I configured, built and booted Linux-2.4.22. There are
some problems.
(1) `dmesg` fails to read the first part of the buffered
kernel log. I have attached two files, dmesg-20 (normal)
and dmesg-22 (bad). File dmesg-22 is from Linux-2.4.22
and dmesg-20 is from Linux-2.4.20. To save space, I
snip everything after 'NET4'.
(2) The ipx module fails to load with undefined symbols.
This module loads fine in Linux-2.4.20.
depmod: *** Unresolved symbols in /lib/modules/2.4.22/kernel/net/ipx/ipx.o
depmod: unregister_8022_client
depmod: make_EII_client
depmod: register_8022_client
depmod: register_snap_client
depmod: make_8023_client
depmod: destroy_8023_client
depmod: destroy_EII_client
depmod: unregister_snap_client
(3) When umounting the root file-system, the machine usually
hangs. The result is a long `fsck` on the next boot. The problem
seems to be that sendmail doesn't get killed during the `init 0`
sequence. It remains with a file open and the root file-system isn't
unmounted. A temporary work-round is to `ifconfig eth0 down` before
starting shutdown. Otherwise, sendmail remains stuck in the 'D' state.
(4) When mounting the DOS file-systems during startup, the echo
on the screen shows about 15 lines of white-space. This never
happened before. When mounting /proc, there are 6 lines of
white-space, also strange.
#
# /etc/fstab
#
# device directory type options freq pass
# -=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-
/dev/sdb1 / ext2 rw,noatime 0 1
/dev/sdc1 /alt ext2 rw,noatime 0 2
/dev/sdb2 none swap defaults 0 2
/dev/sdc2 none swap defaults 0 2
/dev/sdc3 /home/users ext2 rw,noatime 0 2
none /proc proc defaults 0 2
/dev/sda1 /dos/drive_C msdos defaults 0 2
/dev/sda5 /dos/drive_D msdos defaults 0 2
Otherwise it works. I am using it now so I can test any possible
fixes.
Cheers,
Dick Johnson
Penguin : Linux version 2.4.22 on an i686 machine (797.90 BogoMips).
Note 96.31% of all statistics are fiction.
On Tue, 26 Aug 2003, Richard B. Johnson wrote:
>
> I configured, built and booted Linux-2.4.22. There are
> some problems.
>
> (1) `dmesg` fails to read the first part of the buffered
> kernel log. I have attached two files, dmesg-20 (normal)
sounds like the log buffer wrapped around from a lot of printks
> (3) When umounting the root file-system, the machine usually
> hangs. The result is a long `fsck` on the next boot. The problem
> seems to be that sendmail doesn't get killed during the `init 0`
> sequence. It remains with a file open and the root file-system isn't
> unmounted. A temporary work-round is to `ifconfig eth0 down` before
> starting shutdown. Otherwise, sendmail remains stuck in the 'D' state.
this is likely the unshare_files change, which has been mentioned in
several threads as causing similar type issues...i posted a patch that
solves the issue, and Alan included a patch in his -ac series, which
also addresses this issue.
-Jason
On Wed, 27 Aug 2003, Jason Baron wrote:
>
> On Tue, 26 Aug 2003, Richard B. Johnson wrote:
>
> >
> > I configured, built and booted Linux-2.4.22. There are
> > some problems.
> >
> > (1) `dmesg` fails to read the first part of the buffered
> > kernel log. I have attached two files, dmesg-20 (normal)
>
> sounds like the log buffer wrapped around from a lot of printks
>
Perhaps the buffer size was changed?? I'll look at the printks()
I think there are the same number as before.
> > (3) When umounting the root file-system, the machine usually
> > hangs. The result is a long `fsck` on the next boot. The problem
> > seems to be that sendmail doesn't get killed during the `init 0`
> > sequence. It remains with a file open and the root file-system isn't
> > unmounted. A temporary work-round is to `ifconfig eth0 down` before
> > starting shutdown. Otherwise, sendmail remains stuck in the 'D' state.
>
> this is likely the unshare_files change, which has been mentioned in
> several threads as causing similar type issues...i posted a patch that
> solves the issue, and Alan included a patch in his -ac series, which
> also addresses this issue.
>
> -Jason
>
Thanks, I'll check into it.
Cheers,
Dick Johnson
Penguin : Linux version 2.4.22 on an i686 machine (794.73 BogoMips).
Note 96.31% of all statistics are fiction.
On Wed, 27 Aug 2003, Richard B. Johnson wrote:
> On Wed, 27 Aug 2003, Jason Baron wrote:
>
> >
> > On Tue, 26 Aug 2003, Richard B. Johnson wrote:
> >
> > >
> > > I configured, built and booted Linux-2.4.22. There are
> > > some problems.
> > >
> > > (1) `dmesg` fails to read the first part of the buffered
> > > kernel log. I have attached two files, dmesg-20 (normal)
> >
> > sounds like the log buffer wrapped around from a lot of printks
> >
>
I looked into this. The problem I am seeing is not in printk().
It may, therefore, be in klogd.
I replaced the linux-2.4.22 printk() with the one that worked from
2.4.20. The problem presists!
The log-file, when booting the SMP machine is 7,653 bytes when
using 2.4.20. With linux-2.4.22, the first 33 bytes are missing.
Cheers,
Dick Johnson
Penguin : Linux version 2.4.22 on an i686 machine (794.73 BogoMips).
Note 96.31% of all statistics are fiction.
On Tue, 26 Aug 2003, Richard B. Johnson wrote:
>
> I configured, built and booted Linux-2.4.22. There are
> some problems.
