchip Ethernet forum for identifying this issue. 

Known Problems:

1. RTL8019AS MAC layer driver has not been updated for new TCP module. Users requiring RTL8019AS support should continue to use stack version 3.75.
2. I2CEEPROM.c has not been tested or completed. Continue to use I2CEEPROM.c from stack version 3.75 if this file is needed.
3. Telnet server module does not implement a lot of Telnet functions. As a result, it will likely not display correctly or work at all with some Telnet clients. The server was tested with the Microsoft telnet.exe utility which is provided with Microsoft Windows.
4. TFTPc module has not been tested with this version.
5. The default demo web pages which use AJAX do not automatically refresh themselves when viewed in Firefox 2.0.0.1. Earlier Firefox versions (1.5ish) probably work without any problem.
6. Files may be inaccessible in your MPFS if compiled with C18 for internal flash program memory and your total MPFS content is large (around 64KB or larger). The code attempts to access the ROM memory using a near rom pointer when a far rom pointer is needed.
7. If using MPLAB 7.52 all .s files that are compiled with C30 will not have the corresponding object file get stored in the correct directory. As a result, if you are compiling with C30 and with MPFS_USE_EEPROM not defined (i.e. storing web pages in internal program memory), the project won't link (throws a undefined reference to `MPFS_Start'). As a workaround, remove the Intermediates Directory in the MPLAB project. Alternatively upgrade MPLAB to a newer version. MPLAB IDE 7.60+ may have this fixed.
8. If the DHCP client and DHCP server are used at the same time and you connect two similar boards to each other (ex: two PICDEM.net 2 boards connected via a crossover cable), a race condition can occur where both nodes will disable their DHCP server and neither board will get a successful DHCP lease. If this unlikely scenario occurs, as a work around, simply reset one of the boards to renable it's DHCP server.
9. HI-TECH PICC-18 projects may not compile when MPFS_USE_EEPROM is not defined and you are trying to store web page data in internal FLASH program memory.

10.HI-TECH PICC-18 projects may not compile when targeting the external ENC28J60 chip on the PICDEM.net 2 development board (instead of the internal Ethernet controller). This problem only applies when a PIC18F97J60 family part is the target. I.e. it compiles correctly for the HPC_EXPLORER + Ethernet PICtail. 

Testing and Performance Notes:

1. This stack version was compiled and tested with the following tool versions: -MPLAB IDE 7.52 -Microchip C30 version 3.00 -Microchip C18 version 3.10 -HI-TECH PICC-18 version 9.50PL3
2. Using the UDPPerformanceTest.c module, the stack can transmit around 220KBytes/second (1.75Mbits/second) of UDP application data on the PIC18F97J60 with internal Ethernet @ 41.66667MHz core clock, compiled using C18 3.10 with debug optimization settings.
3. Using the UDPPerformanceTest.c module, the stack can transmit around 392KBytes/second (3.14Mbits/second) of UDP application data on the PIC24HJ256GP610 with external ENC28J60 @ 40 MIPS, compiled using C30 3.00 with debug optimization settings.
4. Using the TCPPerformanceTest.c module, the stack can transmit around 58KBytes/second (464Kbits/second) of TCP application data on the PIC18F97J60 with internal Ethernet @ 41.66667MHz core clock, compiled using C18 3.10 with debug optimization settings, over Ethernet when using a tiny 200 byte TX TCP FIFO. Note that performance can be improved significantly by increasing the FIFO size and performance will drop significantly if the round trip TCP acknowledgement time is increased (ex: testing over the Internet instead of Ethernet).
5. Using the TCPPerformanceTest.c module, the stack can transmit around 69KBytes/second (558Kbits/second) of TCP application data on the PIC24HJ256GP610 with external ENC28J60 @ 40 MIPS, compiled using C30 3.00 with debug optimization settings, over Ethernet when using a tiny 200 byte TX TCP FIFO. Note that performance can be improved significantly by increasing the FIFO size and performance will drop significantly if the round trip TCP acknowledgement time is increased (ex: testing over the Internet instead of Ethernet).
6. Using the TCPPerformanceTest.c module, the stack can transmit around 178KBytes/second (1.42Mbits/second) of TCP application data on the PIC24HJ256GP610 with external ENC28J60 @ 40 MIPS, compiled using C30 3.00 with debug optimization settings, over Ethernet when using a larger 2000 byte TX TCP FIFO. Note that performance will drop significantly if the round trip TCP acknowledgement time is increased (ex: testing over the Internet instead of Ethernet).

v4.00RC 28 December 2006

IMPORTANT NOTE: If an external serial EEPROM memory is used to store AppConfig, it's contents will be invalidated the first time you run this version, restoring the AppConfig defaults. The AppConfig structure has been optimized. IMPORTANT NOTE2: If an external serial EEPROM memory for MPFS, you will need to recreate the MPFS image and program your EEPROM. A 32 bit addressing format is now used. 

Changes:

1. Added Simple Mail Transfer Protocol (SMTP) client module and updated MainDemo.c to exercise the Email transmission functionality when a user pushes BUTTON0.
2. Added beta Telnet server module. See Known Problems section.
3. Completely revamped the TCP module. A real transmit FIFO and receive FIFO are allocated for each TCP socket now. This greatly enhances RFC compliance, communications robustness, and makes application development easier. New APIs were added for putting and getting arrays and strings (including ROM variants). Several TCP related bugs are now fixed as a result. Please report any bugs found in the new implementation.
4. Added TCPPutArray() API.
5. Added TCPPutROMArray() API.
6. Added TCPPutString() API.
7. Added TCPPutROMString() API.
8. Added TCPGetArray() API.
9. Changed TCPIsPutReady() API. Instead of returning a BOOL, it now returns a WORD. The WORD is a count of the number of bytes that TCPPut(), TCPPutArray(), etc. can immediately place in the output buffer. MAKE SURE THAT YOUR CODE DOES NOT COMPARE THE RETURN RESULT OF TCPIsPutReady() DIRECTLY TO TRUE. For example, "if(TCPIsPutReady(MySocket) == TRUE){...}" must be converted over to: "if(TCPIsPutReady(MySocket)){...}".

