Bug 11434
Summary: | r8169 exits with error -22 | ||
---|---|---|---|
Product: | Drivers | Reporter: | Georg Grabler (ggrabler) |
Component: | Network | Assignee: | Francois Romieu (romieu) |
Status: | CLOSED CODE_FIX | ||
Severity: | normal | CC: | corsac, michelro, rafael.ristovski |
Priority: | P1 | ||
Hardware: | All | ||
OS: | Linux | ||
Kernel Version: | 2.6.26 | Subsystem: | |
Regression: | No | Bisected commit-id: | |
Attachments: | kernel.log - 2008-08-29 |
Description
Georg Grabler
2008-08-27 12:07:01 UTC
Can you: 1. set the r8169.debug option at 65535 before inserting the module 2. send the complete dmesg with a recent kernel (2.6.26 or above) after 1. It should include a line with the 'XID' pattern. 3. try 2.6.27-rc4. It includes some fixes. -- Ueimor I'll answer this questions tomorrow. For some reason, 1 hour ago it stopped dropping errors (without updating - basically from one insert of the module to the next one). The only answer which I can give for now is: eth0: RTL8168c/8111c at 0xf8878000, 00:21:70:a2:9d:1b, XID 3c4000c0 IRQ 16 I've attached my todays kernel.log, showing the errors (first) and later that it's working. Created attachment 17514 [details]
kernel.log - 2008-08-29
Seems to be fixed in 2.6.27-rc4. It could be consistent with the wrecking of the PCI configuration space which has been fixed in commit 77332894c21165404496c56763d7df6c15c4bb09 between 2.6.26 and 2.6.27-rc1 (more or less random failures). I'll close the bug for now. If it reappears, a bisection could be very helpful. -- Ueimor I'll watch out for the bug reappearing. Currently running the rc4 is "the solution", since I did not encounter any problems since then. It seemed like a PCI bug, reason: -> The module loaded first, but had IRQ problems or similar (didn't respond) -> The module unloaded and loaded again did just crash with error -22 Thanks for the help anyway. I'll report back if it appears again with 2.6.27. Thanks, Georg Hmmh, it seems that I get the same problem. I'm using debian Lenny, which has kernel 2.6.26 (at the moment at version 2.6.26.4). It seems that the fix is in 2.6.26.3, but I still get it. I've opened a bug on Debian BTS where you can find the dmesg (I can provide the complete dmesg) http://bugs.debian.org/cgi-bin/bugreport.cgi?bug=499763 Cheers, It is fixed by patches #0001...#0008 of the series at: http://userweb.kernel.org/~romieu/r8169/2.6.27-rc6/20080913/ -- Ueimor Regarding the Debian bugreport, even if the -22 symptom and the random PCI config writes are fixed in 2.6.22.3, I would not claim the 8101 support to be completely reliable without the aforementioned patches. -- Ueimor Is there a chance that those patches could be backported to 2.6.26? For me personally, the 2.6.27rc1 works pretty well. I had a lot of troubles with 2.6.26.4 as well, the module was not reliable in any way. Running now for some time on 2.6.27rc1 (soon one month). Maybe try to compile the vanilla 2.6.27rc1, using the configuration of debian. It's on a production box. I would prefer using a stock debian kernel for that, so using a custom made kernel is not a good idea, and using a rc one is not really an option :/ corsac@debian.org 2008-09-24 07:16 : > Is there a chance that those patches could be backported to 2.6.26? Patches #0001 .. #0008 can be applied "as is" to 2.6.26.5 but I doubt that people can be convinced that they are kosher for 2.6.26 (once they are included in 2.6.28-rc of course). Hello, with kernel 2.6.38, i have this message: r8169 0000:02:00.0: eth0: unable to apply firmware patch r8169 0000:02:00.0: eth0: link down r8169 0000:02:00.0: eth0: link down r8169 0000:02:00.0: eth0: link up Could someone please remove the alias from this bug report? It messes with the search function. (In reply to Ken Sharp from comment #15) > Could someone please remove the alias from this bug report? It messes with > the search function. +1 for this. The alias makes it impossible to search for `r8169` as it automatically redirects to this bug report. |