Hold the reference count until the CCB is released
When a disk disappears and the periph is invalidated, any I/Os that are pending with the controller can cause a crash when they complete. Move to holding the softc reference count taken in dastart() until the I/O is complete rather than only until xpt_action() returns. (This approach was suggested by Ken Merry.) Sponsored by: Netflix Submitted by: Chuck Silvers Differential Revision: https://reviews.freebsd.org/D15435
This commit is contained in:
parent
646fd30caf
commit
0eedd21317
Notes:
svn2git
2020-12-20 02:59:44 +00:00
svn path=/head/; revision=333648
@ -3300,7 +3300,6 @@ dastart(struct cam_periph *periph, union ccb *start_ccb)
|
||||
cam_periph_unlock(periph);
|
||||
xpt_action(start_ccb);
|
||||
cam_periph_lock(periph);
|
||||
softc->refcount--;
|
||||
|
||||
/* May have more work to do, so ensure we stay scheduled */
|
||||
daschedule(periph);
|
||||
@ -4443,13 +4442,15 @@ dadone(struct cam_periph *periph, union ccb *done_ccb)
|
||||
softc->flags |= DA_FLAG_WAS_OTAG;
|
||||
|
||||
/*
|
||||
* We need to call cam_iosched before we call biodone so that we
|
||||
* don't measure any activity that happens in the completion
|
||||
* routine, which in the case of sendfile can be quite
|
||||
* extensive.
|
||||
* We need to call cam_iosched before we call biodone so that we don't
|
||||
* measure any activity that happens in the completion routine, which in
|
||||
* the case of sendfile can be quite extensive. Release the periph
|
||||
* refcount taken in dastart() for each CCB.
|
||||
*/
|
||||
cam_iosched_bio_complete(softc->cam_iosched, bp, done_ccb);
|
||||
xpt_release_ccb(done_ccb);
|
||||
KASSERT(softc->refcount >= 1, ("dadone softc %p refcount %d", softc, softc->refcount));
|
||||
softc->refcount--;
|
||||
if (state == DA_CCB_DELETE) {
|
||||
TAILQ_HEAD(, bio) queue;
|
||||
|
||||
|
Loading…
Reference in New Issue
Block a user