10.Changed TCPIsGetReady() API. Instead of returning a BOOL, it now returns a WORD. The WORD is a count of the number of bytes that TCPGet() or TCPGetArray() can immediately obtain. MAKE SURE THAT YOUR CODE DOES NOT COMPARE THE RETURN RESULT OF TCPIsGetReady() DIRECTLY TO TRUE. For example, "if(TCPIsGetReady(MySocket) == TRUE){...}" must be converted over to: "if(TCPIsGetReady(MySocket)){...}". 11.Changed TCPDiscard() return type from BOOL to void. 12.Removed TCP_NO_WAIT_FOR_ACK option. It was defaulted to disabled in the last two releases of the stack and is not needed with the new TCP module. 13.Updated DNS module to include two new required APIs: DNSBeginUsage() and DNSEndUsage(). These functions control a one bit ownership semaphore to allow multiple applications to use the DNS module in series. If invoked correctly, this will prevent unintended bugs resulting from two applications trying to use the DNS module at the same time. Old applications, such as those based around the GenericTCPClient.c example must be updated to use these functions. 14.Started using a new project structure and folders. You must use MPLAB 7.41 or higher (stack is tested on MPLAB 7.50) to use the default workspaces/projects, which include files using relative paths. This should improve compatibility with some future code libraries released by Microchip. StackTsk.h was broken into TCPIPConfig.h, HardwareProfile.h, and StackTsk.h. TCPIPConfig.h now includes all stack configuration options and HardwareProfile.h contains all hardware options. No macros need be globally defined in MPLAB project now. TCPIP.h is the only header applications must include now, for any/all modules used. 15.Combined ARP.c/ARP.h and ARPTsk.c/ARPTsk.h into a single file pair: ARP.c/ARP.h. Applications built using a prior stack revision must remove all instances including "ARPTsk.h" and replace it with "ARP.h" instead. The ARP module is now simpler, more linear (easier to read), and being in one source file, allows the C compiler to optimize better. 16.Added PIC18F67J60_TEST_BOARD hardware profile to HardwareProfiles.h. This hardware profile is designed for 05-60091 (Rev 1), a development board that is not in production at this time. 17.Added DSPICDEMNET1 and DSPICDEMNET2 hardware profiles to HardwareProfiles.h for eventual support of the Microchip dsPICDEM.net 1 and dsPICDEM.net 2 demo boards. These two boards use the RTL8019AS Ethernet controller and a 24LC515 EEPROM. These changes are currently incomplete and these profiles cannot be used. 18.Began rewriting I2CEEPROM.c to support 16 bit CPUs, including the dsPIC30F6014 used on the dsPICDEM.net 1 and 2 demo boards. Note that work here is incomplete and cannot be used as a result -- see Known Problems section. 19.Partially updated RTL8019AS.c to support 16 bit CPUs, including the dsPIC30F6014 used on the dsPICDEM.net 1 and 2 demo board. Note that work here is incomplete and cannot be used as a result -- see Known Problems section. 20.Updated SNMP.c to use new typedefs in GenericTypedefs.h. Also SNMP was tested in this version. SNMP.mib was updated some to better reflect current hardware. 21.Added AN870 SNMP callbacks to MainDemo.c (a feature that was missing in 3.xx releases). This code will get compiled when STACK_USE_SNMP_SERVER is defined in TCPIPConfig.h. 22.Removed all instances of MPFS_USE_PGRM for storing in internal FLASH program memory. Storage in internal program memory is now the default. Define MPFS_USE_EEPROM to override the default and store MPFS in an external EEPROM memory. 23.Decreased program memory needed for Announce.c module by about 180 bytes. Multiple inline calls to UDPPut() were removed. 24.UDP checksum checking logic has been improved. The UDP layer now avoids writing the pseudo header checksum in the RX buffer. 25.Swapped endianess of the returned checksum from CalcIPBufferChecksum(). Rewrote CalcIPBufferChecksum() in Helpers.c. This improves consistency. 26.Improved swapl() in Helpers.c. 27.Improved USART baud rate (SPBRG) calculation for PIC18s. Rounding is now done to chose the most optimal value and the code will automatically select high baud rate mode (BRGH=1) if possible. Additional improvements can be made if using a newer PIC18 with the 16 bit baud rate generator. 28.Added GenericTCPServer.c example file to complement GenericTCPClient.c. The server is enabled by defining STACK_USE_GENERIC_TCP_SERVER_EXAMPLE in TCPIPConfig.h. 29.Renamed STACK_USE_GENERIC_TCP_EXAMPLE definition to STACK_USE_GENERIC_TCP_CLIENT_EXAMPLE for consistency with new server example. 30.Defaulted MPFS.exe to generate binary MPFS images using 32 bit addressing. MPFS.h has been modified to also default to use 32 bit addressing of external EEPROM images. You must rebuild any old MPFS images and reprogram them if upgrading from a previous TCP/IP stack revision, which defaulted to use 16 bit addressing. 31.Updated MPFS.exe to #include "TCPIP.h" instead of "..HeadersCompiler.h" in C files generated by the utility. 32.Added MPFSv2.exe PC utility for generating large MPFS images in program memory (ASM30 code) for C30 users. Previously, the C30 compiler placed a limit of less than 32KB of total MPFS size due to the PSV window size limitation on PIC24/dsPIC devices. To get around the limitation, use the new MPFSv2.exe utility to generate an .s file which can be included in your project instead of the .c file generated by the traditional MPFS.exe utility. 

Fixes:

1. Fixed a bug in ARPProcess() which would incorrectly send an ARP response to an incorrect MAC & IP address if a TX buffer wasn't immediately available.
2. Fixed a TCP bug where TCPIsGetReady() would return TRUE even if no data was left in the recieved packet. Previously you had to call TCPGet() one last time and have it fail before TCPIsGetReady() would return FALSE.
3. Modified TCP state machine. Established connections will no longer automatically close if left idle for approximately 45 seconds. Note that your application needs to ensure that no sockets unintentionally get lost (For example: a server socket that received data only is established and the cable breaks while connected. In this case, the socket would never be detected as being disconnected since the server never attempts to transmit anything).
4. Stopped overclocking dsPIC33 and PIC24H devices. Previously PLLFBD was incorrectly set to 39 instead of 38 to yield a resulting Fosc of 84MHz (42MIPS) instead of 80MHz (40MIPS) with the default Explorer 16 development board. Thanks go to Matt Watkins on the Microchip Ethernet Forum for pointing this error out.
5. Corrected a bug in IP.c where IPHeaderLen would not be properly initialized if a NON_MCHP_MAC was used (ex: RTL8019AS) and IPSetRxBuffer() was called. This bug did not affect ENC28J60 or PIC18F97J60 family support. Thanks go to Darren Rook for identifying this issue.
6. Updated checksum checking code in ENC28J60.c for latest silicon DMA checksum errata.
7. Declared TickCount in Tick.c/Tick.h as volatile and implemented an interrupt safe reading procedure in TickGet(). Since this multibyte variable is modified in the ISR and read in the mainline code, these changes are needed to prevent rare inconsistency bugs.
8. Fixed Announce.c so the unicast remoteNode of the requesting packet would be used rather than the remoteNode of the last received packet, which may not be correct when transmitting. Thanks go to Brett Caulton for identifying this issue.
9. Fixed a DHCP bug which would cause DHCP renewals to continually occur after only 60 seconds once the original lease expired. Thanks go to Brett Caulton for identifying this issue and fix.

10.Fixed a potential TCP socket leak in the FTP module. Previously FTPDataSocket would not be reliably initialized nor closed if the connection was killed forcefully (user killed application, cable disconnected while transferring, etc.). 

Known Problems:

1. RTL8019AS MAC layer driver has not been updated for new TCP module. Users requiring RTL8019AS support should continue to use stack version 3.75.
2. I2CEEPROM.c has not been tested or completed. Continue to use I2CEEPROM.c from stack version 3.75 if this file is needed.
3. Telnet server module is still in development. No user authentication features are currently implemented. Some telnet clients may render the telnet server output incorrectly (in the wrong locations or wrong colors). Testing has only been done with the Microsoft Windows telnet.exe utility that comes Windows XP.
4. DHCP will continually send out DHCP Request packets when the lease expires and the original DHCP server that gave the lease is offline. The board will continue to use the expired IP address until the DHCP server comes back online, at which point the lease will be renewed or a new discovery will occur. A new discovery should occur after timing out, instead. It is believed that this problem has always existed in previous stack revisions.
5. DHCP will continually send out DHCP Request packets when the lease expires and the original DHCP server that gave the lease does not include Option 54, the Server Identifier. A new discovery should occur after timing out. It is believed that this problem has always existed in previous stack revisions.
6. TFTPc module has not been tested with this version.
7. Serial numbers >32K will be displayed on the serial port as a negative value when C18 is used and the board is placed in configuration mode (RB0 button is depressed on power up).

v3.75 14 August 2006

Changes:

