Bug#613979: [alsa-devel] Problems with snd_hda_intel in Linux kernel 2.6.38
From: Paul Menzel paulepan...@users.sourceforge.net To: alsa-de...@alsa-project.org Cc: 619...@bugs.debian.org, s...@kth.se, 613...@bugs.debian.org Subject: Re: [alsa-devel] Problems with snd_hda_intel in Linux kernel 2.6.38 Date: Tue, 24 May 2011 12:23:31 +0200 ... The point where it Oops implies that the problem isn't in the sound driver but rather in a breakage in a deeper level, either PCI core, x86 mm or ACPI/BIOS. The problem is still present with 2.6.39. --- Svante, if you have some time, could you please try to follow the Wiki page DebianKernel/GitBisect [1] in the Debian Wiki. Unfortunately I will not have time until the beginning of July. Thanks, Paul [1] http://wiki.debian.org/DebianKernel/GitBisect Sorry, I missed your reply due to [alsa-devel] in the title and the high traffic in the alsa-devel mailing list. I have now unsubscribed from that mailing list since the problems are probably due to other reasons. I will try to bisect when I find the time. Thank you for the link. -- To UNSUBSCRIBE, email to debian-kernel-requ...@lists.debian.org with a subject of unsubscribe. Trouble? Contact listmas...@lists.debian.org Archive: http://lists.debian.org/1308814824.31174.73.ca...@hp.my.own.domain
Bug#613979: [alsa-devel] Problems with snd_hda_intel in Linux kernel 2.6.38
Am Sonntag, den 22.05.2011, 19:56 +0200 schrieb Paul Menzel: Am Montag, den 04.04.2011, 11:21 +0200 schrieb Svante Signell: On Mon, 2011-04-04 at 11:12 +0200, Takashi Iwai wrote: At Mon, 04 Apr 2011 10:42:57 +0200, Svante Signell wrote: On Thu, 2011-03-31 at 00:42 +0200, Svante Signell wrote: On Wed, 2011-03-30 at 15:13 +0200, Clemens Ladisch wrote: ... Anything happening here with respect to this bug? How can I help further? Booting with 2.6.32 all the time does not feel like a good solution in long term. The point where it Oops implies that the problem isn't in the sound driver but rather in a breakage in a deeper level, either PCI core, x86 mm or ACPI/BIOS. The problem is still present with 2.6.39. Any chance to bisect the kernel? Never done that before. Is there a bisect HOWTO somewhere? Ben, I am sorry to bother you directly, but there are so many howtos on the Web, it would be great if you could point us to an “official” one, which has proofed itself. Or maybe you could write up a blog post. ;-) Anyway I did not find any information here. $ ls /usr/share/doc/linux-image-2.6.39-1-amd64/ changelog.Debian.gz copyright Svante, if you have some time, could you please try to follow the Wiki page DebianKernel/GitBisect [1] in the Debian Wiki. Unfortunately I will not have time until the beginning of July. Thanks, Paul [1] http://wiki.debian.org/DebianKernel/GitBisect signature.asc Description: This is a digitally signed message part
Bug#613979: [alsa-devel] Problems with snd_hda_intel in Linux kernel 2.6.38