>
> (1) `dmesg` fails to read the first part of the buffered
> kernel log. I have attached two files, dmesg-20 (normal)
> and dmesg-22 (bad). File dmesg-22 is from Linux-2.4.22
> and dmesg-20 is from Linux-2.4.20. To save space, I
> snip everything after 'NET4'.
>
> (2) The ipx module fails to load with undefined symbols.
> This module loads fine in Linux-2.4.20.
>
> depmod: *** Unresolved symbols in /lib/modules/2.4.22/kernel/net/ipx/ipx.o
> depmod: unregister_8022_client
> depmod: make_EII_client
> depmod: register_8022_client
> depmod: register_snap_client
> depmod: make_8023_client
> depmod: destroy_8023_client
> depmod: destroy_EII_client
> depmod: unregister_snap_client
> (3) When umounting the root file-system, the machine usually
> hangs. The result is a long `fsck` on the next boot. The problem
> seems to be that sendmail doesn't get killed during the `init 0`
> sequence. It remains with a file open and the root file-system isn't
> unmounted. A temporary work-round is to `ifconfig eth0 down` before
> starting shutdown. Otherwise, sendmail remains stuck in the 'D' state.
Which previous kernel didnt show this behaviour?
>
> (4) When mounting the DOS file-systems during startup, the echo
> on the screen shows about 15 lines of white-space. This never
> happened before. When mounting /proc, there are 6 lines of
> white-space, also strange.
Again, which kernel doesnt show that behaviour?
Thanks
On Mon, 1 Sep 2003, Marcelo Tosatti wrote:
>
>
> On Tue, 26 Aug 2003, Richard B. Johnson wrote:
>
> >
> > I configured, built and booted Linux-2.4.22. There are
> > some problems.
> >
> > (1) `dmesg` fails to read the first part of the buffered
> > kernel log. I have attached two files, dmesg-20 (normal)
> > and dmesg-22 (bad). File dmesg-22 is from Linux-2.4.22
> > and dmesg-20 is from Linux-2.4.20. To save space, I
> > snip everything after 'NET4'.
> >
> > (2) The ipx module fails to load with undefined symbols.
> > This module loads fine in Linux-2.4.20.
> >
> > depmod: *** Unresolved symbols in /lib/modules/2.4.22/kernel/net/ipx/ipx.o
> > depmod: unregister_8022_client
> > depmod: make_EII_client
> > depmod: register_8022_client
> > depmod: register_snap_client
> > depmod: make_8023_client
> > depmod: destroy_8023_client
> > depmod: destroy_EII_client
> > depmod: unregister_snap_client
>
> > (3) When umounting the root file-system, the machine usually
> > hangs. The result is a long `fsck` on the next boot. The problem
> > seems to be that sendmail doesn't get killed during the `init 0`
> > sequence. It remains with a file open and the root file-system isn't
> > unmounted. A temporary work-round is to `ifconfig eth0 down` before
> > starting shutdown. Otherwise, sendmail remains stuck in the 'D' state.
>
> Which previous kernel didnt show this behaviour?
>
> >
> > (4) When mounting the DOS file-systems during startup, the echo
> > on the screen shows about 15 lines of white-space. This never
> > happened before. When mounting /proc, there are 6 lines of
> > white-space, also strange.
>
> Again, which kernel doesnt show that behaviour?
Oh, Jason already answered. The fix he talks about is already in
2.4.23-pre.
On Mon, 1 Sep 2003, Marcelo Tosatti wrote:
>
>
> On Tue, 26 Aug 2003, Richard B. Johnson wrote:
>
> >
> > I configured, built and booted Linux-2.4.22. There are
> > some problems.
> >
> > (1) `dmesg` fails to read the first part of the buffered
> > kernel log. I have attached two files, dmesg-20 (normal)
> > and dmesg-22 (bad). File dmesg-22 is from Linux-2.4.22
> > and dmesg-20 is from Linux-2.4.20. To save space, I
> > snip everything after 'NET4'.
> >
> > (2) The ipx module fails to load with undefined symbols.
> > This module loads fine in Linux-2.4.20.
> >
> > depmod: *** Unresolved symbols in /lib/modules/2.4.22/kernel/net/ipx/ipx.o
> > depmod: unregister_8022_client
> > depmod: make_EII_client
> > depmod: register_8022_client
> > depmod: register_snap_client
> > depmod: make_8023_client
> > depmod: destroy_8023_client
> > depmod: destroy_EII_client
> > depmod: unregister_snap_client
>
> > (3) When umounting the root file-system, the machine usually
> > hangs. The result is a long `fsck` on the next boot. The problem
> > seems to be that sendmail doesn't get killed during the `init 0`
> > sequence. It remains with a file open and the root file-system isn't
> > unmounted. A temporary work-round is to `ifconfig eth0 down` before
> > starting shutdown. Otherwise, sendmail remains stuck in the 'D' state.
>
> Which previous kernel didnt show this behaviour?
>
Previous kernel was linux-2.4.20. Linux-2.4.21 would never complete
a boot on my system so I gave up on that one.
> >
> > (4) When mounting the DOS file-systems during startup, the echo
> > on the screen shows about 15 lines of white-space. This never
> > happened before. When mounting /proc, there are 6 lines of
> > white-space, also strange.
>
> Again, which kernel doesnt show that behaviour?
>
Also, I replaced linux-2.4.22/kernel/prink.c with the one from
linux-2.4.21 and it did not help the missing boot data problem.
I also changed the buffer size to 131,072 and it didn't help
either so the problem is not in printk.c, but somewhere else,
perhaps klogd.
> Thanks
>
Cheers,
Dick Johnson
Penguin : Linux version 2.4.22 on an i686 machine (794.73 BogoMips).
Note 96.31% of all statistics are fiction.