1. Added beta DNS client module (DNS.c). DHCP was also updated to obtain a DNS server address. Added AppConfig.PrimaryDNSServer IP address. Added STACK_USE_DNS configuration macro. To use the DNS client, call DNSResolve() with the server name, ex: DNSResolve("www.microchip.com"), and then periodically call DNSIsResolved() until it returns TRUE, ex: DNSIsResolved(&IPAddressDestination). Only one DNS resolution can be in progress at a time. Because the DNS client is a beta module, the API or code may change before being finalized. No formal DNS API documentation is available yet.
2. Added beta NetBIOS Name Service responder module (NBNS.c). Added AppConfig.NetBIOSName string. Added STACK_USE_NBNS configuration macro. Added MY_DEFAULT_HOST_NAME macro in StackTsk.h. Now, whenever a NetBIOS broadcast attempting to resolve AppConfig.NetBIOSName arrives, a response will be made. This form of name resolution only works on a single subnet. Off the subnet, manual registration in a DNS server or other means will be needed to allow the local Host Name to be recognized and translated to an IP address. The default NetBIOS name for the board is "MCHPBOARD". To test the NetBIOS Name Service module, try entering http://MCHPBOARD/ into your web browser instead of the board's IP address.
3. Added beta HTTP client module (GenericTCPClient.c). This module demonstrates how to make a TCP client application. To test this module, uncomment the STACK_USE_GENERIC_TCP_EXAMPLE macro in StackTsk.h, recompile, and then press the BUTTON1 button while the stack is running. RemoteURL[] should be downloaded from ServerName[] and written to the UART. For the default values of ServerName[] and RemoteURL[], the HTML search page for "Microchip" will be fetched from "www.google.com" and written to the serial port. No formal documentation is available for this example yet.
4. Added Embedded Ethernet Device Discoverer PC project to aid in embedded product discovery when connected to a network and demonstrate how to write PC applications which can communicate with embedded devices. The source code for this device is included. It can be built using the Microsoft Visual C# 2005 Express Edition compiler. At the time of stack release, this 3rd party PC development tool can be downloaded at no cost from http://msdn.microsoft.com/vstudio/express/. If using only the Microchip Device Discoverer executable file without the Visual C# compiler, the .NET Framework 2.0 must be installed on the local PC. The application setup utility should allow dynamic downloading of this component if the target machine does not already have it installed.
5. Updated Announce.c to listen and respond to discovery requests sent to UDP port 30303 starting with the character 'D'. To test this functionality, use the Embedded Ethernet Device Discoverer on a PC connected to the same subnet.
6. Updated UART configuration menu to accommodate the new beta module configuration options (DNS server address, device host name).
7. Increased MPFS reserve block to 64 bytes from 32. Also, because the APP_CONFIG structure was updated, all current MPFS images and data stored in deployed EEPROMs needs to be updated.
8. Added a means to erase (invalidate) the onboard EEPROM using the BUTTON0 momentary switch (right-most switch on demo boards with multiple switches). To erase the EEPROM, hold down BUTTON0, RESET the board (press and release MCLR switch), and then continue to hold down BUTTON0 for an additional 4 seconds. If you press MCLR again, the EEPROM contents will now be invalid. If you press '0' on the UART, the same configuration that was read prior to invalidating the contents will be written back into the EEPROM. Invalidating the EEPROM allows the MY_DEFAULT_* constants to get loaded into a previously programmed EEPROM chip. Because of change #7, this procedure should be done for all currently programmed EEPROMs to prevent anomalous values from being read.
9. remoteNode in StackTsk.c was changed from private to global scope. Now external modules can reference the address of the last received packet. Announce.c uses this to send a unicast response to a broadcast discovery request.

10.All stack modules that can be disabled (DHCP.c, FTP.c, etc) now will no longer emit a compiler error if you have it in the project without defining the appropriate macro (STACK_USE_DHCP, STACK_USE_FTP, etc). It will simply generate no machine code when compiled and the stack will not use that module. Make sure the proper macro is defined for each module that you wish to use. 11.Added SetRXHashTableEntry() to ENC28J60.c. This function can be used to set the appropriate bit in the Hash Table registers to join a particular multicast group. 12.Added Realtek RTL8019AS Ethernet controller support to the stack. MAC.c was renamed to RTL8019AS.c. This Ethernet controller is not recommended for new designs. RTL8019AS support was reintroduced to provide ongoing assistance to former Application designs implementing this chip. For new applications, use the Microchip ENC28J60 or PIC18F97J60 family of microcontrollers. 13.Added I2C EEPROM support for MPFS storage. In older 2.xx stack revisions, I2C EEPROM was supported by the XEEPROM.c file. This file has been renamed to I2CEEPROM.c. It is mutually exclusive with SPIEEPROM.c, and only one may be included in the project at a time. 14.Added new hardware definitions to Compiler.h. Pin mappings for the PICDEMNET and PIC18F97J60_TEST_BOARD boards have been added. FS_USB was also defined; however, it is untested and not recommended. See Compiler.h. The PIC18F97J60_TEST_BOARD is a non-production board that some Early Adopters of the PIC18F97J60 family parts have. 15.Changed type definitions for BYTE_VAL, WORD_VAL, DWORD_VAL, and moved the generic typedefs to GenericTypeDefs.h from StackTsk.h. This should improve compatibility with some future code libraries released by Microchip. 16.LCDBlocking.c module was modified to support 4-bit interfaces to LCD modules. The PICDEM.net board has the module wired using a 4-bit bus. 

Fixes:

1. Fixed a serious MAC TXBuffer leak in TCP.c. Previously TCP.c would allocate a buffer for each socket in use, but under heavy traffic conditions (ex: user holds down F5 on web browser), the buffer handle might have been discarded before releasing the buffer. As a result all TCP connections would have lost the ability to send any application data after the TXBuffer pool ran out.
2. In the TCP_SYN_SENT TCP state, ACKs may only be received (as opposed to SYN+ACK packets) if the remote node thinks the connection is already open. A RST is now sent in response to an unexpected ACK, which may improve reconnection time when this (rare) condition occurs.
3. A bug was present in the UDP module where remote MAC addresses would be cached for each socket, even when UDPInit() or UDPClose() was called, or the microcontroller was reset. As a result, responses to incoming packets could have been sent to the wrong MAC address. UDP Sockets are now properly initialized/closed.
4. Fixed a potential timing bug in LCDBlocking.c. For lower values of CLOCK_FREQ, insufficient delay time was given to the LCD module, potentially causing improper operation.
5. Changed PIC24F to default to the XT oscillator fuse rather than HS. The PIC24FJ128GA010 data sheet, rev. C reports that 8MHz should be used with XT mode, not HS mode like prior data sheets.
6. Added a couple of wait states to the Realtek RTL8019AS MAC layer module for NICPut() and NICGet(). Previously, the PICmicro could not operate above approximately 25MHz without losing communication with the RTL8019AS chip.
7. Updated PC based MPFS utility. When generating C files to be added to your MPLAB project, the include path to "Compiler.h" is now "..IncludeCompiler.h". The output file, ex: "MPFSImg.c" should be placed in the "Source" subfolder before compiling. For example, if you are in the main stack folder with the MPLAB projects, type: "mpfs /c WebPages SourceMPFSImg.c"
8. IP Gleaning will now get properly disabled when, through the RS232 configuration application, DHCP and IP Gleaning are disabled. The stack will still respond to ping requests which have the wrong destination IP address, but a correct MAC address. However, the stack will continue to keep its statically defined IP address when DHCP/IP Gleaning are disabled and the ping arrives.
9. SPIEEPROM.c now saves and reconfigures the EEPROM_SPICON1 register (SSPCON1) before reading or writing to the SPI. After the read/write, it restores the saved state. This allows the SPI bus to operate at different speeds, depending on what peripheral is being accessed if other devices share the bus and can support different speeds. In particular, this fixes the SPI @ 10.4MHz problem on the PICDEM.net 2 board when using the ENC28J60.

Known Problems:

1. DHCP will continually send out DHCP Request packets when the lease expires and the original DHCP server that gave the lease is offline. The board will continue to use the expired IP address until the DHCP server comes back online, at which point the lease will be renewed or a new discovery will occur. A new discovery should occur after timing out, instead. It is believed that this problem has always existed in previous stack revisions.
2. DHCP will continually send out DHCP Request packets when the lease expires and the original DHCP server that gave the lease does not include Option 54, the Server Identifier. A new discovery should occur after timing out. It is believed that this problem has always existed in previous stack revisions.
3. When an MPFS .c image file is added to a C30 project, a linking error reporting insufficient contiguous .const memory may occur when too much data is in the MPFS image (PSV window size limitation). Using the PSV window, 1 out of every 3 program memory bytes is wasted.
4. MACSetPMFilter(), MACDisablePMFilter(), and MACCopyRxToTx() have not been tested and possibly do not work.
5. SNMP, TFTPc modules have not been tested with this version.
6. Serial numbers >32K will be displayed on the serial port as a negative value when C18 is used and the board is placed in configuration mode (RB0 button is depressed on power up).
7. The C30 linker may misplace the __CONFIG2 section or disallow usage of MPFS images that are too big (add too much to the .const code section). The consequences of this are that the first configuration word at 0x157FC may not get set through code (must use the Configuration Bits dialog instead), and/or the project will not compile. This problem has been observed with C30 ver. 2.02 on the PIC24FJ128GA010 product. To work around this problem, the p24FJ128GA010.gld linker script has been modified. Specifically, line 68 has been commented out, which causes the linker to place all .text sections after placing all absolute sections. SSR 25966 in the C30 2.02 release notes may be related.
8. It is observed with the Realtek RTL8019AS Ethernet controller and the demo AJAX web page which self refreshes rapidly, that occasional HTTP GET requests sent by the computer do not get received by the HTTP server. This is believed to be a RTL8019AS MAC layer bug. The TCP protocol handles the packet loss, but application performance suffers while waiting for the TCP retransmission. This problem is not observed with ENC28J60.c or ETH97J60.c MAC layers.
9. The HI-TECH compiler version 9.50PL1 crashes when compiling LCDBlocking.c with 4 bit mode (PICDEMNET) and using a warning level of -3 or higher. To work around the problem, the HI TECH projects were set to use warning level -4.