Am Montag, den 04.04.2011, 11:21 +0200 schrieb Svante Signell: On Mon, 2011-04-04 at 11:12 +0200, Takashi Iwai wrote: At Mon, 04 Apr 2011 10:42:57 +0200, Svante Signell wrote: On Thu, 2011-03-31 at 00:42 +0200, Svante Signell wrote: On Wed, 2011-03-30 at 15:13 +0200, Clemens Ladisch wrote: ... Anything happening here with respect to this bug? How can I help further? Booting with 2.6.32 all the time does not feel like a good solution in long term. The point where it Oops implies that the problem isn't in the sound driver but rather in a breakage in a deeper level, either PCI core, x86 mm or ACPI/BIOS. The problem is still present with 2.6.39. Any chance to bisect the kernel? Never done that before. Is there a bisect HOWTO somewhere? Ben, I am sorry to bother you directly, but there are so many howtos on the Web, it would be great if you could point us to an “official” one, which has proofed itself. Or maybe you could write up a blog post. ;-) Anyway I did not find any information here. $ ls /usr/share/doc/linux-image-2.6.39-1-amd64/ changelog.Debian.gz copyright Thanks, Paul signature.asc Description: This is a digitally signed message part
Bug#613979: [alsa-devel] Problems with snd_hda_intel in Linux kernel 2.6.38
At Mon, 04 Apr 2011 10:42:57 +0200, Svante Signell wrote: On Thu, 2011-03-31 at 00:42 +0200, Svante Signell wrote: On Wed, 2011-03-30 at 15:13 +0200, Clemens Ladisch wrote: Svante Signell wrote: Code: f4 01 00 00 ef 31 f6 48 89 df e8 15 dd ff ff 85 c0 0f 88 2b 03 00 00 48 89 ef e8 ee 11 b9 e0 8b 7b 40 e8 9f 25 a7 e0 48 8b 43 38 66 8b 10 66 89 14 24 8b 43 14 83 e8 03 83 f8 01 77 32 31 d2 be 5: 31 f6 xor%esi,%esi 7: 48 89 dfmov%rbx,%rdi a: e8 15 dd ff ff callq 0xdd24 f: 85 c0 test %eax,%eax 11: 0f 88 2b 03 00 00 js 0x342 17: 48 89 efmov%rbp,%rdi 1a: e8 ee 11 b9 e0 callq 0xe0b9120d 1f: 8b 7b 40mov0x40(%rbx),%edi 22: e8 9f 25 a7 e0 callq 0xe0a725c6 27: 48 8b 43 38 mov0x38(%rbx),%rax 2b: 66 8b 10mov(%rax),%dx -- crash here 2e: 66 89 14 24 mov%dx,(%rsp) 32: 8b 43 14mov0x14(%rbx),%eax 35: 83 e8 03sub$0x3,%eax 38: 83 f8 01cmp$0x1,%eax 3b: 77 32 ja 0x6f 3d: 31 d2 xor%edx,%edx This is the azx_readw(chip, GCAP) in azx_create(); chip-remap_addr is 0xc90011c08000 which does look like a valid pointer, but isn't. Thank you Clemens! Maybe your input is sufficient to solve this problem. I have now installed the debug version of the kernel, the objdump output is attached (please let me know if you are missing something).sorry, I don't know where to find the relevant information in this file, but that is all I have (still very large). (Does not include the error messages on stderr, maybe something is still missing.) Anything happening here with respect to this bug? How can I help further? Booting with 2.6.32 all the time does not feel lika a good solution in long term. The point where it Oops implies that the problem isn't in the sound driver but rather in a breakage in a deeper level, either PCI core, x86 mm or ACPI/BIOS. Any chance to bisect the kernel? thanks, Takashi -- To UNSUBSCRIBE, email to debian-kernel-requ...@lists.debian.org with a subject of unsubscribe. Trouble? Contact listmas...@lists.debian.org Archive: http://lists.debian.org/s5h8vvqv0r2.wl%ti...@suse.de
Bug#613979: [alsa-devel] Problems with snd_hda_intel in Linux kernel 2.6.38
On Mon, 2011-04-04 at 11:12 +0200, Takashi Iwai wrote: At Mon, 04 Apr 2011 10:42:57 +0200, Svante Signell wrote: On Thu, 2011-03-31 at 00:42 +0200, Svante Signell wrote: On Wed, 2011-03-30 at 15:13 +0200, Clemens Ladisch wrote: ... Anything happening here with respect to this bug? How can I help further? Booting with 2.6.32 all the time does not feel like a good solution in long term. The point where it Oops implies that the problem isn't in the sound driver but rather in a breakage in a deeper level, either PCI core, x86 mm or ACPI/BIOS. Any chance to bisect the kernel? Never done that before. Is there a bisect HOWTO somewhere? -- To UNSUBSCRIBE, email to debian-kernel-requ...@lists.debian.org with a subject of unsubscribe. Trouble? Contact listmas...@lists.debian.org Archive: http://lists.debian.org/1301908906.32453.94.ca...@s1499.it.kth.se
Bug#613979: [alsa-devel] Problems with snd_hda_intel in Linux kernel 2.6.38
On Tue, 2011-03-29 at 13:10 +0200, Takashi Iwai wrote: At Tue, 29 Mar 2011 12:58:16 +0200, Svante Signell wrote: On Tue, 2011-03-29 at 12:31 +0200, Takashi Iwai wrote: At Tue, 29 Mar 2011 12:24:40 +0200, Svante Signell wrote: ... But let's check the Oops first as below. Also, please try to decode the line from the code shown in the Oops. It's a bit too little information to analyze, unfortunately. ... As mentioned, you can decode the binary dump in Oops to guess which line of the source code corresponds to the Oops point. Use gdb or objdump to figure out the disassembled code. For example, % objdump -D -l /lib/modules/$(uname -r)/kernel/sound/pci/hda/snd-hda-intel.ko Then look for azx_probe. Calculate the position from the offset Oops gave, compare the hex codes with the data show in Code section of Oops. objdump with -l will show the source code line as well, so you'll see now more exactly where it was triggered. Below is the kernel Oops and the objdump output related to azx_probe. Unfortunately I don't know where to find the Oops offset! Kernel Oops when booting: = [4.631033] Oops: 0009 [#1] SMP [4.631187] last sysfs file: /sys/devices/virtual/net/lo/operstate [4.631243] CPU 0 [4.631293] Modules linked in: snd_hda_intel(+) snd_hda_codec tpm_tis tpm pcspkr snd_hwdep tpm_bios shpchp(+) pci_hotplug k8temp nouveau(+) snd_pcm ttm drm_kms_helper drm parport_pc i2c_viapro i2c_algo_bit usblp power_supply i2c_core parport edac_core video edac_mce_amd processor psmouse evdev serio_raw button snd_seq snd_timer snd_seq_device snd soundcore snd_page_alloc thermal_sys ext3 jbd mbcache sg sr_mod cdrom usbhid sd_mod crc_t10dif ata_generic hid sata_via uhci_hcd pata_via libata ehci_hcd usbcore scsi_mod via_rhine floppy mii nls_base [last unloaded: scsi_wait_scan] [4.632005] [4.632005] Pid: 632, comm: work_for_cpu Not tainted 2.6.38-1-amd64 #1 MICRO-STAR INTERNATIONAL CO., LTD MS-7253/MS-7253 [4.632005] RIP: 0010:[a061f416] [a061f416] azx_probe+0x3ad/0x870 [snd_hda_intel] [4.632005] RSP: 0018:88007c05be50 EFLAGS: 00010286 [4.632005] RAX: c90013c98000 RBX: 880036de6000 RCX: 0006 [4.632005] RDX: RSI: 0246 RDI: 0246 [4.632005] RBP: 88007c93d000 R08: R09: 0040 [4.632005] R10: 0286 R11: a971 R12: 880036de5c00 [4.632005] R13: R14: R15: 88007c93d090 [4.632005] FS: 