| Message ID | 20220428124852.80682-1-xiubli@redhat.com (mailing list archive) |
|---|---|
| State | New, archived |
| Headers | show |
| Series | ceph: try to queue a writeback if revoking fails | expand |
Please ignore this, just forgot to add the v2 tag. -- Xiubo On 4/28/22 8:48 PM, Xiubo Li wrote: > If the pagecaches writeback just finished and the i_wrbuffer_ref > reaches zero it will try to trigger ceph_check_caps(). But if just > before ceph_check_caps() the i_wrbuffer_ref could be increased > again by mmap/cache write, then the Fwb revoke will fail. > > We need to try to queue a writeback in this case instead of > triggering the writeback by BDI's delayed work per 5 seconds. > > URL: https://tracker.ceph.com/issues/55377 > URL: https://tracker.ceph.com/issues/46904 > Signed-off-by: Xiubo Li <xiubli@redhat.com> > --- > fs/ceph/caps.c | 28 ++++++++++++++++++++++++---- > 1 file changed, 24 insertions(+), 4 deletions(-) > > diff --git a/fs/ceph/caps.c b/fs/ceph/caps.c > index 906c95d2a4ed..22dae29be64d 100644 > --- a/fs/ceph/caps.c > +++ b/fs/ceph/caps.c > @@ -1912,6 +1912,7 @@ void ceph_check_caps(struct ceph_inode_info *ci, int flags, > struct rb_node *p; > bool queue_invalidate = false; > bool tried_invalidate = false; > + bool queue_writeback = false; > > if (session) > ceph_get_mds_session(session); > @@ -2064,10 +2065,27 @@ void ceph_check_caps(struct ceph_inode_info *ci, int flags, > } > > /* completed revocation? going down and there are no caps? */ > - if (revoking && (revoking & cap_used) == 0) { > - dout("completed revocation of %s\n", > - ceph_cap_string(cap->implemented & ~cap->issued)); > - goto ack; > + if (revoking) { > + if ((revoking & cap_used) == 0) { > + dout("completed revocation of %s\n", > + ceph_cap_string(cap->implemented & ~cap->issued)); > + goto ack; > + } > + > + /* > + * If the "i_wrbuffer_ref" was increased by mmap or generic > + * cache write just before the ceph_check_caps() is called, > + * the Fb capability revoking will fail this time. Then we > + * must wait for the BDI's delayed work to flush the dirty > + * pages and to release the "i_wrbuffer_ref", which will cost > + * at most 5 seconds. That means the MDS needs to wait at > + * most 5 seconds to finished the Fb capability's revocation. > + * > + * Let's queue a writeback for it. > + */ > + if (S_ISREG(inode->i_mode) && ci->i_wrbuffer_ref && > + (revoking & CEPH_CAP_FILE_BUFFER)) > + queue_writeback = true; > } > > /* want more caps from mds? */ > @@ -2137,6 +2155,8 @@ void ceph_check_caps(struct ceph_inode_info *ci, int flags, > spin_unlock(&ci->i_ceph_lock); > > ceph_put_mds_session(session); > + if (queue_writeback) > + ceph_queue_writeback(inode); > if (queue_invalidate) > ceph_queue_invalidate(inode); > }
diff --git a/fs/ceph/caps.c b/fs/ceph/caps.c index 906c95d2a4ed..22dae29be64d 100644 --- a/fs/ceph/caps.c +++ b/fs/ceph/caps.c @@ -1912,6 +1912,7 @@ void ceph_check_caps(struct ceph_inode_info *ci, int flags, struct rb_node *p; bool queue_invalidate = false; bool tried_invalidate = false; + bool queue_writeback = false; if (session) ceph_get_mds_session(session); @@ -2064,10 +2065,27 @@ void ceph_check_caps(struct ceph_inode_info *ci, int flags, } /* completed revocation? going down and there are no caps? */ - if (revoking && (revoking & cap_used) == 0) { - dout("completed revocation of %s\n", - ceph_cap_string(cap->implemented & ~cap->issued)); - goto ack; + if (revoking) { + if ((revoking & cap_used) == 0) { + dout("completed revocation of %s\n", + ceph_cap_string(cap->implemented & ~cap->issued)); + goto ack; + } + + /* + * If the "i_wrbuffer_ref" was increased by mmap or generic + * cache write just before the ceph_check_caps() is called, + * the Fb capability revoking will fail this time. Then we + * must wait for the BDI's delayed work to flush the dirty + * pages and to release the "i_wrbuffer_ref", which will cost + * at most 5 seconds. That means the MDS needs to wait at + * most 5 seconds to finished the Fb capability's revocation. + * + * Let's queue a writeback for it. + */ + if (S_ISREG(inode->i_mode) && ci->i_wrbuffer_ref && + (revoking & CEPH_CAP_FILE_BUFFER)) + queue_writeback = true; } /* want more caps from mds? */ @@ -2137,6 +2155,8 @@ void ceph_check_caps(struct ceph_inode_info *ci, int flags, spin_unlock(&ci->i_ceph_lock); ceph_put_mds_session(session); + if (queue_writeback) + ceph_queue_writeback(inode); if (queue_invalidate) ceph_queue_invalidate(inode); }
If the pagecaches writeback just finished and the i_wrbuffer_ref reaches zero it will try to trigger ceph_check_caps(). But if just before ceph_check_caps() the i_wrbuffer_ref could be increased again by mmap/cache write, then the Fwb revoke will fail. We need to try to queue a writeback in this case instead of triggering the writeback by BDI's delayed work per 5 seconds. URL: https://tracker.ceph.com/issues/55377 URL: https://tracker.ceph.com/issues/46904 Signed-off-by: Xiubo Li <xiubli@redhat.com> --- fs/ceph/caps.c | 28 ++++++++++++++++++++++++---- 1 file changed, 24 insertions(+), 4 deletions(-)