Distribution: Fedora Core 3 test 2, Debian Unstable Hardware Environment: IBM A21p Problem Description: 3c59x internal adapter has stopped working for good. acpi=off used to fix the problem, but now that doesn't work either. There have been problems with this card and acpi since the first 2.6 release. I opened a new bug because there is no simple work around, that I can find, for this problem given the newest "stable" kernel. Steps to reproduce: Attempt to load system using 2.6.8.1.
specific card hardware is: 3Com PCI 3c556B Laptop Hurricane as reported by the kernel.
Do you have any kernel version work with this device? If yes, please collect the following information on working kernel and failed kernel. 1. /proc/interrupt 2. /proc/stat 3. lspci -vvxxx 4. dmesg (with debug option enabled) 5. acpidmp output (file it as attachment)or disassemble version of /proc/acpi/dsdt
Yes, I broke down and just burned cd's of fedora core 3 test 2. I'll attach the requested information shortly.
4. how do i put it (the kernel) in debug mode? 5. can i get more detail on this? .. can't seem to locate a acpidmp command on the system and don't know what a 'disassembled' output of /proc/acpi/dsdt is.
ok, I've got everything ready as far as the 'broken' version EXCEPT the #4 cuz i'm not sure what 'with debug option enabled' means. I compiled a copy of the newest 'iasl' utility from intel's site and generated the dsdt.dsl file. Once I get the information from you on item 4 I'll post everything as attachments. I need to clarify on the 'working' version though. The only way the ethernet has every worked properly on this machine since the 2.6-test versions of the kernel was with 'acpi=off' added as a boot parameter. So, what do you want? Do you want the various outputs with a 2.6 kernel and acpi=off? Or do you want me to try to boot a 2.4 kernel? I might beable to find a livecd version of linux with 2.4 on it
"Debug enabled" means you need to enable some debug kernel options, especially enable acpi debug through kernel config , as well as any other kernel debug options , which you might think they are useful for debugging. It's ok to just post information wrt working kernel ,which was with "acpi=off". Thanks a lot.
I'm just gunna post what I have. I thought by debug you meant some boot options. If its requires recompiling the kernel that'll probably have to wait. Maybe something can be spotted based on the majority of the information you requested. I'll follow with the 5 attachments shortly.
Created attachment 3699 [details] dmesg output
Created attachment 3700 [details] disassembled iasl output
Created attachment 3701 [details] lspci -vvxxx output
Created attachment 3702 [details] /proc/interrupts output
Created attachment 3703 [details] /proc/stat output
Please attach the debug information for working kernel. To narrow down the problem, please try the following debug patch, which will try vortex_start_xmit instead of boomerang_start_xmit, which seems to be impuissant. [root@titan1 work]# vi 3c59x.diff --- /usr/out/linux-2.6.8.1/drivers/net/3c59x.c 2004-08-14 18:54:46.000000000 +0800 +++ linux-2.6.8.1/drivers/net/3c59x.c 2004-10-09 18:05:19.000000000 +0800 @@ -1347,7 +1347,7 @@ /* Extract our information from the EEPROM data. */ vp->info1 = eeprom[13]; vp->info2 = eeprom[15]; - vp->capabilities = eeprom[16]; + vp->capabilities = eeprom[16] & 0xFFDF; if (vp->info1 & 0x8000) { vp->full_duplex = 1;
I have 3C905-TX card in DEC Venturis GL 6200 with kernel 2.6.5-1.358 (Fedore Core 2). The driver, 3c59x, is loaded, but the network is unoperational. I also get dump on the screen for each broadcast message sent over local LAN (I have a Di-624 router).
since this fails also with acpi=off, i'm moving it out of the ACPI category.
I believe this and bug #3122, and bug #3889 are dupes of bug #541 and related to bug #1569. I am having this problem with an IBM Thinkpad T20, and vanilla kernel 2.6.10 with ACPI. I'll see if I can insert the code in comment #13, but I see in bug #541 that Dave Jones wrote: --- Note the ff's everywhere. Looks like it didn't power up the eeprom. --- If this is true, then would the fix in comment #13 be of any use? I'll see if there is any way to try to manually (on the command line) try to power up the network interface. FWIW, comparing lspci -vv between a 2.4.28 kernel (without acpi) and the 2.6.10 kernel (with acpi) gives me the following differences: 2.4.28 w/o ACPI Latency: 80 (cache line size 08) Region 0: I/O ports at 1800 Region 1: mem@e8101400 size=128 Region 2: mem@e8101000 size=128 2.6.10 Latency: 64 (no cache line size information) Region 0: I/O ports at 1400 Region 1: mem@18000000 size=128 Region 2: mem@18000080 size=128 I don't know if any of those differences are interesting.
What is the status of this issue in recent 2.6 kernels?
Please reopen this bug if it's still present in kernel 2.6.17.