I have a Dell Latitude D600 laptop with an on-board Broadcom BCM5702X Ethernet controller. Under all test kernels later than 2.6.11-1.1177_FC4, attempting to ifup the interface results in no traffic, and this error message is logged whenver I interrupt the ifup command: Apr 14 00:46:58 farslayer kernel: tg3: tg3_stop_block timed out, ofs=4800 enable_bit=2 Kernel 2.6.11-1.1177_FC4 works properly. Under FC3, all 2.6.10 kernels worked properly. (I reinstalled the FC4 test release before the last 2.6.11 kernel update, so I have not tested it.)
Created attachment 113355 [details] lspci -v -v -v output (using kernel 2.6.11-1.1177_FC4)
Please try the test kernels here: http://people.redhat.com/linville/kernels/rhel4/ I think it should work (well enough) on an FC4 runtime. The tg3 driver there is very similar to what is current in FC4. I'd be curious to know if it works any differently. Please give it a try and let me know the results. Thanks!
I also have a Dell Latitude D600 and had been experiencing the same issue as reported by James Ralston. The tg3 driver seems to work fine with the "jwltest" kernels while not working with the 1268_FC4 and 1240_FC4 kernels that I have tried. It should be noted that the newly released 1275_FC4 causes a system hang now when X comes up...
Created attachment 113800 [details] 1275_FC4 kernel panic Kernel panic while loading kernel modules as part of booting process on a Dell Latitude D600 running the 1275_FC4 kernel.
Created attachment 113803 [details] dmesg from a successful boot using the jwltest19 kernel
Created attachment 113804 [details] lspci and lspci -v -v -v from jwltest19
I haven't tried the kernels mentioned in comment #2 yet, but I did find another development kernel that works. Specifically: 2.6.11-1.1253_FC4 works. However, no kernel I've pulled since then (including 2.6.11-1.1275_FC4) works. I'll try the kernels mentioned in comment #2 next.
Ok, the tg3 driver in 2.6.9-6.39.EL.jwltest.19 works.
So, it would appear that something is broken between the 3.25-based driver in jwltest.19 and the driver that is in the later rawhide kernels...well, that gives me a place to start looking...thanks!
Don't forget that 2.6.11-1.1253_FC4 also works, even though kernels preceding it (except for 2.6.11-1.1177_FC4) and following it don't.
I just pulled 2.6.11-1.1282_FC4, and it works correctly.
Now I have some FC3 test kernels for you. (I'm not currently building FC4 test kernels, but I guess I need to start soon...) Please look here: http://people.redhat.com/linville/kernels/fc3/ These include the same driver as the later FC4 kernels, with a few changes backed-out. I'm trying to narrow down the source of the problem, so please bear with me. This may take a few revolutions... :-) Please give those a try and post the results here. Thanks!
These kernels both work: 2.6.11-1.1286_FC4 2.6.11-1.20_FC3.jwltest.8
Basically all of the FC4 kernels mentioned (both "working" and "non-working") have identical or nearly identical tg3 drivers. I'm inclined to think that the difference between working and non-working is as much luck as anything. Discussion w/ the tg3 maintainers indicates that the "tg3_stop_block timed out" messages are not always actually indicative of a problem. In fact, there are recent patches proposed to quiet those messages in certain circumstances. I wonder if you aren't seeing just those circumstances. I've applied the patches in question to my new FC4 test kernels. Wanna try them? Here is the link: http://people.redhat.com/linville/kernels/fc4/ Please let me know if you still see the messages and/or if you experience actual network (i.e. connectivity) problems.
I just reinstalled with FC4t3. The kernel it includes (2.6.11-1.1286_FC4) does *not* work, even though 2.6.11-1.1286_FC4 *did* work when I was running FC4t2. I am perfectly willing to accept that this problem depends on something other than the specific code in the tg3 driver itself. But the problem does seem to be following specific kernels. I will try both your latest test kernels and the latest development kernels, and report back.
Putting in NEEDINFO pending further test results...
2.6.11-1.1305_FC4 works. 2.6.11-1.1315_FC4 doesn't work. 2.6.11-1.1318.2.2_FC4.jwltest.2 doesn't work. 2.6.11-1.1323_FC4 doesn't work. 2.6.11-1.1329_FC4 doesn't work. It would seem as if something changed between 1305 and 1315...
The CVS log shows no changes to tg3 between 1305 and 1315 revision 1.1315 enable xen again ---------------------------- revision 1.1314 fix Xen compile ---------------------------- revision 1.1313 missing exports. ---------------------------- revision 1.1312 doh, missed off the pci id ---------------------------- revision 1.1311 e1k update. ---------------------------- revision 1.1310 add new PCI ID to Intel E1000 driver. ---------------------------- revision 1.1309 Slab debug single-bit error improvement. ---------------------------- revision 1.1308 SCSI CAM geometry fix. ---------------------------- revision 1.1307 Tickle the NMI watchdog when we're doing serial writes. ---------------------------- revision 1.1306 Bump maximum supported CPUs on x86-64 to 32. (Also fix dwmw2's time-bending antics) ---------------------------- revision 1.1305 Update audit patches
Well, as I said in comment #15, I'm perfectly willing to believe that the problem depends on something other than the actual tg3 code. But I'm not what it could be. I'm no slouch when it comes to patching and building RPMs, so I'd be willing to explore different avenues. But right now, I don't know where to take this. The only other person who (so far) has reported experiencing the same problems also has a Dell Latitude D600. Perhaps there's something flaky with this particular adapter with this particular laptop?
2.6.11-1.1363_FC4 works.
Hmmm...there has been no mention of using "acpi=off" on the kernel command line. Does that have any effect on the functionality of the adapter?
Mass update for bugs reported against -test: Updating version field to FC4 final. Please retest with final FC4 release if you have not already done so. Thanks.
Closed due to inactivity. Please reopen when requested information becomes available. Thanks!
Despite my best efforts, I've been unable to reproduce this problem with any of the FC4 kernels that have been released. I hate it when things mysteriously stop and start working, but until/unless this bug resurfaces, there's nothing further I can add. (I've left this bug as CLOSED, but changed the resolution to WORKSFORME, which I think is slightly more accurate.)