Skip to content

Commit

Permalink
vnode_pager_setsize.9: Some clarifications on the manpage
Browse files Browse the repository at this point in the history
A number of changes:
- Clarifies the locking rules when calling the routine.
- Correct the description regarding the content range to be purged.
- Document the effects on page fault handler.

MFC with:	86a52e2
Sponsored by:	The FreeBSD Foundation
Reviewed by:	bcr, kib
Approved by:	philip (mentor)
Differential Revision:	https://reviews.freebsd.org/D29637

(cherry picked from commit b77f5f5)
  • Loading branch information
khng300 committed Apr 27, 2021
1 parent 68a43a5 commit 17db2ae
Showing 1 changed file with 23 additions and 18 deletions.
41 changes: 23 additions & 18 deletions share/man/man9/vnode_pager_setsize.9
Original file line number Diff line number Diff line change
Expand Up @@ -27,7 +27,7 @@
.\" OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
.\" SUCH DAMAGE.
.\"
.Dd April 6, 2021
.Dd April 8, 2021
.Dt VNODE_PAGER_SETSIZE 9
.Os
.Sh NAME
Expand All @@ -41,26 +41,31 @@
.Fn vnode_pager_setsize "struct vnode *vp" "vm_ooffset_t nsize"
.Sh DESCRIPTION
.Nm
lets the VM system know about a change in size for a file.
Content beyond the new EOF specified by the
.Fa nsize
argument will be purged from the cache.
This function is useful for use within file system code to implement
truncation in
.Xr VOP_SETATTR 9 .
.Sh IMPLEMENTATION NOTES
lets the VM system know about a change in size for a file,
and updates the object size and vnode pager size of the vm object in
.Fa vp
with
.Fa nsize .
Page faults on the object mapping with offset beyond the new object
size results in
.Va SIGBUS .
.Pp
Pages between the old EOF and the new EOF are removed from the object queue
if the file size shrinks.
In case the new EOF specified by the
.Fa nsize
argument is not aligned to page boundaries,
partial-page area starting beyond the EOF will be zeroed.
In partial-page area,
for content occupying whole blocks within block
boundaries,
the dirty bits for the corresponding blocks will be cleared.
.Sh LOCKING
Writer lock of the VM object of
argument is not aligned to page boundary,
partial-page area starting beyond the EOF is zeroed and marked invalid.
if the page exists resident.
.Pp
In case the vnode
.Fa vp
will be held within the function.
does not have a VM object allocated, the effect of calling this function is no-op.
.Pp
This function must be used within file system code to implement truncation
if the file system allocates vm objects for vnodes.
.Sh LOCKS
The vnode should be exclusively locked on entry and will still be locked on exit.
.Sh SEE ALSO
.Xr vnode 9
.Sh HISTORY
Expand Down

0 comments on commit 17db2ae

Please sign in to comment.