rare, but serious deadlock in rpm
Bug #910731 reported by
Jeff Johnson
This bug affects 1 person
Affects | Status | Importance | Assigned to | Milestone | |
---|---|---|---|---|---|
RPM |
Invalid
|
Medium
|
Jeff Johnson | ||
Mandriva |
Incomplete
|
Medium
|
Bug Description
this is hard to reproduce, but it can happen, if you kill (have to kill) a
running rpm instance. Since those instances of rpm that "hang" can only be
killed by SIGKILL, the process cannot guarantee a fault free rpm-db.
The error lies in the fact, that on startup of rpm it does some mmap's of __db*
and if one of them is locked, it stays there forever - there is no simple tool
to say: clear your lock.
I had to edit __db.002 and set the value at adress 0xf18 to 0 to get rpm
running again.
Since this has happend the 3 time in different situations, I'd suggest another
soulution than just trust this byte if the rpm-db is locked.
[reply] [-] Comment 1
tags: | added: mandriva rpmdb |
Changed in mandriva: | |
importance: | Unknown → Medium |
status: | Unknown → Incomplete |
Changed in rpm: | |
importance: | Undecided → Medium |
To post a comment you must log in.
RPM registers all locker processes/threads and checks/clears
stale locks on next invocation.