From 494ddd11be3e2621096bb425eed2886f8e8446d4 Mon Sep 17 00:00:00 2001 From: majianpeng Date: Tue, 16 Jul 2013 19:36:21 +0800 Subject: ceph: Don't forget the 'up_read(&osdc->map_sem)' if met error. CC: stable@vger.kernel.org Signed-off-by: Jianpeng Ma Reviewed-by: Sage Weil --- fs/ceph/ioctl.c | 4 +++- 1 file changed, 3 insertions(+), 1 deletion(-) (limited to 'fs') diff --git a/fs/ceph/ioctl.c b/fs/ceph/ioctl.c index e0b4ef31d3c8..a5ce62eb7806 100644 --- a/fs/ceph/ioctl.c +++ b/fs/ceph/ioctl.c @@ -196,8 +196,10 @@ static long ceph_ioctl_get_dataloc(struct file *file, void __user *arg) r = ceph_calc_file_object_mapping(&ci->i_layout, dl.file_offset, len, &dl.object_no, &dl.object_offset, &olen); - if (r < 0) + if (r < 0) { + up_read(&osdc->map_sem); return -EIO; + } dl.file_offset -= dl.object_offset; dl.object_size = ceph_file_layout_object_size(ci->i_layout); dl.block_size = ceph_file_layout_su(ci->i_layout); -- cgit v1.2.3-59-g8ed1b From c338c07c51e3106711fad5eb599e375eadb6855d Mon Sep 17 00:00:00 2001 From: Nathaniel Yazdani Date: Sun, 4 Aug 2013 21:04:30 -0700 Subject: ceph: fix null pointer dereference When register_session() is given an out-of-range argument for mds, ceph_mdsmap_get_addr() will return a null pointer, which would be given to ceph_con_open() & be dereferenced, causing a kernel oops. This fixes bug #4685 in the Ceph bug tracker . Signed-off-by: Nathaniel Yazdani Reviewed-by: Sage Weil --- fs/ceph/mds_client.c | 3 +++ 1 file changed, 3 insertions(+) (limited to 'fs') diff --git a/fs/ceph/mds_client.c b/fs/ceph/mds_client.c index 3eb1b4470c85..6b40d8112c64 100644 --- a/fs/ceph/mds_client.c +++ b/fs/ceph/mds_client.c @@ -414,6 +414,9 @@ static struct ceph_mds_session *register_session(struct ceph_mds_client *mdsc, { struct ceph_mds_session *s; + if (mds >= mdsc->mdsmap->m_max_mds) + return ERR_PTR(-EINVAL); + s = kzalloc(sizeof(*s), GFP_NOFS); if (!s) return ERR_PTR(-ENOMEM); -- cgit v1.2.3-59-g8ed1b From ad88f23f42a9b34a0b29a5b19d37251ccb7dd776 Mon Sep 17 00:00:00 2001 From: "Yan, Zheng" Date: Sun, 21 Jul 2013 20:25:25 +0800 Subject: ceph: drop CAP_LINK_SHARED when sending "link" request to MDS To handle "link" request, the MDS need to xlock inode's linklock, which requires revoking any CAP_LINK_SHARED. Signed-off-by: Yan, Zheng Reviewed-by: Sage Weil --- fs/ceph/dir.c | 2 ++ 1 file changed, 2 insertions(+) (limited to 'fs') diff --git a/fs/ceph/dir.c b/fs/ceph/dir.c index f02d82b7933e..0e4da4a9c213 100644 --- a/fs/ceph/dir.c +++ b/fs/ceph/dir.c @@ -796,6 +796,8 @@ static int ceph_link(struct dentry *old_dentry, struct inode *dir, req->r_locked_dir = dir; req->r_dentry_drop = CEPH_CAP_FILE_SHARED; req->r_dentry_unless = CEPH_CAP_FILE_EXCL; + /* release LINK_SHARED on source inode (mds will lock it) */ + req->r_old_inode_drop = CEPH_CAP_LINK_SHARED; err = ceph_mdsc_do_request(mdsc, dir, req); if (err) { d_drop(dentry); -- cgit v1.2.3-59-g8ed1b From 85ce127a9adf5ab9e9d57ddf64c858927d5e546d Mon Sep 17 00:00:00 2001 From: "Yan, Zheng" Date: Sun, 21 Jul 2013 20:25:26 +0800 Subject: ceph: wake up writer if vmtruncate work get blocked To write data, the writer first acquires the