Index: share/man/man9/vnode_pager_setsize.9 =================================================================== --- share/man/man9/vnode_pager_setsize.9 +++ share/man/man9/vnode_pager_setsize.9 @@ -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 @@ -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