Opened 15 years ago
Closed 15 years ago
#5404 closed bug (fixed)
Invalid concurrent page access
Reported by: | andreasf | Owned by: | bonefish |
---|---|---|---|
Priority: | high | Milestone: | R1 |
Component: | - General | Version: | R1/Development |
Keywords: | Cc: | Jens.Arm@… | |
Blocked By: | Blocking: | #5409 | |
Platform: | x86 |
Description
Got this at hrev35458 while running git-rebase
, while it was re-applying the local commits:
PANIC: Invalid concurrent access to page 0x82c03ddc (start), currently accessed by: 7 Welcome to Kernel Debugging Land... Thread 22630 "git" running on CPU 1 kdebug>
The system is frozen, so KDL cannot be used to get more info.
It looks similar to my earlier #5359, but there have been no further problems at boot time.
Change History (7)
follow-up: 5 comment:1 by , 15 years ago
Cc: | removed |
---|---|
Owner: | changed from | to
Status: | new → assigned |
comment:3 by , 15 years ago
Cc: | added |
---|
comment:4 by , 15 years ago
Blocking: | 5409 removed |
---|---|
Cc: | removed |
Status: | assigned → in-progress |
comment:5 by , 15 years ago
Replying to bonefish:
Do you know by any chance whether git uses mmap() for I/O?
Yes. There's a Makefile variable NO_MMAP
but I don't set it, so it will use mmap
at some point.
comment:6 by , 15 years ago
Blocking: | 5409 added |
---|---|
Cc: | added |
Replying to andreasf:
Yes. There's a Makefile variable NO_MMAP but I don't set it, so it will use mmap at some point.
Thanks, that supports my suspicion that this is indeed the same problem as #5409 where an mmap()ed area is deleted while the page writer writes back one of its modified pages.
@jahaiku: Sorry for removing you from the CC. Trac is really annoying w.r.t. concurrent ticket changes.
Assuming that thread 7 is the page writer, that might be a problem Axel pointed me to recently. Do you know by any chance whether git uses mmap() for I/O?
Regarding the hanging KDL, in case you use a USB keyboard, you need to enter KDL manually at least once or it won't work on programmatic panic()s.