i_mutex, then try getting caps. The writer may sleep while holding the i_mutex. If the MDS revokes Fb cap in this case, vmtruncate work can't do its job because i_mutex is locked. We should wake up the writer and let it truncate the pages. Signed-off-by: Yan, Zheng Reviewed-by: Sage Weil --- fs/ceph/inode.c | 9 ++++++++- 1 file changed, 8 insertions(+), 1 deletion(-) (limited to 'fs') diff --git a/fs/ceph/inode.c b/fs/ceph/inode.c index 4906ada4a97c..55aaddb4047e 100644 --- a/fs/ceph/inode.c +++ b/fs/ceph/inode.c @@ -1465,7 +1465,14 @@ static void ceph_vmtruncate_work(struct work_struct *work) struct inode *inode = &ci->vfs_inode; dout("vmtruncate_work %p\n", inode); - mutex_lock(&inode->i_mutex); + if (!mutex_trylock(&inode->i_mutex)) { + /* + * the i_mutex can be hold by a writer who is waiting for + * caps. wake up waiters, they will do pending vmtruncate. + */ + wake_up_all(&ci->i_cap_wq); + mutex_lock(&inode->i_mutex); + } __ceph_do_pending_vmtruncate(inode); mutex_unlock(&inode->i_mutex); iput(inode); -- cgit v1.2.3-59-g8ed1b From ca20c991917ef6a98d6b40184fefe981727d9328 Mon Sep 17 00:00:00 2001 From: "Yan, Zheng" Date: Sun, 21 Jul 2013 10:07:51 +0800 Subject: ceph: trim deleted inode The MDS uses caps message to notify clients about deleted inode. when receiving a such message, invalidate any alias of the inode. This makes the kernel release the inode ASAP. Signed-off-by: Yan, Zheng Reviewed-by: Sage Weil --- fs/ceph/caps.c | 41 ++++++++++++++++++++++++++++++++++++++++- 1 file changed, 40 insertions(+), 1 deletion(-) (limited to 'fs') diff --git a/fs/ceph/caps.c b/fs/ceph/caps.c index 25442b40c25a..430121a795bd 100644 --- a/fs/ceph/caps.c +++ b/fs/ceph/caps.c @@ -2333,6 +2333,38 @@ void ceph_put_wrbuffer_cap_refs(struct ceph_inode_info *ci, int nr, iput(inode); } +/* + * Invalidate unlinked inode's aliases, so we can drop the inode ASAP. + */ +static void invalidate_aliases(struct inode *inode) +{ + struct dentry *dn, *prev = NULL; + + dout("invalidate_aliases inode %p\n", inode); + d_prune_aliases(inode); + /* + * For non-directory inode, d_find_alias() only returns + * connected dentry. After calling d_delete(), the dentry + * become disconnected. + * + * For directory inode, d_find_alias() only can return + * disconnected dentry. But directory inode should have + * one alias at most. + */ + while ((dn = d_find_alias(inode))) { + if (dn == prev) { + dput(dn); + break; + } + d_delete(dn); + if (prev) + dput(prev); + prev = dn; + } + if (prev) + dput(prev); +} + /* * Handle a cap GRANT message from the MDS. (Note that a GRANT may * actually be a revocation if it specifies a smaller cap set.) @@ -2363,6 +2395,7 @@ static void handle_cap_grant(struct inode *inode, struct ceph_mds_caps *grant, int writeback = 0; int revoked_rdcache = 0; int queue_invalidate = 0; + int deleted_inode = 0; dout("handle_cap_grant inode %p cap %p mds%d seq %d %s\n", inode, cap, mds, seq, ceph_cap_string(newcaps)); @@ -2407,8 +2440,12 @@ static void handle_cap_grant(struct inode *inode, struct ceph_mds_caps *grant, from_kgid(&init_user_ns, inode->i_gid)); } - if ((issued & CEPH_CAP_LINK_EXCL) == 0) + if ((issued & CEPH_CAP_LINK_EXCL) == 0) { set_nlink(inode, le32_to_cpu(grant->nlink)); + if (inode->i_nlink == 0 && + (newcaps & (CEPH_CAP_LINK_SHARED | CEPH_CAP_LINK_EXCL))) + deleted_inode = 1; + } if ((issued & CEPH_CAP_XATTR_EXCL) == 0 && grant->xattr_len) { int len = le32_to_cpu(grant->xattr_len); @@ -2517,6 +2554,8 @@ static void handle_cap_grant(struct inode *inode, struct ceph_mds_caps *grant, ceph_queue_writeback(inode); if (queue_invalidate) ceph_queue_invalidate(inode); + if (deleted_inode) + invalidate_aliases(inode); if (wake) wake_up_all(&ci->i_cap_wq); -- cgit v1.2.3-59-g8ed1b From 688bac461ba3e9d221a879ab40b687f5d7b5b19c Mon Sep 17 00:00:00 2001 From: Dan Carpenter Date: Tue, 23 Jul 2013 16:48:01 +0300 Subject: ceph: cleanup types in striped_read() We pass in a u64 value for "len" and then immediately truncate away the upper 32 bits. Signed-off-by: Dan Carpenter Reviewed-by: Sage Weil Reviewed-by: Alex Elder --- fs/ceph/file.c | 8 ++++---- 1 file changed, 4 insertions(+), 4 deletions(-) (limited to 'fs') diff --git a/fs/ceph/file.c b/fs/ceph/file.c index a44d5153179b..7d4e769f1d3d 100644 --- a/fs/ceph/file.c +++ b/fs/ceph/file.c @@ -313,9 +313,9 @@ static int striped_read(struct inode *inode, { struct ceph_fs_client *fsc = ceph_inode_to_client(inode); struct ceph_inode_info *ci = ceph_inode(inode); - u64 pos, this_len; + u64 pos, this_len, left; int io_align, page_align; - int left, pages_left; + int pages_left; int read; struct page **page_pos; int ret; @@ -346,7 +346,7 @@ more: ret = 0; hit_stripe = this_len < left; was_short = ret >= 0 && ret < this_len; - dout("striped_read %llu~%u (read %u) got %d%s%s\n", pos, left, read, + dout("striped_read %llu~%llu (read %u) got %d%s%s\n", pos, left, read, ret, hit_stripe ? " HITSTRIPE" : "", was_short ? " SHORT" : ""); if (ret > 0) { @@ -378,7 +378,7 @@ more: if (pos + left > inode->i_size) left = inode->i_size - pos; - dout("zero tail %d\n", left); + dout("zero tail %llu\n", left); ceph_zero_page_vector_range(page_align + read, left, pages); read += left; -- cgit v1.2.3-59-g8ed1b From 7ab9b3807097fcb87b0e85a9ad82b12322d9cc63 Mon Sep 17 00:00:00 2001 From: majianpeng Date: Thu, 27 Jun 2013 15:56:06 +0800 Subject: ceph: Don't use ceph-sync-mode for synchronous-fs. Sending reads and writes through the sync read/write paths bypasses the page cache, which is not expected or generally a good idea. Removing the write check is safe as there is a conditional vfs_fsync_range() later in ceph_aio_write that already checks for the same flag (via IS_SYNC(inode)). Signed-off-by: Jianpeng Ma Reviewed-by: Sage Weil --- fs/ceph/file.c | 2 -- 1 file changed, 2 deletions(-) (limited to 'fs') diff --git a/fs/ceph/file.c b/fs/ceph/file.c index 7d4e769f1d3d..63ec830b9306 100644 --- a/fs/ceph/file.c +++ b/fs/ceph/file.c @@ -659,7 +659,6 @@ again: if ((got & (CEPH_CAP_FILE_CACHE|CEPH_CAP_FILE_LAZYIO)) == 0 || (iocb->ki_filp->f_flags & O_DIRECT) || - (inode->i_sb->s_flags & MS_SYNCHRONOUS) || (fi->flags & CEPH_F_SYNC)) /* hmm, this isn't really async... */ ret = ceph_sync_read(filp, base, len, ppos, &checkeof); @@ -763,7 +762,6 @@ retry_snap: if ((got & (CEPH_CAP_FILE_BUFFER|CEPH_CAP_FILE_LAZYIO)) == 0 || (iocb->ki_filp->f_flags & O_DIRECT) || - (inode->i_sb->s_flags & MS_SYNCHRONOUS) || (fi->flags & CEPH_F_SYNC)) { mutex_unlock(&inode->i_mutex); written = ceph_sync_write(file, iov->iov_base, count, -- cgit v1.2.3-59-g8ed1b From 2fbcbff1d6b9243ef71c64a8ab993bc3c7bb7af1 Mon Sep 17 00:00:00 2001 From: majianpeng Date: Fri, 2 Aug 2013 18:14:48 +0800 Subject: ceph: Add check returned value on func ceph_calc_ceph_pg. Func ceph_calc_ceph_pg maybe failed.So add check for returned value. Signed-off-by: Jianpeng Ma Reviewed-by: Sage Weil Signed-off-by: Sage Weil --- fs/ceph/ioctl.c | 8 ++++++-- 1 file changed, 6 insertions(+), 2 deletions(-) (limited to 'fs') diff --git a/fs/ceph/ioctl.c b/fs/ceph/ioctl.c index a5ce62eb7806..669622fd1ae3 100644 --- a/fs/ceph/ioctl.c +++ b/fs/ceph/ioctl.c @@ -211,8 +211,12 @@ static long ceph_ioctl_get_dataloc(struct file *file, void __user *arg) snprintf(dl.object_name, sizeof(dl.object_name), "%llx.%08llx", ceph_ino(inode), dl.object_no); - ceph_calc_ceph_pg(&pgid, dl.object_name, osdc->osdmap, - ceph_file_layout_pg_pool(ci->i_layout)); + r = ceph_calc_ceph_pg(&pgid, dl.object_name, osdc->osdmap, + ceph_file_layout_pg_pool(ci->i_layout)); + if (r < 0) { + up_read(&osdc->map_sem); + return r; + } dl.osd = ceph_calc_pg_primary(osdc->osdmap, pgid); if (dl.osd >= 0) { -- cgit v1.2.3-59-g8ed1b From 6f60f889470aecf747610279545c054a99aadca3 Mon Sep 17 00:00:00 2001 From: "Yan, Zheng" Date: Wed, 24 Jul 2013 12:22:11 +0800 Subject: ceph: fix freeing inode vs removing session caps race remove_session_caps() uses iterate_session_caps() to remove caps, but iterate_session_caps() skips inodes that are being deleted. So session->s_nr_caps can be non-zero after iterate_session_caps() return. We can fix the issue by waiting until deletions are complete. __wait_on_freeing_inode() is designed for the job, but it is not exported, so we use lookup inode function to access it. Signed-off-by: Yan, Zheng --- fs/ceph/inode.c | 8 ++++++++ fs/ceph/mds_client.c | 31 +++++++++++++++++++++++++++++++ fs/ceph/super.h | 2 ++ 3 files changed, 41 insertions(+) (limited to 'fs') diff --git a/fs/ceph/inode.c b/fs/ceph/inode.c index 55aaddb4047e..3b0abed667c2 100644 --- a/fs/ceph/inode.c +++ b/fs/ceph/inode.c @@ -61,6 +61,14 @@ struct inode *ceph_get_inode(struct super_block *sb, struct ceph_vino vino) return inode; } +struct inode *ceph_lookup_inode(struct super_block *sb, struct ceph_vino vino) +{ + struct inode *inode; + ino_t t = ceph_vino_to_ino(vino); + inode = ilookup5_nowait(sb, t, ceph_ino_compare, &vino); + return inode; +} + /* * get/constuct snapdir inode for a given directory */ diff --git a/fs/ceph/mds_client.c b/fs/ceph/mds_client.c index 6b40d8112c64..cbf08203e00d 100644 --- a/fs/ceph/mds_client.c +++ b/fs/ceph/mds_client.c @@ -1031,6 +1031,37 @@ static void remove_session_caps(struct ceph_mds_session *session) { dout("remove_session_caps on %p\n", session); iterate_session_caps(session, remove_session_caps_cb, NULL); + + spin_lock(&session->s_cap_lock); + if (session->s_nr_caps > 0) { + struct super_block *sb = session->s_mdsc->fsc->sb; + struct