Opened 15 years ago
Closed 15 years ago
#175 closed defect (fixed)
Kernel hangs in ipc_cleanup()
| Reported by: | Jakub Jermář | Owned by: | Jakub Jermář |
|---|---|---|---|
| Priority: | major | Milestone: | 0.4.2 |
| Component: | helenos/kernel/generic | Version: | mainline |
| Keywords: | Cc: | ||
| Blocker for: | Depends on: | ||
| See also: |
Description
The hang described in ticket #169 is actually caused by kernel looping on the answerbox spinlock in ipc_cleanup().
The klog read from gdb attached to qemu says:
...A...s...s...e...r...t...i...o...n... ...f...a...i...l...e...d... ...(...n...o...d...e...p...-...>...s...i...z...e... ...%... ...b...p...s... ...=...=... ...0...)... ...a...t... ...f...i...l...e... ...'...f...a...t..._...o...p...s.......c...'...,... ...l...i...n...e... ...1...2...0...7....... ...0...x...3...0...f...4...4...:... ...0...x...1...3...e...1...a...(...)... ...0...x...3...0...f...b...4...:... ...0...x...3...2...9...2...(...)... ...0...x...3...0...f...f...4...:... ...0...x...1...2...f...1...(...)... ...0...x...3...1...0...2...4...:... ...0...x...1...0...d...7...b...(...)... ...0...x...3...1...0...4...4...:... ...0...x...b...5...f...7...(...)... ...c...p...u...0...:... ...l...o...o...p...i...n...g... ...o...n... ...s...p...i...n...l...o...c...k... ...8...0...0...2...d...0...5...8...:...i...p...c..._...b...o...x..._...l...o...c...k...,... ...c...a...l...l...e...r...=...8...0...1...3...1...f...c...1...(...g...e...n...e...r...i...c.../...s...r...c.../...i...p...c.../...i...p...c.......o...:...i...p...c..._...c...l...e...a...n...u...p...)...
Note:
See TracTickets
for help on using tickets.
The problem seems to be related to homogenous mounts (e.g. FAT on FAT) or possibly other situations when the task sends itself a call. In ipc_cleanup(), it locks TASK→answerbox and
attempts to answer all unanswered calls. For calls that the task sent itself, this will
hang the system, because _ipc_answer_free_call(), which is used during this operation, locks the
callerbox answerbox, which is unfortunately the same as the already locked TASK→answerbox.