2.5.15 IDE 62
Quoted 8 lines · Martin Dalecki
Oops. Indeed I see now that the
ide_lockis exported to the upper layers above it inide-probe.c
blk_init_queue(q, do_ide_request, &ide_lock);But this is problematic in itself, since it means that we are basically serialiazing between *all* requests on all channels.
Surely not. If you look at the line above the one you quoted above, you see the per-channel serialization being requested:
q->queuedata = drive->channel;
cheers,
Neil
PS: 2.4 doesn't even have the spinlock as a parameter to blk_init_queue().
Quoted 2 lines · Neil Conway
PS: 2.4 doesn't even have the spinlock as a parameter to
blk_init_queue().
Right, the "queue" (well global in 2.4) lock has always been grabbed prior to request_fn entry.
Signature · 1 line
Jens Axboe