Memory tables corruption - kernel v2.4.33-rc1
Please fwd to 2.4.33 Team.
One of my customers reported in tests that, after some runtime, his 59MB ram drive suddenly, randomly shrinks.
Same customer s/w has been running fine with 2.4.31 kernel.
Report follows:
I have two Advantech (i486 Geode's w/64MB ram, ~49MB usable, as configured in the kernel make) test units where Dosemu crashed this weekend. For some reason the Via unit was fine.
Here is what happened:
1. Linux was O.K., dosemu 1.3.2 had shut down.
It appears both units were out of RAM.
2. The first unit showed the RAM drive size to be 14661 1k blocks using the df command. The 1K blocks should be 47595, as they were after boot. The other unit showed the correct value of 1K blocks=47595. Even after restarting the Advantech, the 1K blocks would not go back to normal, so my software cannot even load.
3. The second Advantech unit showed correct values with the df command, but when I used the du -cbs command, I found the /tmp directory had only 110796 bytes, instead of the normal ~15 MB. When I restarted Linux, all the RAM came back, and the unit started working again.
---------
Thanks, I will check further with this customer.
Tony Borras
Fall is my favorite season in Los Angeles, watching the birds change color and fall from the trees.
David Letterman (1947 - )
--
__ __ _ I N C. http://www.sysdev.org
/ __|\\// __|| \ __ __ / [email protected] \__ \ \/\__ \||)|/ O_)\/ / \/ System Tools / Utilities |___/ || ___/|_ /\___|\_/ WIntel / Linux Device Drivers
Hello,
Quoted 32 lines · Tony Borras
Please fwd to 2.4.33 Team.
One of my customers reported in tests that, after some runtime, his 59MB ram drive suddenly, randomly shrinks.
Same customer s/w has been running fine with 2.4.31 kernel.
Report follows:
I have two Advantech (i486 Geode's w/64MB ram, ~49MB usable, as configured in the kernel make) test units where Dosemu crashed this weekend. For some reason the Via unit was fine.
Here is what happened:
1. Linux was O.K., dosemu 1.3.2 had shut down.
It appears both units were out of RAM.2. The first unit showed the RAM drive size to be 14661 1k blocks using the df command. The 1K blocks should be 47595, as they were after boot. The other unit showed the correct value of 1K blocks=47595. Even after restarting the Advantech, the 1K blocks would not go back to normal, so my software cannot even load.
3. The second Advantech unit showed correct values with the df command, but when I used the du -cbs command, I found the /tmp directory had only 110796 bytes, instead of the normal ~15 MB. When I restarted Linux, all the RAM came back, and the unit started working again.
Quoted 3 lines
---------Thanks, I will check further with this customer.
2.4.33-rc1 has a bug in vfs_unlink(). Probably you've been hitting it when removing lots of files from /tmp which might not have been really freed. -rc2 has this fixed. You should try it instead. Other people reported kernel panics on -rc1 with NFS.
Quoted 1 line
Tony Borras
Regards,
Willy