Guiding Notes:

1. To use the stack on a classic PICDEM.net demo board with the Realtek Ethernet controller, a PIC18F452 processor, and Microchip C18: -Use the C18EEPROM MPLAB project -Change the processor in the MPLAB IDE -Change linker script to "18f452i.lkr" in the MPLAB project. Use the one provided in the Linker subfolder, it has been modified to make more RAM available. -Update the hardware definitions macro. Click on Project -> Build Options... -> Project -> MPLAB C18 -> Add PICDEMNET, remove HPC_EXPLORER) -Remove ENC28J60.c from the project -Remove SPIEEPROM.c from the project -Add RTL8019AS.c to the project -Add I2CEEPROM.c to the project -Enable all compiler optimizations (Project -> Build Options... -> Project -> MPLAB C18 -> Categories Optimization -> Enable all)

v3.60 12 July 2006

General Information: This stack version is being publicly released, so the following changes are with respect to the prior public stack release (v3.02). Interim stack changes for version 3.16 and 3.50 are documented below for those using non-public releases, but can be ignored by most people. 

Troubleshooting notes:

1. If you have an Ethernet PICtail revision 2.1 and are having reliability issues when viewing the fast-refresh demo web page, you may need to install resistors in series with the ENC28J60 SI, nCS, and SCK pins. The recommended value is 100 to 200 ohms. This will reduce signal undershoot caused by long traces (parasitic inductance), which can violate the absolute maximum electrical specs and cause SPI data corruption. The HPC Explorer Rev 5 has fairly long traces to the PICtail connector.
2. Enabling C30 2.02 compiler optimizations on the dsPIC33FJ256GP710, PIC24HJ256GP610 ES chips may produce unreliable code.
3. When changing a C30 project to a PIC24H or dsPIC33F processor on the Explorer 16 demo board, the JTAG configuration fuse should be disabled to free the I/O pins associated with it. JTAG is enabled by default.
4. This stack release was tested using MPLAB 7.40, C18 version 3.03, C30 version 2.02, and HI TECH PICC18 version 9.50PL1.
5. When using the Ethernet PICtail board and HPC Explorer demo boards, make sure to plug the power into the Ethernet PICtail and not the HPC Explorer. The HPC Explorer's power regulator cannot provide enough current.

Changes:

1. Source files have been split into separate directories. To compile old applications with this new stack, application source files may need to be updated to include the proper path to the stack header files.
2. New MPLAB projects have been created: -C18EEPROM: Equivalent to the previously named "mpnicee" project. Designed for PIC18's using the C18 compiler. Web page content, board's IP address, MAC address, DHCP enabled state, etc. is stored in an external SPI EEPROM (25LC256 on demo boards). FTP Server demo is included. -C30EEPROM: New supporting PIC24 and dsPIC controllers using the C30 compiler. Similar to C18EEPROM. -C18ProgramMem: Equivalent to the previously named "mpnicpg" project. Web page content stored in internal FLASH program memory. Board's IP address, MAC address, DHCP enabled state, etc. is stored only in RAM and defaults are loaded from MY_DEFAULT_* constants in StackTsk.h. FTP Server demo is not included. Web pages cannot be updated remotely. -C30ProgramMem: New supporting PIC24 and dsPIC controllers using the C30 compiler. Similar to C18ProgramMem. -HTC18EEPROM: Equivalent to the previously named "htnicee" project. Designed for PIC18's using the HI TECH PICC18 compiler. Similar to C18EEPROM. -HTC18ProgramMem: Equivalent to the previously named "htnicpg" project. Designed for PIC18's using the HI TECH PICC18 compiler. Similar to C18ProgramMem.
3. Created hardware definitions (pins, interrupt flags, special registers, etc) in Compiler.h for easy changing of hardware. Four demo board combinations are supported out-of-box now: -EXPLORER_16: Explorer 16 motherboard + Ethernet PICtail Plus daughter card. Tested with dsPIC33FJ256GP710, PIC24HJ256GP610, and PIC24F128GA010 ES PIMs. -HPC_EXPLORER: PICDEM HPC Explorer motherboard + Ethernet PICtail daughter card. Tested with PIC18F8722 onboard and PIC18F87J10 PIM. -DSPICDEM11: dsPICDEM 1.1 motherboard + Ethernet PICtail daughter card (manually air wired). See Compiler.h for proper pins to air wire. Tested with dsPIC30F6014A PIM. -PICDEMNET2: PICDEM.net 2 motherboard (PIC18F97J60) Change boards by changing the defined macro (Project -> Build Options... -> Project -> MPLAB Cxx -> Add macro). When moving to custom hardware, add an appropriate profile to Compiler.h. YOUR_BOARD is present as a placeholder.
4. Added Ethernet PICtail Plus schematic (reference ENC28J60 daughter card design for Explorer 16 demo board). These boards have a Microchip part number of AC164123.
5. Latest ENC28J60 rev. B5 errata workarounds added. The code checks the EREVID register and implements the appropriate workarounds as needed for the silicon revision, so rev. B1, B4, and B5 are all supported in this stack release.
6. Significantly revised demonstration web page content in WebPages folder to use AJAX technology. Using asynchronous JavaScript code executing in the web browser, the status sections of the page are updated rapidly from the web server without doing a full page refresh. As a result, a virtually real time update of the potentiometer and button values can be displayed. Due to the constant use of new TCP sockets, multiple simultaneous users are not recommended. See the Index.cgi file for a simple static method of retrieving dynamic variables from the HTTP server.
7. Changed IP Gleaning procedure. Now, if DHCP is enabled, the DHCP module will continue to look for a new IP address/renew existing IP address if the IP address is configured using IP Gleaning. Previously, the DHCP module would be disabled once a successful ICMP packet was received and used to configure the IP address.
8. MAX_RETRY_COUNTS is 3 (previously it was 3, but an interim release changed it to 5).
9. Updated TCP state machine. It now includes the TCP_FIN_WAIT_2 state. Some other changes were made to handle errors more robustly.