7f1cd2afb7a0() GS:88007fc0() knlGS: [4.632005] CS: 0010 DS: ES: CR0: 8005003b [4.632005] CR2: c90013c98000 CR3: 7c00f000 CR4: 06f0 [4.632005] DR0: DR1: DR2: [4.632005] DR3: DR6: 0ff0 DR7: 0400 [4.632005] Process work_for_cpu (pid: 632, threadinfo 88007c05a000, task 88003713a880) [4.632005] Stack: [4.632005] 0005 81240724 0001 880036de5c00 [4.632005] 00013700 88007c93d000 88007c93d090 88007a7c7dc8 [4.632005] 88007c93d200 811b1a42 [4.632005] Call Trace: [4.632005] [81240724] ? __pm_runtime_set_status +0x162/0x186 [4.632005] [811b1a42] ? local_pci_probe+0x49/0x92 [4.632005] [8105aad2] ? do_work_for_cpu+0x0/0x1b [4.632005] [8105aad2] ? do_work_for_cpu+0x0/0x1b [4.632005] [8105aadd] ? do_work_for_cpu+0xb/0x1b [4.632005] [8105fcdf] ? kthread+0x7a/0x82 [4.632005] [8100a764] ? kernel_thread_helper+0x4/0x10 [4.632005] [8105fc65] ? kthread+0x0/0x82 [4.632005] [8100a760] ? kernel_thread_helper+0x0/0x10 [4.632005] Code: f4 01 00 00 ef 31 f6 48 89 df e8 15 dd ff ff 85 c0 0f 88 2b 03 00 00 48 89 ef e8 ee 11 b9 e0 8b 7b 40 e8 9f 25 a7 e0 48 8b 43 38 66 8b 10 66 89 14 24 8b 43 14 83 e8 03 83 f8 01 77 32 31 d2 be [4.632005] RIP [a061f416] azx_probe+0x3ad/0x870 [snd_hda_intel] [4.632005] RSP 88007c05be50 [4.632005] CR2: c90013c98000 [4.632005] ---[ end trace c6748815fe9ff43b ]--- objdump -D -l /lib/modules/$(uname -r)/kernel/sound/pci/hda/snd-hda-intel.ko /lib/modules/2.6.38-1-amd64/kernel/sound/pci/hda/snd-hda-intel.ko: file format elf64-x86-64 01fd azx_probe: azx_probe(): (see attachment) /lib/modules/2.6.38-1-amd64/kernel/sound/pci/hda/snd-hda-intel.ko: file format elf64-x86-64 01fd azx_probe: azx_probe(): 1fd: 41 57 push %r15 1ff: 41 56 push %r14 201: 41 be ed ff ff ff mov$0xffed,%r14d 207: 41 55 push %r13 209: 41 54 push %r12 20b: 55
Bug#613979: [alsa-devel] Problems with snd_hda_intel in Linux kernel 2.6.38
At Wed, 30 Mar 2011 12:25:44 +0200, Svante Signell wrote: On Tue, 2011-03-29 at 13:10 +0200, Takashi Iwai wrote: At Tue, 29 Mar 2011 12:58:16 +0200, Svante Signell wrote: On Tue, 2011-03-29 at 12:31 +0200, Takashi Iwai wrote: At Tue, 29 Mar 2011 12:24:40 +0200, Svante Signell wrote: ... But let's check the Oops first as below. Also, please try to decode the line from the code shown in the Oops. It's a bit too little information to analyze, unfortunately. ... As mentioned, you can decode the binary dump in Oops to guess which line of the source code corresponds to the Oops point. Use gdb or objdump to figure out the disassembled code. For example, % objdump -D -l /lib/modules/$(uname -r)/kernel/sound/pci/hda/snd-hda-intel.ko Then look for azx_probe. Calculate the position from the offset Oops gave, compare the hex codes with the data show in Code section of Oops. objdump with -l will show the source code line as well, so you'll see now more exactly where it was triggered. Below is the kernel Oops and the objdump output related to azx_probe. Unfortunately I don't know where to find the Oops offset! This is shown in below: [4.632005] RIP: 0010:[a061f416] [a061f416] azx_probe+0x3ad/0x870 [snd_hda_intel] The offset is 0x3ad. As azx_probe() in the disassembled code begins with 0x1fd, it points to 0x5aa (0x1fd + 0x3ad). You can see the disassembled code matches with the dump in Code: in Oops. However, you objdump output doesn't give the line number. Did you install the corresponding debug package? Usually this is stripped in the main package and provided as an add-on. thanks, Takashi -- To UNSUBSCRIBE, email to debian-kernel-requ...