Hi,
I've been having an issue with blktrace -k, basically I could only run
it once per boot. Am I the only one running into this issue?
This may be due to some local modification I've done, but I took a
look at the code and there seems to be two issues with the current
code:
# TEARDOWN:
> int blk_trace_remove(struct request_queue *q)
> {
> struct blk_trace *bt;
>
> bt = xchg(&q->blk_trace, NULL);
> if (!bt)
> return -EINVAL;
>
> if (bt->trace_state == Blktrace_setup ||
> bt->trace_state == Blktrace_stopped)
> blk_trace_cleanup(bt);
When this "if" fails then we have lost the only reference to
q->blk_trace?
>
> return 0;
> }
The other issue seems to be that blk_trace_cleanup will fail when if
we run blk_trace_cleanup when user space has the traceN files
open:
> static void blk_trace_cleanup(struct blk_trace *bt)
> {
> relay_close(bt->rchan);
This will only decrease the refcount of the files by one, but user
space still holds one reference so blk_remove_buf_file_callback will
not be called.
> debugfs_remove(bt->dropped_file);
This will run OK. (I'm not sure if blktrace user space has read this
file yet?)
> blk_remove_tree(bt->dir);
This will fail because dir is not empty.
> free_percpu(bt->sequence);
> free_percpu(bt->msg_data);
> kfree(bt);
> }
Workaround: use "killall blktrace" or similar?
If time permits I'll write a patch later this week.
--
Carl Henrik
--
To unsubscribe from this list: send the line "unsubscribe linux-btrace" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
