Yet more followup with myself.... I can reproduce this problem on
2.4.0-test10-pre1 every time. I'm using the ide-scsi and usb-storage
modules to trigger the bug -- loading and then unloading either one causes
/proc/scsi to not be cleaned up properly.
As yet, nobody has indicated to me that they are looking into this problem.
I've taken a few experimental pokes at it, and it seems that the SCSI layer
is, in fact, calling the procfs function to remove the entries. At least,
I think it is -- I'm not sure if it's the right entry.
Will someone _please_ look at this? I consider this a critical item for
2.4.0, and I hope others do too.
Matt
On Mon, Oct 02, 2000 at 06:00:53PM -0700, Matthew Dharm wrote:
> Following up yet some more with myself... is anyone actually looking at
> this stuff? I can provide even more information if desired, but I'd really
> like to know that at least one person who understands this code is looking
> at it....
>
> Anyway, I managed to get a better OOPS trace. Here it is:
>
> ksymoops 0.7c on i586 2.4.0-test9. Options used
> -V (default)
> -k /proc/ksyms (default)
> -l /proc/modules (default)
> -o /lib/modules/2.4.0-test9/ (default)
> -m /boot/System.map (default)
>
> Warning: You did not tell me where to find symbol information. I will
> assume that the log matches the kernel and modules that are running
> right now and I'll use the default options above for symbol resolution.
> If the current kernel and/or modules do not match the log, you can get
> more accurate output by telling me the kernel version and where to find
> map, modules, ksyms etc. ksymoops -h explains the options.
>
> Reading Oops report from the terminal
> Unable to handle kernel paging request at virtual address d38c2010
> c0145032
> *pde = 01594063
> Oops: 0002
> CPU: 0
> EIP: 0010:[<c0145032>]
> Using defaults from ksymoops -t elf32-i386 -a i386
> EFLAGS: 00010286
> eax: d38c2000 ebx: cf93f0a0 ecx: c1466fc0 edx: 00000023
> esi: c6c1a360 edi: cf93f0f4 ebp: ffffffea esp: c8be7ef0
> ds: 0018 es: 0018 ss: 0018
> Process ls (pid: 11563, stackpage=c8be7000)
> Stack: c7f81360 c49c17e0 c0146a9c c144b800 00001190 cf93f0a0 fffffff4 c7f81360
> c49c17e0 c49c1834 c0137083 c49c17e0 c7f81360 00000000 00000000 c8be7f68
> c9f0f013 c01377c4 c7f812e0 c8be7f68 00000000 c9f0f000 00000000 c8be7fa4
> Call Trace: [<c0146a9c>] [<c0137083>] [<c01377c4>] [<f27a545f>] [<c0137daa>]
>[<c0134a71>] [<c010a413>]
> Code: ff 40 10 8b 43 24 80 48 14 18 66 8b 43 08 25 00 f0 ff ff 66
>
> >>EIP; c0145032 <proc_get_inode+96/108> <=====
> Trace; c0146a9c <proc_lookup+68/8c>
> Trace; c0137083 <real_lookup+4f/b8>
> Trace; c01377c4 <path_walk+5a0/7ec>
> Trace; f27a545f <END_OF_CODE+1eebbdb8/????>
> Trace; c0137daa <__user_walk+3a/54>
> Trace; c0134a71 <sys_newlstat+15/70>
> Trace; c010a413 <system_call+33/40>
> Code; c0145032 <proc_get_inode+96/108>
> 00000000 <_EIP>:
> Code; c0145032 <proc_get_inode+96/108> <=====
> 0: ff 40 10 incl 0x10(%eax) <=====
> Code; c0145035 <proc_get_inode+99/108>
> 3: 8b 43 24 movl 0x24(%ebx),%eax
> Code; c0145038 <proc_get_inode+9c/108>
> 6: 80 48 14 18 orb $0x18,0x14(%eax)
> Code; c014503c <proc_get_inode+a0/108>
> a: 66 8b 43 08 movw 0x8(%ebx),%ax
> Code; c0145040 <proc_get_inode+a4/108>
> e: 25 00 f0 ff ff andl $0xfffff000,%eax
> Code; c0145045 <proc_get_inode+a9/108>
> 13: 66 00 00 addb %al,(%eax)
>
>
> 1 warning issued. Results may not be reliable.
>
> This is under the same test scenario as before -- load ide-scsi, unload
> ide-scsi, ls /proc/scsi
>
> It's interesting to note that 'cat /proc/scsi/scsi' works just fine -- only
> getting the directory listing seems to be broken, not the actual reading of
> files.
>
> Again, this was collected on 2.4.0-test9-pre7 -- but the behavior is the
> same for 2.4.0-test9-pre9.
>
> Matt
>
> On Sun, Oct 01, 2000 at 06:32:38PM -0700, Matthew Dharm wrote:
> > Just to follow-up to my own post, I have some more datapoints...
