d/freebsd-dev
1
0
Fork 0
Code Issues Pull Requests Packages Projects Releases Wiki Activity

Remove an overly-aggressive assertion.

Browse Source 
While it is true that the new vmspace passed to vmspace_switch_aio
will always have a valid reference due to the AIO job or the extra
reference on the original vmspace in the worker thread, it is not true
that the old vmspace being switched away from will have more than one
reference.

Specifically, when a process with queued AIO jobs exits, the exit hook
in aio_proc_rundown will only ensure that all of the AIO jobs have
completed or been cancelled.  However, the last AIO job might have
completed and woken up the exiting process before the worker thread
servicing that job has switched back to its original vmspace.  In that
case, the process might finish exiting dropping its reference to the
vmspace before the worker thread resulting in the worker thread
dropping the last reference.

Reported by:	np
Reviewed by:	alc, markj, np, imp
MFC after:	2 weeks
Sponsored by:	Chelsio Communications
Differential Revision:	https://reviews.freebsd.org/D20542
This commit is contained in:
John Baldwin 2019-06-10 19:01:54 +00:00
parent 2dd9a967d2
commit 0b96ca3310
Notes: svn2git 2020-12-20 02:59:44 +00:00 
svn path=/head/; revision=348874
1 changed files with 11 additions and 15 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

26
sys/vm/vm_map.c
View File

@ -454,18 +454,17 @@ vmspace_acquire_ref(struct proc *p)
/*
* Switch between vmspaces in an AIO kernel process.
*
* The AIO kernel processes switch to and from a user process's
* vmspace while performing an I/O operation on behalf of a user
* process. The new vmspace is either the vmspace of a user process
* obtained from an active AIO request or the initial vmspace of the
* AIO kernel process (when it is idling). Because user processes
* will block to drain any active AIO requests before proceeding in
* exit() or execve(), the vmspace reference count for these vmspaces
* can never be 0. This allows for a much simpler implementation than
* the loop in vmspace_acquire_ref() above. Similarly, AIO kernel
* processes hold an extra reference on their initial vmspace for the
* life of the process so that this guarantee is true for any vmspace
* passed as 'newvm'.
* The new vmspace is either the vmspace of a user process obtained
* from an active AIO request or the initial vmspace of the AIO kernel
* process (when it is idling). Because user processes will block to
* drain any active AIO requests before proceeding in exit() or
* execve(), the reference count for vmspaces from AIO requests can
* never be 0. Similarly, AIO kernel processes hold an extra
* reference on their initial vmspace for the life of the process. As
* a result, the 'newvm' vmspace always has a non-zero reference
* count. This permits an additional reference on 'newvm' to be
* acquired via a simple atomic increment rather than the loop in
* vmspace_acquire_ref() above.
*/
void
vmspace_switch_aio(struct vmspace *newvm)
@ -490,9 +489,6 @@ vmspace_switch_aio(struct vmspace *newvm)
/* Activate the new mapping. */
pmap_activate(curthread);
/* Remove the daemon's reference to the old address space. */
KASSERT(oldvm->vm_refcnt > 1,
("vmspace_switch_aio: oldvm dropping last reference"));
vmspace_free(oldvm);
}