ceph: avoid carrying Fw cap during write into page cache
The generic_file_aio_write call may block on balance_dirty_pages while we
flush data to the OSDs. If we hold a reference to the FILE_WR cap during
that interval revocation by the MDS (e.g., to do a stat(2)) may be very
slow.
Reviewed-by: Yehuda Sadeh <yehuda@hq.newdream.net>
Signed-off-by: Sage Weil <sage@newdream.net>
diff --git a/fs/ceph/file.c b/fs/ceph/file.c
index 44e4fe9..6c90cf0 100644
--- a/fs/ceph/file.c
+++ b/fs/ceph/file.c
@@ -713,7 +713,7 @@
want = CEPH_CAP_FILE_BUFFER;
ret = ceph_get_caps(ci, CEPH_CAP_FILE_WR, want, &got, endoff);
if (ret < 0)
- goto out;
+ goto out_put;
dout("aio_write %p %llx.%llx %llu~%u got cap refs on %s\n",
inode, ceph_vinop(inode), pos, (unsigned)iov->iov_len,
@@ -726,8 +726,18 @@
ret = ceph_sync_write(file, iov->iov_base, iov->iov_len,
&iocb->ki_pos);
} else {
- ret = generic_file_aio_write(iocb, iov, nr_segs, pos);
+ /*
+ * buffered write; drop Fw early to avoid slow
+ * revocation if we get stuck on balance_dirty_pages
+ */
+ int dirty;
+ spin_lock(&inode->i_lock);
+ dirty = __ceph_mark_dirty_caps(ci, CEPH_CAP_FILE_WR);
+ spin_unlock(&inode->i_lock);
+ ceph_put_cap_refs(ci, got);
+
+ ret = generic_file_aio_write(iocb, iov, nr_segs, pos);
if ((ret >= 0 || ret == -EIOCBQUEUED) &&
((file->f_flags & O_SYNC) || IS_SYNC(file->f_mapping->host)
|| ceph_osdmap_flag(osdc->osdmap, CEPH_OSDMAP_NEARFULL))) {
@@ -735,7 +745,12 @@
if (err < 0)
ret = err;
}
+
+ if (dirty)
+ __mark_inode_dirty(inode, dirty);
+ goto out;
}
+
if (ret >= 0) {
int dirty;
spin_lock(&inode->i_lock);
@@ -745,12 +760,13 @@
__mark_inode_dirty(inode, dirty);
}
-out:
+out_put:
dout("aio_write %p %llx.%llx %llu~%u dropping cap refs on %s\n",
inode, ceph_vinop(inode), pos, (unsigned)iov->iov_len,
ceph_cap_string(got));
ceph_put_cap_refs(ci, got);
+out:
if (ret == -EOLDSNAPC) {
dout("aio_write %p %llx.%llx %llu~%u got EOLDSNAPC, retrying\n",
inode, ceph_vinop(inode), pos, (unsigned)iov->iov_len);