> >
> > The bug definatley seems to be in either the SCSI layer or the procfs
> > layer. The behavior is the same if I use either ide-scsi or usb-storage,
> > which are the only two SCSI modules I can test.
> >
> > Matt
> >
> > On Fri, Sep 29, 2000 at 03:19:23PM -0700, Matthew Dharm wrote:
> > > I'm using 2.4.0-test9-pre7 and have a _very_ reproducable OOPS with the
> > > SCSI layer. Everything relevant is compiled as a module (except for the
> > > /proc support).
> > >
> > > The test scenario is this:
> > > (1) Boot the machine
> > > (2) modprobe ide-scsi (note this autoloads scsi_mod but nothing else)
> > > (3) rmmod ide-scsi
> > > (4) ls /proc/scsi
> > > Then the OOPS happens.
> > >
> > > If, instead of step (4), I instead re-modprobe ide-scsi again, then an 'ls
> > > /proc/scsi/' will show (without an immediate OOPS):
> > >
> > > [root@zen mdharm]# ls /proc/scsi/
> > > ide-scsi/ ide-scsi/ scsi
> > >
> > > No, the double ide-scsi entry isn't a typo -- this is what ls reports as
> > > being there.
> > >
> > > The OOPS case is decoded via ksymoops below. My intuition suggests that
> > > this is related to the fact that /proc seems to always be busy (and
> > > therefore not umountable) at shutdown.
> > >
> > > I'm more than willing to test patches that might fix the problem.
> > >
> > > Sep 29 15:05:01 zen kernel: Unable to handle kernel paging request at
> > > virtual address d38c2010
> > > Sep 29 15:05:01 zen kernel: c0145032
> > > Sep 29 15:05:01 zen kernel: *pde = 01594063
> > > Sep 29 15:05:01 zen kernel: Oops: 0002
> > > Sep 29 15:05:01 zen kernel: CPU: 0
> > > Sep 29 15:05:01 zen kernel: EIP: 0010:[proc_get_inode+150/264]
> > > Sep 29 15:05:01 zen kernel: EFLAGS: 00010286
> > > Sep 29 15:05:01 zen kernel: eax: d38c2000 ebx: cf687520 ecx: c1466fc0 edx:
>00000023
> > > Sep 29 15:05:01 zen kernel: esi: cd990340 edi: cf687574 ebp: ffffffea esp:
>cd99def0
> > > Sep 29 15:05:01 zen kernel: ds: 0018 es: 0018 ss: 0018
> > > Sep 29 15:05:01 zen kernel: Process ls (pid: 705, stackpage=cd99d000)
> > > Sep 29 15:05:01 zen kernel: Stack: cda17ec0 cd9901c0 c0146a9c c144b800 00001190
>cf687520 fffffff4 cda17ec0
> > > Sep 29 15:05:01 zen kernel: cd9901c0 cd990214 c0137083 cd9901c0 cda17ec0
>00000000 00000000 cd99df68
> > > Sep 29 15:05:01 zen kernel: cf115013 c01377c4 cda17e40 cd99df68 00000000
>cf115000 00000000 cd99dfa4
> > > Sep 29 15:05:01 zen kernel: Call Trace: [proc_lookup+104/140]
>[real_lookup+79/184] [path_walk+1440/2028] [<f27a545f>] [__user_walk+58/84]
>[sys_newlstat+21/112] [system_call+51/64]
> > > Sep 29 15:05:01 zen kernel: Code: ff 40 10 8b 43 24 80 48 14 18 66 8b 43 08 25
>00 f0 ff ff 66
> > > Using defaults from ksymoops -t elf32-i386 -a i386
> > >
> > > Code; 00000000 Before first symbol
> > > 00000000 <_EIP>:
> > > Code; 00000000 Before first symbol
> > > 0: ff 40 10 incl 0x10(%eax)
> > > Code; 00000003 Before first symbol
> > > 3: 8b 43 24 movl 0x24(%ebx),%eax
> > > Code; 00000006 Before first symbol
> > > 6: 80 48 14 18 orb $0x18,0x14(%eax)
> > > Code; 0000000a Before first symbol
> > > a: 66 8b 43 08 movw 0x8(%ebx),%ax
> > > Code; 0000000e Before first symbol
> > > e: 25 00 f0 ff ff andl $0xfffff000,%eax
> > > Code; 00000013 Before first symbol
> > > 13: 66 00 00 addb %al,(%eax)
> > >
--
Matthew Dharm Home: [EMAIL PROTECTED]
Maintainer, Linux USB Mass Storage Driver
P: How about "Web Designer"?
DP: I'd like a name that people won't laugh at.
-- Pitr and Dust Puppy
User Friendly, 12/6/1997
PGP signature