10.AN0String and AN1String now return all characters excluding the null terminator when the HTTP server calls HTTPGetVar (except when the string is 0 length). Previously, the null terminator was returned as well. 11.Dynamic pages (ie: .cgi files) are now served with an expired HTTP header to prevent browser caching and allow more dynamic content to be displayed. 12.Support for the HI TECH PICC18 compiler has changed. Special Function Register bits and other definitions have changed substantially from the previous HI TECH PICC18 projects in TCP/IP stack version 3.02 and earlier. The C18/C30 SFR and SFRbits naming conventions are now used and special remapping macros in Compiler.h are used to maintain a consistent syntax. The HI TECH PICC18 projects were tested with compiler version 9.50PL1 on the HPC Explorer board (PIC18F8722). 13.FTP client hash printing has been added to the FTP server. Now, whenever a chunk of data is successfully uploaded to the device, a '#' character will appear on the FTP client screen. The numbers of bytes each '#' represents is variable. 14.To improve maintainability, built in support for the "Compatible" A/D converter present on older PIC18 parts (ex: PIC18F452) has been removed. 15.Removed old LCD code originally provided for the PICDEM.net demo board. 16.Added LCDBlocking.c and LCDBlocking.h, which implement simple routines for writing to the LCD module on the Explorer 16 and PICDEM.net 2 development boards. The LCD on the dsPICDEM 1.1 board is not supported. The stack version and IP address are shown on the LCD on power up. 17.UART functions in MainDemo.c were replaced with C18 and C30 peripheral library functions. However, because the UART peripheral libraries are not being updated for newer silicon devices, the code was copied into UART.c and is compiled with the stack. 18.Multiple TX buffer support has been implemented. Most stack layers have been touched. ENC28J60.c has the most extensive changes. Each socket may use only one TX buffer. 19.Implemented TCP retransmission support regardless of if TCP_NO_WAIT_FOR_ACK is defined or not. 20.TCP_NO_WAIT_FOR_ACK in StackTsk.h has been undefined by default. This should increase default TCP connection robustness. Packets sent from the stack to the remote node will now be detected and retransmitted if lost or corrupted. 21.All TCP packets are now retransmitted immediately after being initially transmitted when TCP_NO_WAIT_FOR_ACK is undefined. This improves throughput greatly when communicating with systems which wait a long time before transmitting ACKs. TCP/IP stacks, such as that used by Microsoft Windows, implement the TCP Delayed Acknowledgement algorithm, which is why this retransmission is necessary for high performance. The double transmission feature can be disabled in the Microchip TCP/IP stack by defining "DEBUG" either in the TCP.c file or the project compiler macros section. Using DEBUG mode can be useful when trying to look for errors using Ethreal [ http://www.ethereal/ ]. 22.Lowered TCP_START_TIMEOUT_VAL from 60 seconds to 3 seconds. 60 seconds is an unreasonably long timeout for modern day network speeds. 23.Native support for the SLIP module has been dropped. 

Fixes:

1. A new IP address obtained via IP Gleaning will now update the LCD (if present), invoke the Announce module (for MCHPDetect.exe), and output the new address out the RS232 port.
2. DHCP client will now correctly use the first DHCP offer received when connected to a network running multiple DHCP servers. Previously, the board would get no IP address when attached to a network with multiple DHCP servers (unless the DHCP request was transmitted before a second DHCP offer was received -- a relatively rare event). Additionally, DHCPLeaseTime does not get reset to 60 seconds or the value stored in the last DHCP packet received prior to receiving the ACK.
3. UDPProces() will now correctly process received UDP packets that have a 0x0000 checksum field. The UDP protocol specifies that 0x0000 means the checksum is disabled. Packets with a 0x0000 checksum were previously thrown away unless the calculated checksum also happened to be 0x0000.
4. The TCPIsPutReady() function will now honor the remote node's TCP window size. In other words, if the remote application pauses or cannot handle the incoming data rate, the TCP flow control feature will correctly function. Previously, if the remote node ran out of incoming buffer memory, the TCP layer would still allow more data to be transmitted. This would result in the loss or corruption of application data, with a potentially broken connection. The change requires 2 more bytes of RAM per TCP socket (TCB array).

Known Problems:

1. On PICDEM.net 2 board ENC28J60 and 25LC256 EEPROM share the same SPI1 module. At 3.3V, the 25LC256 is only rated to 5MHz SPI clock, but the code is setting it to 10.4MHz because the MACInit() function reconfigures the same SPI1 module.
2. DHCP will continually send out DHCP Request packets when the lease expires and the original DHCP server that gave the lease is offline. The board will continue to use the expired IP address until the DHCP server comes back online, at which point the lease will be renewed or a new discovery will occur. A new discovery should occur after timing out, instead. It is believe that this problem has always existed in previous stack revisions.
3. DHCP will continually send out DHCP Request packets when the lease expires and the original DHCP server that gave the lease does not include Option 54, the Server Identifier. A new discovery should occur after timing out. It is believe that this problem has always existed in previous stack revisions.
4. The MPFS utility has not been updated. When creating a .c image file, the include path for the Compiler.h file will be incorrect and need to be manually updated to "..IncludeCompiler.h".
5. When an MPFS .c image file is added to a C30 project, a linking error reporting insufficient contiguous .const memory may occur when too much data is in the MPFS image (PSV window size limitation). Using the PSV window, 1 out of every 3 program memory bytes is wasted.
6. MACSetPMFilter(), MACDisablePMFilter(), and MACCopyRxToTx() have not been tested and possibly do not work.
7. SNMP, TFTPc modules have not been tested with this version.
8. Serial numbers >32K will be displayed on the serial port as a negative value when C18 is used and the board is placed in configuration mode (RB0 button is depressed on power up).
9. IP Gleaning may not get disabled when, through the RS232 configuration application, DHCP and IP Gleaning are disabled.

10.The C30 linker may misplace the __CONFIG2 section or disallow usage of MPFS images that are too big (add too much to the .const code section). The consequences of this are that the first configuration word at 0x157FC may not get set through code (must use the Configuration Bits dialog instead), and/or the project will not compile. This problem has been observed with C30 ver. 2.02 on the PIC24FJ128GA010 product. To work around this problem, the p24FJ128GA010.gld linker script has been modified. Specifically, line 68 has been commented out, which causes the linker to place all .text sections after placing all absolute sections. SSR 25966 in the C30 2.02 release notes may be related. 

Guiding Notes:

1. To change processors using a C18* project: -Change the processor in the MPLAB IDE -Change linker script (ex: 18f87j10i.lkr) in the MPLAB project. Use *i.lkr if the ICD2 is going to be used to debug with. -Update the hardware definitions in Compiler.h or change your demo board selection macro. (Project -> Build Options... -> Project -> MPLAB Cxx -> PICDEMNET2, etc)
2. To change processors using a HTC18* project: -Change the processor in the MPLAB IDE -Update the hardware definitions in Compiler.h or change your demo board selection macro. (Project -> Build Options... -> Project -> MPLAB Cxx -> PICDEMNET2, etc)
3. To change processors using a C30* project: -Change the processor in the MPLAB IDE -Change linker script (ex: p33FJ256GP710.gld) in the MPLAB project. -Update the hardware definitions in Compiler.h or change your demo board selection macro. (Project -> Build Options... -> Project -> MPLAB Cxx -> DSPICDEM11, etc) -Disable JTAG configuration fuse, if enabled
4. When using the PICDEM.net 2 board, to write code targeting the PIC18F97J60 family Ethernet module: -Remove ENC28J60.c from the project -Add ETH97J60.c to the project -Plug the Ethernet cable into the left-most RJ45 jack (next to LCD)
5. When using the PICDEM.net 2 board, to write code targeting the ENC28J60 Ethernet device: -Make sure ENC28J60.c is in the project -Make sure that ETH97J60.c is not in the project -Plug the Ethernet cable into the right-most RJ45 jack (next to board edge)
6. When using the PICDEM.net 2 board, to write code targeting an Ethernet PICtail module (ENC28J60): -Make sure ENC28J60.c is in the project -Make sure that ETH97J60.c is not in the project -Make sure that the Ethernet PICtail J9 jumper is in the 2-3 position (default). -Properly update the hardware profile in Compiler.h. ENC_CS_TRIS and ENC_CS_IO need to be changed from D3 to B3. -Plug the Ethernet cable into the PICtail -Plug power into the PICDEM.net 2 board
7. When using the Explorer 16 and Ethernet PICtail Plus demo boards, make sure to mate the PICtail to the motherboard using the topmost socket position, leaving the cable hanging over prototyping area. If SPI2 is desired, the PICtail should have the same orientation but be installed in the middle slot. Using SPI2, the hardware profile will need to be updated in Compiler.h.

v3.50 13 April 2006

Changes:

1. Improved dsPIC33F and PIC24H support. UART functions are included now instead of precompiled object files for the PIC24F. The 12-bit A/D converter is now shown in use on the demo web content. When changing a C30 project to a PIC24H or dsPIC33F processor on the Explorer 16 demo board, the JTAG configuration fuse should be disabled to free the I/O pins associated with it. JTAG is enabled by default.
2. Added LCDBlocking.c and LCDBlocking.h, which implement simple routines for writing to the LCD module on the Explorer 16 development board. The stack version and IP address are shown on the LCD on power up.
3. Added "C18ProgramMem" and "C30ProgramMem" MPLAB projects for MPFS storage (web page content) on on-chip program memory. These projects are equivalent to the previously named "mpnicpg" project in prior stack releases.
4. Multiple TX buffer support has been implemented. Most stack layers have been touched. ENC28J60.c has the most extensive changes. Each socket may use only one TX buffer.
5. Implemented TCP retransmission support when TCP_NO_WAIT_FOR_ACK is undefined.
6. TCP_NO_WAIT_FOR_ACK in StackTsk.h has been undefined by default. This should increase default TCP connection robustness.
7. All TCP packets are now retransmitted immediately after being initially transmitted when TCP_NO_WAIT_FOR_ACK is undefined. This improves throughput greatly when communicating with systems which wait a long time before transmitting ACKs.
8. Lowered TCP_START_TIMEOUT_VAL from 60 seconds to 3 seconds.
9. Increased MAX_RETRY_COUNTS from 3 to 5 times.
10. The example HTTP server now returns a content expiration date which has already past. This prevents web browser caching and allows more dynamic content to be displayed.
11. Added WebPages_JScript folder, with new web pages that support dynamic page updates without a full page reload. A tiny page of dynamic variables is returned by the web server and Javascript executing on the target web browser changes DOM elements as needed. Button S5 (RA7) on the Explorer 16 demo board and S1 (RB0) on the HPC Explorer demo board changes the page color scheme. The rapid dynamic updates do not work on some web browsers (Internet Explorer works, Firefox does not).

Known Problems:

1. MPFS utility has not been updated. When creating a .c image file, the include path for the compiler.h file will be incorrect and need to be manually updated.
2. When an MPFS .c image file is added to a C30 project, a linking error reporting insufficient contiguous .const memory may occur (PSV window size limitation).
3. MACSetPMFilter(), MACDisablePMFilter(), and MACCopyRxToTx() have not been tested and possibly do not work.
4. SNMP, TFTPc, SLIP modules have not been tested with this version.
5. Serial numbers >32K will be displayed on the serial port as a negative value when C18 is used and the board is placed in configuration mode (RB0 button is depressed on power up).
6. IP Gleaning may not get disabled when, through the RS232 configuration application, DHCP and IP Gleaning are disabled.
7. The IP address being outputted out the RS232 port and through the Announce module does not happen when the IP address is configured using IP Gleaning.
8. On the PIC24F with C30 compiler optimizations enabled (such as Option 3, maximum speed), the project may not work. The PIC24F headers that come with C30 ver. 2.01 declare several SFRs without using the volatile keyword.
9. dsPIC30 support is incomplete. Currently PIC18, PIC24F, PIC24H, and dsPIC33F processors are supported.

v3.16.00: 06 March 2006

Changes:

1. Added unified support for both the Microchip C18 and C30 compilers. The intention is to allow one code base to be compiled for any PIC18, PIC24F/H, dsPIC30, or dsPIC33 product (with adequate memory). See the "Tested Using" section for what is known to work.
2. To improve maintainability, support for the HI-TECH PICC18 compiler has been dropped.
3. New project workspaces have been created, "C30EEPROM.mcw" and "C18EEPROM.mcw". C18EEPROM.mcw is equivalent to the previously named "mpnicee.mcw." C30EEPROM is intended to be used for PIC24 and dsPIC 16-bit controllers.
4. Source files have been split into separate directories.
5. Latest ENC28J60 rev. B5 errata workarounds added. The code checks the EREVID register and implements the appropriate workarounds as needed for the silicon revision, so rev. B1, B4, and B5 are all supported in this stack release.
6. Removed old LCD code originally provided for the PICDEM.net demo board.
7. To improve maintainability, built in support for the "Compatable" A/D converter present on older PIC18 parts (ex: PIC18F452) has been removed.
8. UART functions in MainDemo.c were replaced with C18 and C30 peripheral library functions.

Tested Using:

1. Software: -MPLAB version 7.31.01 -C18 version 3.02 -C30 version 2.01
2. Hardware: -PICDEM HPC Explorer rev. 4 (PIC18F8722) + Ethernet PICtail Daughter Board (ENC28J60

B1) -Explorer 16 rev. 4 (PIC24FJ128GA010 ES and dsPIC33FJ256GP710 ES) + Ethernet PICtail+ Daughter card (ENC28J60 B1).

3. Notes: -MPLAB 7.31.01 is a development build. The publicly available version 7.31 should work fine, with the exception of being unable to program dsPIC33 and PIC24H parts with the ICD 2. -No dsPIC30 or PIC24H parts have been tested yet.

Known Problems:

1. MPFS utility has not been updated. When creating a .c image file, the include path for the compiler.h file will be incorrect and need to be manually updated.
2. When an MPFS .c image file is added to a C30 project, a linking error reporting insufficient contiguous .const memory may occur.
3. On the PIC24FJ128GA010, it is observed that some inbound packets are lost from time to time with no anticipated reason.
4. MACSetPMFilter(), MACDisablePMFilter(), and MACCopyRxToTx() have not been tested and possibly do not work.
5. SNMP, TFTPc, SLIP modules have not been tested with this version.
6. Serial numbers >32K will be displayed on the serial port as a negative value when C18 is used and the board is placed in configuration mode (RB0 button is depressed on power up).
7. IP Gleaning may not get disabled when, through the RS232 configuration application, DHCP and IP Gleaning are disabled.
8. The IP address being outputted out the RS232 port and through the Announce module does not happen when the IP address is configured using IP Gleaning.
9. Multiple TX buffer support is not fully inplemented in the MAC layer, ENC28J60.c. Stack behavior when TCP_NO_WAIT_FOR_ACK is undefined may be unexpected.

v3.02.00: 20 Feb 2006

Fixes:

1. Changed TXSTART in ENC28J60.c to stop wasting a byte.
2. Changed RXSTOP in ENC28J60.c to always be an odd value to properly implement an ENC28J60 silicon errata workaround.
3. Changed initialization of ERXRDPT in MACInit() to agree with the current errata.

Changes:

1. Licence agreement
2. Schematics and other board files to the Ethernet PICtail Daughter Board have been updated to revision 5. Of significant note, the nRESET pin has been freed and 200 ohm resistors were added to the ENC28J60 SI, nCS, and SCK pins. The added resistors reduce undershoot caused by stray trace inductance and strong host output drivers.

Known Problems:

1. Testing on the PICDEM.net demo board with the Realtek RTL8019AS Ethernet controller has not been done. Moving to the HPC Explorer demo board has resulted in pinout and other hardware changes.
2. MACSetPMFilter(), MACDisablePMFilter(), and MACCopyRxToTx() have not been tested and possibly do not work.
3. SNMP, TFTPc, LCD, SLIP modules have not been tested with this version.
4. The stack may behave incorrectly if compiled using the Hitech compiler with a high optimizations setting.
5. Serial numbers >32K will be displayed on the serial port as a negative value when C18 is used and the board is placed in configuration mode (RB0 button is depressed on power up).
6. IP Gleaning may not get disabled when, through the RS232 configuration application, DHCP and IP Gleaning are disabled.
7. The IP address being outputted out the RS232 port and through the Announce module does not happen when the IP address is configured using IP Gleaning.
8. Multiple TX buffer support is not fully inplemented in the MAC layer, ENC28J60.c. Stack behavior when TCP_NO_WAIT_FOR_ACK is undefined may be unexpected.

v3.01.00: 18 Jan 2006

Fixes:

1. Implemented latest ENC28J60 silicon errata workarounds.
2. Fixed a bug in TCP.c and UDP.c which would incorrectly write the packet checksum into the RX buffer incorrectly when the checksum field was exactly spanning the RX wrapparound boundary in the ENC28J60. This problem would have caused packets to be discarded in rare circumstances

Known Problems:

1. Testing on the PICDEM.net demo board with the Realtek RTL8019AS Ethernet controller has not been done. Moving to the HPC Explorer demo board has resulted in pinout and other hardware changes.
2. MACSetPMFilter(), MACDisablePMFilter(), and MACCopyRxToTx() have not been tested and possibly do not work.
3. SNMP, TFTPc, LCD, SLIP modules have not been tested with this version.
4. The stack may behave incorrectly if compiled using the Hitech compiler with a high optimizations setting.
5. Serial numbers >32K will be displayed on the serial port as a negative value when C18 is used and the board is placed in configuration mode (RB0 button is depressed on power up).
6. IP Gleaning may not get disabled when, through the RS232 configuration application, DHCP and IP Gleaning are disabled.
7. The IP address being outputted out the RS232 port and through the Announce module does not happen when the IP address is configured using IP Gleaning.
8. Multiple TX buffer support is not fully inplemented in the MAC layer, ENC28J60.c. Stack behavior when TCP_NO_WAIT_FOR_ACK is defined may be unexpected.

v3.00.00: 16 Jan 2006

Changes:

1. The stack now targets the PICDEM HPC Explorer demo board (PIC18F8722 MCU) with an attached Ethernet PICtail Daughter Board (with the Microchip ENC28J60 Ethernet controller).
2. IP Gleaning is no longer enabled (STACK_USE_IP_GLEANING is not defined) by any of the default project files.
3. The IP address, whenever it changes, is outputted out the RS232 serial port in human readable form. Any terminal program, such as HyperTerminal can be used to read it. This allows the IP address to be easily determined when DHCP is used. The serial port defaults to 19200 baud when CLOCK_FREQ in Compiler.h is properly defined.

Additions:

1. Microchip ENC28J60 Ethernet controller support. Support is included through the ENC28J60.c and ENC28J60.h files. Various other files were modified to take advantage of ENC28J60 specific features, like the hardware DMA/IP checksum engine. This new MAC driver incorporates several new functions which can be called from any layer above the MAC. The functions are: MACSetDuplex() MACPowerDown() MACPowerUp() MACSetPMFilter() MACDisablePMFilter() CalcIPBufferChecksum() MACCalcRxChecksum() MACCalcTxChecksum() MACCopyRxToTx() See the ENC28J60.c file comments for function descriptions. The ENC28J60.c file also incroporates TestMemory() which can do a power on self test of various hardware functions. TestMemory() is included and used when MAC_POWER_ON_TEST is defined in StackTsk.h. It is undefined by default. Defining it will require some program memory.
2. Announce module. Announce.c and announce.h have been added. When included in the project, STACK_USE_ANNOUNCE must be defined. This module will broadcast a UDP message to port 30303 containing the local MAC address whenever the local IP address changes. This addition is intended to facilitate device discovery on DHCP enabled networks and eliminate the need for an RS232 connection if board reconfiguration is not needed. To retrieve the UDP message on your computer, use the new MCHPDetect.exe program included in the MCHPDetect subfolder.
3. The spieeprom.c file was added to support SPI EEPROM chips for MPFS storage. ENC28J60.c and spieeprom.c may both be included and they will share the same SPI module.

Improvements:

1. Renamed files/edited files so that the HI-TECH compiler won't raise messages stating that include files were spelled wrong.
2. Moved MAX_ICMP_DATA_LEN from StackTsk.c to ICMP.h file for easier maintenance.
3. Corrected STACK_USE_SIIP typo in dhcp.c file - Thanks to Gisle J.B.
4. Implemented UDP checksum logic in UDPProcess() in UDP.c file.
5. Renamed CalcTCPChecksum() in tcp.c file to CalcIPBufferChecksum().
6. Moved CalcIPBufferChecksum() to helpers.c to reuse it for UDP checksum calculation.
7. Modified UDPProcess() in UDP.c and TCPProcess() in TCP.c to include localIP as third new parameter. This makes pseudo header checksum calculation correct in both functions. StackTsk.h, UDP.h and TCP.h files were also modified to reflect these changes.
8. Modified TCP.C file to include compile-time check of STACK_USE_TCP define. If it is not defined, an error will be displayed.
9. Removed an unnecessary call to MACDiscardRx() when an IP packet is received but fails version, options length, or header checksum tests.
10. Changed LCD code to be compile time removable by undefining USE_LCD.

Fixes:

1. IPHeaderLen in IP.c is initialized properly now when IPGetHeader() is called.
2. Under some circumstances, HandleTCPSeg() would acknowlege, but throw valid received TCP packets away, resulting in loss of application data. An invalid comparison in HandleTCPSeg() has been fixed to prevent this situation from occuring. *** Thanks go to Richard Shelquist for identifying this problem.
3. Fixed StackTsk.c file so that if a static IP address is used and the LINK is removed, the node IP address is not cleared.
4. Invalid ICMP echo replies are no longer generated for echo requests with a data length of 33 (one more than the configured maximum).
5. Changed MAX_OPTIONS_LEN from 20 to 40. The maximum IP options length is now in agreement with the IP RFC.
6. Changed IPSetRxBuffer() from a macro to a function. The function takes into account any options which may be present in the header of received IP packets. Previously, possible options were not taken into account when calculating the offset.

Known Problems:

1. Testing on the PICDEM.net demo board with the Realtek RTL8019AS Ethernet controller has not been done. Moving to the HPC Explorer demo board has resulted in pinout and other hardware changes.
2. Sometimes when the FTP sever is used, an attempt to put a file is unsuccessful. The problem may be caused when an HTTP request to GET a file is made at the wrong time.
3. MACSetPMFilter(), MACDisablePMFilter(), and MACCopyRxToTx() have not been tested and possibly do not work.
4. SNMP, TFTPc, LCD, SLIP modules have not been tested with this version.
5. The stack may behave incorrectly if compiled using the Hitech compiler with a high optimizations setting.
6. Serial numbers >32K will be displayed on the serial port as a negative value when C18 is used and the board is placed in configuration mode (RB0 button is depressed on power up).
7. IP Gleaning may not get disabled when, through the RS232 configuration application, DHCP and IP Gleaning are disabled.
8. The IP address being outputted out the RS232 port and through the Announce module does not happen when the IP address is configured using IP Gleaning.
9. Multiple TX buffer support is not fully inplemented in the MAC layer, ENC28J60.c. Stack behavior when TCP_NO_WAIT_FOR_ACK is defined may be unexpected.

v2.20.04.01: 9/24/03

1. Recreated MPLAB projects to avoid problems when source is not at MCHPStack location.

v2.20.04: 9/5/03

Fixes:

1. Modified DHCPReset() in DHCP.c to not reset DHCP state machine if it was previously disabled using DHCPDisable(). This would make sure that if DHCP module was enabled and application had run-time disabled DHCP and network cable is disconnected, stack will not clear its IP address.

2. Rebuilt mib2bib.exe file with static library options. This fixes problem where one tries to execute this exe, an error occurs about missing DLLs.

v2.20.03:

Improvements:

1. When DHCP is enabled, LINK is monitored and IP address is reset on disconnect. New IP configuration is obtained on LINK reconnect. - For RealTek only. Modified DHCP.c to add DHCPReset() Modified MAC.c to add MACIsLinked() Modified StackTsk.h to add BYTE_VAL def.

Changes:

1. Modified SMSC91c111.c to add empty MACIsLinked() - will be populated in next rev.

Bug Fixes:

1. Corrected DHCP logic to accept first DHCP offer instead of second response.
2. Corrected DHCP logic to check for chaddr in DHCP offer and accept one that matches with local MAC address. This will fix problem where if multiple nodes were on bus and all requested DHCP address, all would accept response from one server instead of verifying who was intended node.
3. Fixed UDPClose() in UDP.c to use INVALID_UDP_PORT instead of INVALID_UDP_SOCKET because of which a closed socket would not be scanned correctly.
4. Modified UDP.h to use long contsant designators for INVALID_UDP_OPRT to explicitly state that it is a long.

v2.20.02:

Beta version containing TFTP client module. 

Addition:

1. TFTP Client module - See TFTPc.* and TFTPcDemo.c for more information. See MpTFTPcDemo and HtTFTPcDemo projects for build information.