inode *inode; + struct ceph_cap *cap, *prev = NULL; + struct ceph_vino vino; + /* + * iterate_session_caps() skips inodes that are being + * deleted, we need to wait until deletions are complete. + * __wait_on_freeing_inode() is designed for the job, + * but it is not exported, so use lookup inode function + * to access it. + */ + while (!list_empty(&session->s_caps)) { + cap = list_entry(session->s_caps.next, + struct ceph_cap, session_caps); + if (cap == prev) + break; + prev = cap; + vino = cap->ci->i_vino; + spin_unlock(&session->s_cap_lock); + + inode = ceph_lookup_inode(sb, vino); + iput(inode); + + spin_lock(&session->s_cap_lock); + } + } + spin_unlock(&session->s_cap_lock); + BUG_ON(session->s_nr_caps > 0); BUG_ON(!list_empty(&session->s_cap_flushing)); cleanup_cap_releases(session); diff --git a/fs/ceph/super.h b/fs/ceph/super.h index cbded572345e..afcd62a68916 100644 --- a/fs/ceph/super.h +++ b/fs/ceph/super.h @@ -677,6 +677,8 @@ extern void ceph_destroy_inode(struct inode *inode); extern struct inode *ceph_get_inode(struct super_block *sb, struct ceph_vino vino); +extern struct inode *ceph_lookup_inode(struct super_block *sb, + struct ceph_vino vino); extern struct inode *ceph_get_snapdir(struct inode *parent); extern int ceph_fill_file_size(struct inode *inode, int issued, u32 truncate_seq, u64 truncate_size, u64 size); -- cgit v1.2.3-59-g8ed1b From 0e5dd45ce4c41d3e3857116a77f34f04c99e78ad Mon Sep 17 00:00:00 2001 From: majianpeng Date: Thu, 8 Aug 2013 15:32:19 +0800 Subject: ceph: Move the place for EOLDSNAPC handle in ceph_aio_write to easily understand Only for ceph_sync_write, the osd can return EOLDSNAPC.so move the related codes after the call ceph_sync_write. Signed-off-by: Jianpeng Ma Reviewed-by: Sage Weil --- fs/ceph/file.c | 16 +++++++++------- 1 file changed, 9 insertions(+), 7 deletions(-) (limited to 'fs') diff --git a/fs/ceph/file.c b/fs/ceph/file.c index 63ec830b9306..7478d5dbd1aa 100644 --- a/fs/ceph/file.c +++ b/fs/ceph/file.c @@ -766,6 +766,15 @@ retry_snap: mutex_unlock(&inode->i_mutex); written = ceph_sync_write(file, iov->iov_base, count, pos, &iocb->ki_pos); + if (written == -EOLDSNAPC) { + dout("aio_write %p %llx.%llx %llu~%u" + "got EOLDSNAPC, retrying\n", + inode, ceph_vinop(inode), + pos, (unsigned)iov->iov_len); + mutex_lock(&inode->i_mutex); + hold_mutex = true; + goto retry_snap; + } } else { written = generic_file_buffered_write(iocb, iov, nr_segs, pos, &iocb->ki_pos, @@ -796,13 +805,6 @@ retry_snap: written = err; } - if (written == -EOLDSNAPC) { - dout("aio_write %p %llx.%llx %llu~%u got EOLDSNAPC, retrying\n", - inode, ceph_vinop(inode), pos, (unsigned)iov->iov_len); - mutex_lock(&inode->i_mutex); - hold_mutex = true; - goto retry_snap; - } out: if (hold_mutex) mutex_unlock(&inode->i_mutex); -- cgit v1.2.3-59-g8ed1b From 2f75e9e17911524f294aa7b3bf0d7233f99a3218 Mon Sep 17 00:00:00 2001 From: Sage Weil Date: Fri, 9 Aug 2013 09:57:58 -0700 Subject: ceph: replace hold_mutex flag with goto All of the early exit paths need to drop the mutex; it is only the normal path through the function that does not. Skip the unlock in that case with a goto