| 1 |
"Juan Fco. Giordana" <juangiordana@×××××.com> posted |
| 2 |
49C8B0B3.1000709@×××××.com, excerpted below, on Tue, 24 Mar 2009 07:06:43 |
| 3 |
-0300: |
| 4 |
|
| 5 |
> I'm using the driver e1000e for these NICs since it's the one suggested |
| 6 |
> for these Intel cards and I've found lot of information about a problem |
| 7 |
> on this driver that leads to the the firmware being corrupted. I don't |
| 8 |
> think I'm affected though since the first card is working properly and |
| 9 |
> I've dumped the eprom for both adapters and they doesn't show any |
| 10 |
> difference. |
| 11 |
|
| 12 |
On the corrupted firmware bit, the problem ultimately was a combination |
| 13 |
of the dunder-headed hardware/firmware implementation (blindly accepting |
| 14 |
any random write as new firmware without even verifying a checksum!), now |
| 15 |
corrected in newer firmware I believe, and a bug in ftrace. If you never |
| 16 |
had the ftrace kernel option turned on in kernels you ran on that |
| 17 |
hardware, as is presumably the case with production server hardware, or |
| 18 |
if you had it turned on but only in kernels AFTER the fix, the single |
| 19 |
known triggering bug vector will have never been available, so it's very |
| 20 |
unlikely you have it. |
| 21 |
|
| 22 |
LWN article: The source of the e1000e corruption bug: |
| 23 |
http://lwn.net/Articles/304105/ |
| 24 |
|
| 25 |
Other than that, I don't have any direct experience with that hardware, |
| 26 |
so I'm not going to be of much help, I'm afraid. |
| 27 |
|
| 28 |
-- |
| 29 |
Duncan - List replies preferred. No HTML msgs. |
| 30 |
"Every nonfree program has a lord, a master -- |
| 31 |
and if you use the program, he is your master." Richard Stallman |
Archives