Bug Fix:

1. UDPIsGetReady() was modified to overcome compiler rule where only 8-bit value was used to evaluate non-zero condition.
2. ARPResolve() in ARPTsk was fixed to clear Cache.IPAddr value.

v2.20.01:

Bug fix:

1. Fixed SMSC91C111.c where MACInit() would hand if ethernet link is not detected.

v2.20:

Bug Fixes:

1. General - Removed most of harmless warnings.
2. C18Cfg.asm - Fixed "include" instead of "define".
3. DHCP.c - Increased DHCP_TIMEOUT_VAL to 2 seconds. Fixed problem where UDP active socket was not set before calling UDP

functions in SM_DHCP_BROADCAST state.

4. MAC.c - Fixed MACIsTxReady() where under heavy traffic it would always return

FALSE. This fixes bug where all high level applications would stop transmitting.

5. TCP.c - Enabled portion of code that performs timeout logic even if

TCP_NO_WAIT_ACK is defined. This fixes bug where occasionally, tcp applications such as HTTP server would stop working after few hours.

6. UDP.c - Fixed UDPGet() where it would return FALSE on last good byte. Fixed UDPProcess() where it was calculating incorrect length.

Added bFirstRead flag with UDP sockets similar to TCP sockets so that whenever first UDP byte is read, MAC read pointer will be reset to begining of correct packet. This change fixes problem where if one transmits a packet while UDP packet is pending in a socket, next get to pending UDP socket would return wrong data. (This is apparent only when there is heavy network traffic) 

Known Issues:

1. HiTech v8.20 PL4 with all optimization enabled may not work properly.
2. C18 "Static" and "Auto" mode may not be used - there are too many local variables to

fit in standard stack of 256 bytes. One may modify linker script file to avoid this limitation. 

Improvements:

1. Modified TICK def. in Tick.h to unsigned long to support 32-bit wide SNMP tick.
2. Added SNMP Module (SNMP.c)
3. Added Two new demo projects - DemoSNMPApp and HtDemoSNMPApp.
4. Created MPLAB 6.X projects for different demo configurations.
5. MAC.c - Added MACGetTxOffset().
6. MPFS.c - Added MPFSSeek(), MPFSTell().
7. MPFSImg.*- Rebuilt to reflect v2.20, footprint changes etc.
8. StackTsk.h- Enhanced WORD_VAL, DWORD_VAL defs. Added STACK_USE_SNMP and related compile-time checks.
9. UDP.h - Added UDPSetTx and UDPSetRx macros. Moved UDP_SOCKET_INFO structure to header file.
10. WebSrvr.c- Modifed MCHPStack version message and added DATE info to BoardSetup menu.
11. Added support for SMSC LAN91C111 10/100 Non-PCI ethernet controller Use "SMSC91C111.C" instead of MAC.c. "mpnicee_smsc" is a sample project that uses PIC18F8720 and SMSC NIC. "MasterDemo.c" is a main source file for above project that includes all modules - must use device with more than 32KB of memory.

v2.11:

Bug Fixes:

1. Fixed dhcp.c to make it work with new C18 startup code.

Improvements:

1. Modified websrvr.c DownloadMPFS() to make use of compiler allocated XMODEM data block rather than use fixed address block starting at 0x400.

v2.10: 7/9/02

Bug Fixes:

1. Fixed HTTP Server bug where a form submission with empty parameter value would not parse correctly.

v2.0: 5/22/02

New Modules:

1. Added UDP, DHCP, FTP and IP Gleaning
2. Added PICDEM.net LCD support
3. Added board setup through RS-232.

Improvements:

1. Optimized serial EEPROM access routines in terms of speed and size (Replaced ee256.* files with eeprom*.h)
2. Improved board setup through RS-232.

Known Issues:

1. LCD may not display properly on MCLR only. Workaround: 1. Debug XLCDInit() routine in "xlcdlh"
   2. Always do POR reset.

2. SLIP connection is not very robust. Workaround: None at this time.

3. Hi-Tech Compiler:
   1. Aggressive optimization breaks the functionality. Workaround: Apply optimization listed in each source file comment

header.

2. In order to use V8.12, you will need to remove "FTP Server" from

Ht*.pjt. You will also need to disable all optimizations. 

4. C18 Compler: 1. Static model does not compile. Workaround: None at this time.
   2. Overlay model breaks the functionality. Workaround: None at this time.
   3. All modules does not fit in 32KB memory. Workaround: 1. None at this time.
      2. Sample project disables some modules.

New Files:

======================================================================================== ==================================== File Purpose ======================================================================================== ====================================

1. delay.* Provides CLOCK_FREQ depenent delay routines.
2. dhcp.* DHCP client support
3. ftp.* FTP server
4. udp.* UDP socket support
5. xeeprom.* Improved ee256.* and renamed.
6. xlcd.* External LCD support.
7. version.log To track changes and history.

Changes:

======================================================================================== ==================================== File Change To-do for v1.0 stack applications ======================================================================================== ====================================

1. arptsk.c 1. Fixed STACK_CLIENT_MODE compile errors.

None

2. Modifed ARPIsResolved() to support IP Gleaning

None 

2. c18cfg.asm 1. Added PIC18F452 configuration

None

2. Fixed "include" errors.

None 

3. compiler.h 1. Included "stdlib.h" in both C18 and Hi-Tech compilers.

None

2. Moved CLOCK_FREQ from "stacktsk.h" to this file.

None

3. Added PORTA defs.

None 

4. htnicee.pjt 1. Removed "ee256.c".

None

2. Added "udp.c", "dhcp.c", "ftp.c", "xlcd.c", "xeeprom.c" files

Add these files if needed. 

5. htnicpg.pjt None

6. htslee.pjt 1. Removed "ee256.c".

None

2. Added "ftp.c", "xlcd.c", "xeeprom.c" files

None 

7. http.c 1. Included compile-time verification that HTTP module is included.

None

2. Put HTTP message strings into one array "HTTPMessages".

None

3. Modified to return "Service Unavailable" message if MPFS is being

None remotely programmed.

4. Modified SendFile() to make use of sequential EEPROM read.

None 

8. ip.c 1. Added one more paramter to IPGetHeader() to support IP Gleaning

Custom apps using IP needs to be 

modified. 

9. mac.c 1. Replaced fixed delay routines with CLOCK_FREQ dependent

None routines 

10. mpfs.c 1. Replaced ee256.h with xeeprom.h.

None

2. Added MPFSFormat(), MPFSPut() etc. routines

None

3. Added sequential read and page write operations

Custom apps using MPFS directly 

needs to be modified.

4. Defined MPFS_WRITE_PAGE_SIZE for MPFSPut operations.

Apps using different EEPROM page size 

needs to be modified. 

11. mpnicee.pjt 1. Removed "ee256.c"

None

2. Added "xcld.c", "xeeprom.c" files

None 

12. stacktsk.c 1. Replaced ee256.h with xeeprom.h

None

2. Added IP Gleaning and DHCP support.

None 

13. stacktsk.h 1. Moved CLOCK_FREQ to compiler.h

None

2. Added STACK_USE_DHCP, STACK_USE_FTP_SERVER etc. options

None

3. Added compile-time enable/disable of modules based on selection of higher level modules.

None

4. Modified MY_DEFAULT_MAC_BYTE? to use Microchip OUI id.

None

5. Added compiler-time check to confirm available TCP sockets

None

6. Added MSB and LSB macros.

None

7. Added SerialNumber etc. to AppConfig structure

None

8. Commented module selection defines: They are defined by cmopiler

None command-line options. Real application should define them here in this file. 

14. tcp.c 1. Moved TCP_STATE and TCP_INFO to .h file.

None

2. Fixed TCPIsConnected()

None

3. Fixed TCPDisconnect()

None

4. Modified TransmitTCP() to set receive window of one segment

None

5. Modified TransmitTCP() to use max segment size equal to predefined value.

None

6. Improved TCP State machine

None 

15. tick.c 1. Modified TICK type to 16-bit.

None

2. Made use of TICK_PRESCALE_VALUE

None

3. Added code to blink PICDEM.net "System LED"

Remove if not required. 

16. websrvr.c 1. Added LCD support

N/A

2. Made TickUpdate() on Timer0 interrupt

N/A

3. Added code to save/restore board configuration

N/A

4. Added board setup via RS-232.

N/A

5. Added call to FTP modules

If needed, add this.