out_unlocked. Signed-off-by: Sage Weil Reviewed-by: Jianpeng Ma --- fs/ceph/file.c | 11 ++++------- 1 file changed, 4 insertions(+), 7 deletions(-) (limited to 'fs') diff --git a/fs/ceph/file.c b/fs/ceph/file.c index 7478d5dbd1aa..a17ffe4ec3ca 100644 --- a/fs/ceph/file.c +++ b/fs/ceph/file.c @@ -710,13 +710,11 @@ static ssize_t ceph_aio_write(struct kiocb *iocb, const struct iovec *iov, &ceph_sb_to_client(inode->i_sb)->client->osdc; ssize_t count, written = 0; int err, want, got; - bool hold_mutex; if (ceph_snap(inode) != CEPH_NOSNAP) return -EROFS; mutex_lock(&inode->i_mutex); - hold_mutex = true; err = generic_segment_checks(iov, &nr_segs, &count, VERIFY_READ); if (err) @@ -772,7 +770,6 @@ retry_snap: inode, ceph_vinop(inode), pos, (unsigned)iov->iov_len); mutex_lock(&inode->i_mutex); - hold_mutex = true; goto retry_snap; } } else { @@ -781,7 +778,6 @@ retry_snap: count, 0); mutex_unlock(&inode->i_mutex); } - hold_mutex = false; if (written >= 0) { int dirty; @@ -805,11 +801,12 @@ retry_snap: written = err; } + goto out_unlocked; + out: - if (hold_mutex) - mutex_unlock(&inode->i_mutex); + mutex_unlock(&inode->i_mutex); +out_unlocked: current->backing_dev_info = NULL; - return written ? written : err; } -- cgit v1.2.3-59-g8ed1b From fe2a801b50c0bb8039d627e5ae1fec249d10ff39 Mon Sep 17 00:00:00 2001 From: Milosz Tanski Date: Fri, 9 Aug 2013 12:59:39 -0400 Subject: ceph: Remove bogus check in invalidatepage The early bug checks are moot because the VMA layer ensures those things. 1. It will not call invalidatepage unless PagePrivate (or PagePrivate2) are set 2. It will not call invalidatepage without taking a PageLock first. 3. Guantrees that the inode page is mapped. Signed-off-by: Milosz Tanski Reviewed-by: Sage Weil --- fs/ceph/addr.c | 4 ---- 1 file changed, 4 deletions(-) (limited to 'fs') diff --git a/fs/ceph/addr.c b/fs/ceph/addr.c index afb2fc241061..f1d6c60ab229 100644 --- a/fs/ceph/addr.c +++ b/fs/ceph/addr.c @@ -149,10 +149,6 @@ static void ceph_invalidatepage(struct page *page, unsigned long offset) struct ceph_inode_info *ci; struct ceph_snap_context *snapc = page_snap_context(page); - BUG_ON(!PageLocked(page)); - BUG_ON(!PagePrivate(page)); - BUG_ON(!page->mapping); - inode = page->mapping->host; /* -- cgit v1.2.3-59-g8ed1b From b150f5c1c759d551da9146435d3dc9df5f7e15ef Mon Sep 17 00:00:00 2001 From: Milosz Tanski Date: Fri, 9 Aug 2013 12:59:55 -0400 Subject: ceph: cleanup the logic in ceph_invalidatepage The invalidatepage code bails if it encounters a non-zero page offset. The current logic that does is non-obvious with multiple if statements. This should be logically and functionally equivalent. Signed-off-by: Milosz Tanski Reviewed-by: Sage Weil --- fs/ceph/addr.c | 29 +++++++++++++++-------------- 1 file changed, 15 insertions(+), 14 deletions(-) (limited to 'fs') diff --git a/fs/ceph/addr.c b/fs/ceph/addr.c index 722585cd5c7e..cb78ce81d6a6 100644 --- a/fs/ceph/addr.c +++ b/fs/ceph/addr.c @@ -151,6 +151,13 @@ static void ceph_invalidatepage(struct page *page, unsigned int offset, struct ceph_snap_context *snapc = page_snap_context(page); inode = page->mapping->host; + ci = ceph_inode(inode); + + if (offset != 0 || length != PAGE_CACHE_SIZE) { + dout("%p invalidatepage %p idx %lu partial dirty page %u~%u\n", + inode, page, page->index, offset, length); + return; + } /* * We can get non-dirty pages here due to races between @@ -160,21 +167,15 @@ static void ceph_invalidatepage(struct page *page, unsigned int offset, if (!PageDirty(page)) pr_err("%p invalidatepage %p page not dirty\n", inode, page); - if (offset == 0 && length == PAGE_CACHE_SIZE) - ClearPageChecked(page); + ClearPageChecked(page); - ci = ceph_inode(inode); - if (offset == 0 && length == PAGE_CACHE_SIZE) { - dout("%p invalidatepage %p idx %lu full dirty page\n", - inode, page, page->index); - ceph_put_wrbuffer_cap_refs(ci, 1, snapc); - ceph_put_snap_context(snapc); - page->private = 0; - ClearPagePrivate(page); - } else { - dout("%p invalidatepage %p idx %lu partial dirty page %u(%u)\n", - inode, page, page->index, offset, length); - } + dout("%p invalidatepage %p idx %lu full dirty page\n", + inode, page, page->index); + + ceph_put_wrbuffer_cap_refs(ci, 1, snapc); + ceph_put_snap_context(snapc); + page->private = 0; + ClearPagePrivate(page); } /* just a sanity check */ -- cgit v1.2.3-59-g8ed1b From b0d7c2231015b331b942746610a05b6ea72977ab Mon Sep 17 00:00:00 2001 From: "Yan, Zheng" Date: Mon, 12 Aug 2013 21:42:15 -0700 Subject: ceph: introduce i_truncate_mutex I encountered below deadlock when running fsstress wmtruncate work truncate MDS --------------- ------------------ -------------------------- lock i_mutex <- truncate file lock i_mutex (blocked) <- revoking Fcb (filelock to MIX) send request -> handle request (xlock filelock) At the initial time, there are some dirty pages in the page cache. When the kclient receives the truncate message, it reduces inode size and creates some 'out of i_size' dirty pages. wmtruncate work can't truncate these dirty pages because it's blocked by the i_mutex. Later when the kclient receives the cap message that revokes Fcb caps, It can't flush all dirty pages because writepages() only flushes dirty pages within the inode size. When the MDS handles the 'truncate' request from kclient, it waits for the filelock to become stable. But the filelock is stuck in unstable state because it can't finish revoking kclient's Fcb caps. The truncate pagecache locking has already caused lots of trouble for use. I think it's time simplify it by introducing a new mutex. We use the new mutex to prevent concurrent truncate_inode_pages(). There is no need to worry about race between buffered write and truncate_inode_pages(), because our "get caps" mechanism prevents them from concurrent execution. Reviewed-by: Sage Weil Signed-off-by: Yan, Zheng --- fs/ceph/caps.c | 4 ---- fs/ceph/file.c | 8 +++++++- fs/ceph/inode.c | 39 ++++++++++++++++++++++----------------- fs/ceph/super.h | 1 + 4 files changed, 30 insertions(+), 22 deletions(-) (limited to 'fs') diff --git a/fs/ceph/caps.c b/fs/ceph/caps.c index 430121a795bd..0e94d27fa284 100644 --- a/fs/ceph/caps.c +++ b/fs/ceph/caps.c @@ -2072,11 +2072,7 @@ static int try_get_cap_refs(struct ceph_inode_info *ci, int need, int want, /* finish pending truncate */ while (ci->i_truncate_pending) { spin_unlock(&ci->i_ceph_lock); - if (!