Bug#711113: linux-image-3.2.0-4-686-pae: Please merge commit "x86-32: Fix ... incomplete TLB invalidate with PAE pagetables" from upstream
Hi Ben,
Thank you very much for your timely and thoughtful replies.
> This is commit 1de14c3c5cbc9bb17e9dcc648cda51c0c85d54b9. The full
> commit message says about the bug: "Ingo guesses this got introduced by
> commit 611ae8e3f520 ("x86/tlb: enable tlb flush range support for x86")
> since that code started to free unused pagetables." And that was in
> Linux 3.6. So, what makes you think that you're seeing the same bug in
> Linux 3.2?
At http://lwn.net/Articles/547282/ Linus wrote that
commit 1de14c3c5cbc9bb17e9dcc648cda51c0c85d54b9 may
corrupted-page-table fix bugs triggered by Google Chrome. I, too,
have been getting corrupted-page-table bugs triggered by Chromium, so
I assumed
that commit 1de14c3c5cbc9bb17e9dcc648cda51c0c85d54b9 was the solution.
But your replies made me realize that perhaps I was wrong. I did
upgrade from a non-PAE kernel (on Squeeze) to a PAE kernel (on
Wheezy), so in truth, the bugs could be caused by anything.
> [...]
>
> Please send the log of the *first* BUG/oops message, which will include
> the words 'Not tainted'.
I couldn't find that message in my /var/log directory. So, maybe an
hour ago, I rebooted back into the 3.2 kernel. Here's the
bad-pagetable error I got some time after that reboot:
[ 371.239303] chromium: Corrupted page table at address b1748d40
[ 371.239388] *pdpt = 000000001c606001 *pde = 000000003a212067
[ 371.239465] Bad pagetable: 001d [#1] SMP
[ 371.239524] Modules linked in: parport_pc ppdev lp parport radeon
ttm drm_kms_helper drm i2c_algo_bit binfmt_misc fuse
snd_hda_codec_realtek snd_hda_codec_si3054 snd_hda_intel snd_hda_codec
snd_hwdep snd_pcm_oss snd_mixer_oss snd_pcm joydev snd_page_alloc
snd_seq_midi snd_seq_midi_event snd_rawmidi arc4 ath5k ath mac80211
snd_seq pcmcia cfg80211 snd_seq_device snd_timer i2c_piix4 snd
yenta_socket pcmcia_rsrc pcmcia_core shpchp soundcore i2c_core psmouse
serio_raw battery ac power_supply pcspkr evdev video processor button
toshiba_acpi sparse_keymap rfkill ext3 mbcache jbd sg sr_mod sd_mod
crc_t10dif cdrom ata_generic usbhid hid pata_atiixp sata_sil libata
ohci_hcd thermal fan thermal_sys ehci_hcd scsi_mod 8139too 8139cp mii
usbcore usb_common [last unloaded: scsi_wait_scan]
[ 371.240019]
[ 371.240019] Pid: 3801, comm: chromium Not tainted 3.2.0-4-686-pae
#1 Debian 3.2.41-2 TOSHIBA Satellite A100/SB450
[ 371.240019] EIP: 0073:[<b1748d40>] EFLAGS: 00010286 CPU: 0
[ 371.240019] EIP is at 0xb1748d40
[ 371.240019] EAX: b898a8fc EBX: b174eff4 ECX: bfe12728 EDX: 00000030
[ 371.240019] ESI: b8abe280 EDI: 00000000 EBP: b8fef530 ESP: bfe1261c
[ 371.240019] DS: 007b ES: 007b FS: 0000 GS: 0033 SS: 007b
[ 371.240019] Process chromium (pid: 3801, ti=e4066000 task=f31d0de0
task.ti=e4066000)
[ 371.240019]
[ 371.240019] EIP: [<b1748d40>] 0xb1748d40 SS:ESP 007b:bfe1261c
[ 371.254907] ---[ end trace 2980570486fbde78 ]---
> I looked at this some more, and this fix at least is only applicable to
> the code Ingo referred to, which doesn't exist in Linux 3.2.
OK. So the subject line of my bug is invalid.
The 3.2 kernel has been giving me at least a couple bad-pagetable
errors per hour. But despite this, I haven't found a reliable way to
trigger a bad-pagetable error.
Should I file a new bug with a better subject line, and after I do so,
you can close this bug as "invalid"?
Alternatively, should this bug be closed as "invalid" and no
replacement be filed? I can keep on using my Squeeze kernel for the
next few years: it works fine on Wheezy, and my machine is merely a
home laptop.
Cheers!
--unforgettableid
Reply to: