[Devel] Re: /proc/*/pagemap BUG: sleeping function called from invalid context

[Matt Mackall]

On Mon, Apr 09, 2007 at 12:25:54PM +0400, Alexey Dobriyan wrote:
> After
> 	cat /proc/self/pagemap
> 
> BUG: sleeping function called from invalid context at include/asm/uaccess.h:453
> in_atomic():1, irqs_disabled():0
> 1 lock held by cat/14183:
>  #0:  (&mm->mmap_sem){----}, at: [<c017d17b>] pagemap_read+0x11f/0x21b
>  [<c01b7bc7>] copy_to_user+0x37/0x4c
>  [<c017cf92>] add_to_pagemap+0x49/0x6f
>  [<c017d034>] pagemap_pte_range+0x56/0x7e
>  [<c017cfde>] pagemap_pte_range+0x0/0x7e
>  [<c01b481d>] walk_page_range+0xf1/0x1a0
>  [<c017d1e3>] pagemap_read+0x187/0x21b
>  [<c01fdafa>] tty_write+0x1bb/0x1cc
>  [<c017d05c>] pagemap_read+0x0/0x21b
>  [<c01589b0>] vfs_read+0x72/0x95
>  [<c0158cce>] sys_read+0x41/0x67
>  [<c0103d96>] sysenter_past_esp+0x8f/0x99
>  [<c0103d66>] sysenter_past_esp+0x5f/0x99

I'm still waking up but this is odd. mm->mmap_sem is an rwsem,
sleeping inside of it should be perfectly valid.

Oh. You've got CONFIG_HIGHPTE and the note is really about the
kmap_atomic inside pte_offset_map. I'll fix that up.

-- 
Mathematics is the supreme nostalgia of our time.