@lists.debian.org with a subject of unsubscribe. Trouble? Contact listmas...@lists.debian.org Archive: http://lists.debian.org/s5hpqp827sv.wl%ti...@suse.de
Bug#613979: [alsa-devel] Problems with snd_hda_intel in Linux kernel 2.6.38
On Wed, 2011-03-30 at 12:59 +0200, Takashi Iwai wrote: At Wed, 30 Mar 2011 12:25:44 +0200, Svante Signell wrote: On Tue, 2011-03-29 at 13:10 +0200, Takashi Iwai wrote: At Tue, 29 Mar 2011 12:58:16 +0200, Svante Signell wrote: On Tue, 2011-03-29 at 12:31 +0200, Takashi Iwai wrote: At Tue, 29 Mar 2011 12:24:40 +0200, Svante Signell wrote: ... But let's check the Oops first as below. ... The offset is 0x3ad. As azx_probe() in the disassembled code begins with 0x1fd, it points to 0x5aa (0x1fd + 0x3ad). You can see the disassembled code matches with the dump in Code: in Oops. Sorry, I did not see any match in the objdump and Oops Code. Code: f4 01 00 00 ef 31 f6 48 89 df e8 15 dd ff ff 85 c0 0f 88 2b 03 00 00 48 89 ef e8 ee 11 b9 e0 8b 7b 40 e8 9f 25 a7 e0 48 8b 43 38 66 8b 10 66 89 14 24 8b 43 14 83 e8 03 83 f8 01 77 32 31 d2 be Never mind, tanks anyway. However, you objdump output doesn't give the line number. Did you install the corresponding debug package? Usually this is stripped in the main package and provided as an add-on. You mean that I need to install a debug version of the kernel? If yes I will do that when having physical access to that computer again (probably tonight or tomorrow) -- To UNSUBSCRIBE, email to debian-kernel-requ...@lists.debian.org with a subject of unsubscribe. Trouble? Contact listmas...@lists.debian.org Archive: http://lists.debian.org/1301486370.32453.9.ca...@s1499.it.kth.se
Bug#613979: [alsa-devel] Problems with snd_hda_intel in Linux kernel 2.6.38
Svante Signell wrote: Code: f4 01 00 00 ef 31 f6 48 89 df e8 15 dd ff ff 85 c0 0f 88 2b 03 00 00 48 89 ef e8 ee 11 b9 e0 8b 7b 40 e8 9f 25 a7 e0 48 8b 43 38 66 8b 10 66 89 14 24 8b 43 14 83 e8 03 83 f8 01 77 32 31 d2 be 5: 31 f6 xor%esi,%esi 7: 48 89 dfmov%rbx,%rdi a: e8 15 dd ff ff callq 0xdd24 f: 85 c0 test %eax,%eax 11: 0f 88 2b 03 00 00 js 0x342 17: 48 89 efmov%rbp,%rdi 1a: e8 ee 11 b9 e0 callq 0xe0b9120d 1f: 8b 7b 40mov0x40(%rbx),%edi 22: e8 9f 25 a7 e0 callq 0xe0a725c6 27: 48 8b 43 38 mov0x38(%rbx),%rax 2b: 66 8b 10mov(%rax),%dx -- crash here 2e: 66 89 14 24 mov%dx,(%rsp) 32: 8b 43 14mov0x14(%rbx),%eax 35: 83 e8 03sub$0x3,%eax 38: 83 f8 01cmp$0x1,%eax 3b: 77 32 ja 0x6f 3d: 31 d2 xor%edx,%edx This is the azx_readw(chip, GCAP) in azx_create(); chip-remap_addr is 0xc90011c08000 which does look like a valid pointer, but isn't. Regards, Clemens -- To UNSUBSCRIBE, email to debian-kernel-requ...@lists.debian.org with a subject of unsubscribe. Trouble? Contact listmas...@lists.debian.org Archive: http://lists.debian.org/4d932c71.10...@ladisch.de