Skip to contentSkip to conversation

2.6.14-mm2 bugs and fixes

PATCH Andrew Morton 2 participants 3 messages Nov 13, 2005 1 message not in the archive

lore
Share & tools
lore
#1 lore raw
In reply to a message not in the archive
Quoted 4 lines · Bernhard Rosenkraenzer

Some problems with 2.6.14-mm2, and fixes for some:

drivers/input/power.ko needs unresolved symbol PM_IS_ACTIVE -- needs to either include <linux/pm_legacy.h> or remove the usage -- attached patch #1 does the latter.

Thanks.

Quoted 1 line

drivers/usb/core/message.c no longer exports usb_get_string, this breaks some external drivers (probably most notably ndiswrapper) -- attached patch #2 reverts this.

Agree. We shouldn't be ripping out exported symbols with zero notice like this. We have a process for doing this.

#3 lore raw
In reply to a message not in the archive
Quoted 42 lines · Bernhard Rosenkraenzer

Some problems with 2.6.14-mm2, and fixes for some:

...
There's another problem for which I haven't found a fix yet:
With 2.6.14-mm1 and 2.6.14-mm2 [on an x86 box], I'm getting a number of bad page errors (system seems to work fine nevertheless):
Bad page state at prep_new_page (in process 'tar', page c10000a0)
flags:0x00000400 mapping:00000000 mapcount:0 count:0

Backtrace:
[<c013f85b>] bad_page+0x7b/0xc0
[<c01400a5>] buffered_rmqueue+0x115/0x2c0
[<c01ae3fc>] journal_stop+0x11c/0x1e0
[<c0140349>] __alloc_pages+0xf9/0x470
[<c013b0d0>] find_lock_page+0x30/0x90
[<c013b9f8>] generic_file_buffered_write+0x178/0x6b0
[<c01a0ca9>] add_dirent_to_buf+0x159/0x380
[<c011ec5e>] current_fs_time+0x4e/0x60
[<c0175aeb>] file_update_time+0x4b/0xc0
[<c013c219>] __generic_file_aio_write_nolock+0x2e9/0x520
[<c019c012>] ext3_mark_iloc_dirty+0x1b2/0x3e0
[<c013d878>] generic_file_aio_write+0x78/0xf0
[<c019a864>] ext3_file_write+0x44/0xd0
[<c015a610>] do_sync_write+0xd0/0x110
[<c012d480>] autoremove_wake_function+0x0/0x60
[<c0167ca7>] pipe_read+0x37/0x40
[<c015b175>] vfs_write+0xd5/0x1b0
[<c015b86b>] sys_write+0x4b/0x80
[<c0102f0f>] sysenter_past_esp+0x54/0x75
Trying to fix it up, but a reboot is needed
Bad page state at prep_new_page (in process 'bzip2', page c10000c0)
flags:0x00000400 mapping:00000000 mapcount:0 count:0
Backtrace:
[<c013f85b>] bad_page+0x7b/0xc0
[<c01400a5>] buffered_rmqueue+0x115/0x2c0
[<c0140349>] __alloc_pages+0xf9/0x470
[<c0175aeb>] file_update_time+0x4b/0xc0
[<c0167661>] pipe_writev+0x331/0x590
[<c015b305>] vfs_read+0xb5/0x1b0
[<c01678f7>] pipe_write+0x37/0x40
[<c015b175>] vfs_write+0xd5/0x1b0
[<c015b86b>] sys_write+0x4b/0x80
[<c0102f0f>] sysenter_past_esp+0x54/0x75
Quoted 3 lines

Trying to fix it up, but a reboot is needed

So far I've seen those only while unpacking large tar.bz2 files (such as linux-2.6.14.tar.bz2).

That's PG_reserved. I assume someone is freeing up pages which have PG_reserved set. Do you get any traces on the __free_pages_ok() path so we can see where they're coming from?

#2 lore raw
In reply to Andrew Morton
Quoted 7 lines · Andrew Morton
Quoted 3 lines · Bernhard Rosenkraenzer

drivers/usb/core/message.c no longer exports usb_get_string, this breaks some external drivers (probably most notably ndiswrapper) -- attached patch #2 reverts this.

Agree. We shouldn't be ripping out exported symbols with zero notice like this. We have a process for doing this.

I had no idea no one else used this symbol, as nothing in-kernel does, sorry. Care to tell me why you need this function? A USB driver should not use it, but rather use the already-cached strings the USB core provides for you.

thanks,

greg k-h