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

In vm_fault()'s loop over the shadow chain, move a comment describing our

Browse Source 
invariants to a better place.  Also, add two comments concerning the
relationship between the map and vnode locks.

Reviewed by:	kib
MFC after:	3 days
This commit is contained in:
Alan Cox 2016-11-03 16:44:55 +00:00
parent 2ad1d09f16
commit 857025056f
1 changed files with 18 additions and 5 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

23
sys/vm/vm_fault.c
View File

@ -563,6 +563,14 @@ fast_failed:
}
readrest:
/*
* At this point, we have either allocated a new page or found
* an existing page that is only partially valid.
*
* We hold a reference on the current object and the page is
* exclusive busied.
*/
/*
* If the pager for the current object might have the page,
* then determine the number of additional pages to read and
@ -632,18 +640,23 @@ readrest:
*/
if (fs.object->type != OBJT_DEFAULT) {
/*
* We have either allocated a new page or found an
* existing page that is only partially valid. We
* hold a reference on fs.object and the page is
* exclusive busied.
* Release the map lock before locking the vnode or
* sleeping in the pager. (If the current object has
* a shadow, then an earlier iteration of this loop
* may have already unlocked the map.)
*/
unlock_map(&fs);
if (fs.object->type == OBJT_VNODE &&
(vp = fs.object->handle) != fs.vp) {
/*
* Perform an unlock in case the desired vnode
* changed while the map was unlocked during a
* retry.
*/
unlock_vp(&fs);
locked = VOP_ISLOCKED(vp);
locked = VOP_ISLOCKED(vp);
if (locked != LK_EXCLUSIVE)
locked = LK_SHARED;