Opened 5 years ago
Last modified 2 years ago
#14762 new bug
KDL with bfs checkfs on x86_gcc2
| Reported by: | korli | Owned by: | axeld |
|---|---|---|---|
| Priority: | normal | Milestone: | Unscheduled |
| Component: | File Systems/BFS | Version: | R1/Development |
| Keywords: | Cc: | ||
| Blocked By: | Blocking: | #17718 | |
| Platform: | x86 |
Description
it's a 50GB partition. Tried two times on x86_gcc2 with the same KDL. Works successfully on x86_64. Probably the number of blocks is too high for 32-bit, but it should be possible to handle the problem with more checks.
Attachments (3)
Change History (10)
by , 5 years ago
| Attachment: | SAM_6814.JPG added |
|---|
comment:1 by , 5 years ago
IIRC this is actually an out-of-address-space issue. I know bfs_recover definitely runs into a similar one on 32-bit systems.
by , 5 years ago
| Attachment: | checkfs_kdl.png added |
|---|
comment:4 by , 5 years ago
Ah, didn't check syslog and the next attempt at running checkfs succeeded with no KDL.
comment:5 by , 2 years ago
| Blocking: | 17718 added |
|---|
by , 2 years ago
| Attachment: | syslog.old added |
|---|
comment:6 by , 2 years ago
Does running with PAE improve things? In x86 virtual machines running in 32 bit mode, with PAE off you get 32 bit virtual addresses, with PAE on, you get 64 bit virtual addresses. The extra space avoids some problems (like a VirtualBox crash on the file cache using up all memory, note it's a VM crash, not Haiku!) and may help with things like using a swap file.
Though that crash sounds more like corrupted data blowing up the disk check algorithm. You'd think a 50GB disk wouldn't need that much RAM to check.
KDL screenshot