Revision: 2.85

September 2012

Specification Update

29

30. PCIe Hot Reset Can Lead to a Firmware Hang

Problem:

A PCIe hot reset prevents the firmware from accessing internal registers. When the hot

reset occurs while the firmware is performing initializations, the firmware might fail to

clear a semaphore bit. As a result, the firmware goes into an infinite polling loop when

the next initialization sequence occurs.

Implication: PHY initialization from the EEPROM is performed by firmware, so if the firmware hangs,

the initialization is not performed and the Ethernet link might not operate correctly.

Additionally, the software driver might fail to load as a result of a hung semaphore bit.

When using manageability, the firmware hang causes a watchdog timer to expire and

then resets the firmware, requiring re-initialization of the manageability.

Workaround: Use new EEPROM images that include a firmware workaround for this issue: 1.9 or later.

Also, do not keep the device in hot reset for more than 20 ms. Hot reset should be

performed by setting and then immediately clearing the Secondary Bus Reset bit in the

upstream PCIe device.

Status:

A1, A2 NoFix

▼ Return to Summary

31. MACSec: Replay Protect Not Working In Check Mode

Problem:

When working in check mode, replay packets are not dropped and they are counted in

MACSec RX Octets Validated LSECRXOCTP (0x4320).

Implication: In check mode, replay protect is not working and LSECRXOCTP counts incorrectly.

Note: In strict mode, which is the normal operation mode, there is no problem.

Workaround: Drivers can provide a workaround by working in “no-header strip” mode and drop the

replay packet in the driver while correcting the LSECRXOCTP and the LSECRXLATE

statistics (maintained in the driver as 64-bit counters).

Status:

A1, A2 NoFix

▼ Return to Summary

32. MACSec: Packets With E=0, C=1 Should Not Be Handled as a Authenticated MACSec Packet

Problem:

MACSec Packets with E=0, C=1 is a reserved future mode, and should not be handled as

a Authenticated MACSec packet.

Implication: If MACSec Packets with E=0, C=1 are sent it will be handled as a Authenticated MACSec

packet. This a reserved mode and there is no implication as long as the specification does

not change in the future.

Workaround: None.

Status:

A1, A2 NoFix

▼ Return to Summary

33. MACSec: Packets With PN = 0 In The SECTAG Are Not Dropped

Problem:

According to the MACSec specification, frames with PN=0 in the sectag should be counted

as bad tags/packets. The 82576 will accept such packets. Setting the first expected PN to