linux-kernel - 2.6.20-rc7 regression on intel-agp

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [thread-next>] [day] [month] [year] [list]

Date:	Sun, 04 Feb 2007 16:51:38 +0100
From:	Eric Piel <Eric.Piel@...mplin-utc.net>
To:	davej@...emonkey.org.uk, Wang Zhenyu <zhenyu.z.wang@...el.com>
CC:	linux-kernel@...r.kernel.org, Adrian Bunk <bunk@...sta.de>
Subject: 2.6.20-rc7 regression on intel-agp

Hello,

I've got a regression in 2.6.20-rc7 (-rc6 was fine) due to commit 
4b95320fc4d21b0ff2f8604305dd6c851aff6096 ([AGPGART] intel_agp: restore 
graphics device's pci space early in resume).

Although the patch does have good effects after a suspend to disk, 
because it avoid to have a garbled console, it also generates an oops 
(breaking the X session):
BUG: unable to handle kernel paging request at virtual address 0001c000
  printing eip:
c023d954
*pde = 1b480067
*pte = 00000000
Oops: 0002 [#1]
PREEMPT
Modules linked in: i810 drm snd_seq_oss snd_seq_midi_event snd_seq 
snd_seq_device e100 mii snd_pcm_oss snd_mixer_oss af_packet snd_intel8x0 
snd_ac97_codec ac97_bus snd_pcm snd_timer snd_page_alloc ipv6 snd 
soundcore ipt_MASQUERADE iptable_nat nf_nat ipt_REJECT xt_tcpudp 
nf_conntrack_ipv4 xt_state nf_conntrack iptable_filter ip_tables 
x_tables binfmt_misc loop nls_utf8 nls_cp437 vfat fat video thermal 
processor fan button battery orinoco_cs ac orinoco hermes pcmcia 
firmware_class joydev yenta_socket rsrc_nonstatic pcmcia_core 
cpufreq_ondemand cpufreq_conservative cpufreq_powersave speedstep_ich 
speedstep_lib evdev psmouse pcspkr wistron_btns uhci_hcd usbcore sr_mod 
cdrom
CPU:    0
EIP:    0060:[<c023d954>]    Not tainted VLI
EFLAGS: 00213206   (2.6.20-rc7-eric #9)
EIP is at intel_i810_insert_entries+0xf4/0x210
eax: 03000003   ebx: 0000c004   ecx: 03001000   edx: 0000c000
esi: 00003001   edi: dd662720   ebp: 00000000   esp: de955eb0
ds: 007b   es: 007b   ss: 0068
Process Xorg (pid: 2425, ti=de955000 task=de4f4ab0 task.ti=de955000)
Stack: e0820000 dfc177a0 00003400 00003000 c02e5f00 de955edc c03a0060 
dd662720
        00003000 dd780c40 c023a9d8 00203097 c03a4364 de955f18 dcc59c00 
00003000
        e4b99263 00000000 00000000 00000000 00000000 e4b992b0 dcc59c00 
00000036
Call Trace:
  [<c023a9d8>] agp_bind_memory+0x38/0x80
  [<e4b99263>] drm_agp_bind+0x73/0xc0 [drm]
  [<e4b992b0>] drm_agp_bind_ioctl+0x0/0x50 [drm]
  [<e4b992e6>] drm_agp_bind_ioctl+0x36/0x50 [drm]
  [<e4b992b0>] drm_agp_bind_ioctl+0x0/0x50 [drm]
  [<e4b94c1c>] drm_ioctl+0xac/0x200 [drm]
  [<c016c9c8>] do_ioctl+0x78/0x90
  [<c016ca3c>] vfs_ioctl+0x5c/0x2b0
  [<c015fe00>] do_sync_write+0x0/0x110
  [<c016cd02>] sys_ioctl+0x72/0x90
  [<c0103032>] sysenter_past_esp+0x5f/0x85
  =======================
Code: 24 0c 01 c8 39 c6 73 36 89 f3 89 f1 c1 e3 02 c1 e1 0c 8b 2d 5c 76 
40 c0 89 c8 89 da 83 c8 03 46 83 c3 04 81 c1 00 10 00 00 01 ea <89> 82 
00 00 01 00 8b 44 24 0c 8b 6f 10 01 e8 39 f0 77 d4 8d 04
EIP: [<c023d954>] intel_i810_insert_entries+0xf4/0x210 SS:ESP 0068:de955eb0

(full dmesg is attached).

Reverting the commit does avoid the oops, and permits to resume (from 
disk) to the X session fine (as it used to). Maybe it worths noting that 
my graphic chipset is a "Intel 82815 CGC (rev 11)" and in addition to X, 
I use i810fb with those parameters:
video=i810fb:xres:1024,yres:768,hsync1:32,hsync2:48,vsync1:50,vsync2:70,vram:2,bpp:16,mtrr:1

Hopefully we can find a solution which can at the same time not garble 
the console and not oops...

See you,
Eric

View attachment "i810-ooops-resumefromdisk.dmesg" of type "text/plain" (41170 bytes)