(need & CEPH_CAP_FILE_WR)) - mutex_lock(&inode->i_mutex); __ceph_do_pending_vmtruncate(inode); - if (!(need & CEPH_CAP_FILE_WR)) - mutex_unlock(&inode->i_mutex); spin_lock(&ci->i_ceph_lock); } diff --git a/fs/ceph/file.c b/fs/ceph/file.c index bc0735498d29..abc0e0759bdc 100644 --- a/fs/ceph/file.c +++ b/fs/ceph/file.c @@ -773,6 +773,13 @@ retry_snap: goto retry_snap; } } else { + /* + * No need to acquire the i_truncate_mutex. Because + * the MDS revokes Fwb caps before sending truncate + * message to us. We can't get Fwb cap while there + * are pending vmtruncate. So write and vmtruncate + * can not run at the same time + */ written = generic_file_buffered_write(iocb, iov, nr_segs, pos, &iocb->ki_pos, count, 0); @@ -819,7 +826,6 @@ static loff_t ceph_llseek(struct file *file, loff_t offset, int whence) int ret; mutex_lock(&inode->i_mutex); - __ceph_do_pending_vmtruncate(inode); if (whence == SEEK_END || whence == SEEK_DATA || whence == SEEK_HOLE) { ret = ceph_do_getattr(inode, CEPH_STAT_CAP_SIZE); diff --git a/fs/ceph/inode.c b/fs/ceph/inode.c index 98b6e50bde04..602ccd8e06b7 100644 --- a/fs/ceph/inode.c +++ b/fs/ceph/inode.c @@ -352,6 +352,7 @@ struct inode *ceph_alloc_inode(struct super_block *sb) for (i = 0; i < CEPH_FILE_MODE_NUM; i++) ci->i_nr_by_mode[i] = 0; + mutex_init(&ci->i_truncate_mutex); ci->i_truncate_seq = 0; ci->i_truncate_size = 0; ci->i_truncate_pending = 0; @@ -463,16 +464,20 @@ int ceph_fill_file_size(struct inode *inode, int issued, dout("truncate_seq %u -> %u\n", ci->i_truncate_seq, truncate_seq); ci->i_truncate_seq = truncate_seq; + + /* the MDS should have revoked these caps */ + WARN_ON_ONCE(issued & (CEPH_CAP_FILE_EXCL | + CEPH_CAP_FILE_RD | + CEPH_CAP_FILE_WR | + CEPH_CAP_FILE_LAZYIO)); /* * If we hold relevant caps, or in the case where we're * not the only client referencing this file and we * don't hold those caps, then we need to check whether * the file is either opened or mmaped */ - if ((issued & (CEPH_CAP_FILE_CACHE|CEPH_CAP_FILE_RD| - CEPH_CAP_FILE_WR|CEPH_CAP_FILE_BUFFER| - CEPH_CAP_FILE_EXCL| - CEPH_CAP_FILE_LAZYIO)) || + if ((issued & (CEPH_CAP_FILE_CACHE| + CEPH_CAP_FILE_BUFFER)) || mapping_mapped(inode->i_mapping) || __ceph_caps_file_wanted(ci)) { ci->i_truncate_pending++; @@ -1427,18 +1432,20 @@ static void ceph_invalidate_work(struct work_struct *work) u32 orig_gen; int check = 0; + mutex_lock(&ci->i_truncate_mutex); spin_lock(&ci->i_ceph_lock); dout("invalidate_pages %p gen %d revoking %d\n", inode, ci->i_rdcache_gen, ci->i_rdcache_revoking); if (ci->i_rdcache_revoking != ci->i_rdcache_gen) { /* nevermind! */ spin_unlock(&ci->i_ceph_lock); + mutex_unlock(&ci->i_truncate_mutex); goto out; } orig_gen = ci->i_rdcache_gen; spin_unlock(&ci->i_ceph_lock); - truncate_inode_pages(&inode->i_data, 0); + truncate_inode_pages(inode->i_mapping, 0); spin_lock(&ci->i_ceph_lock); if (orig_gen == ci->i_rdcache_gen && @@ -1453,6 +1460,7 @@ static void ceph_invalidate_work(struct work_struct *work) ci->i_rdcache_revoking); } spin_unlock(&ci->i_ceph_lock); + mutex_unlock(&ci->i_truncate_mutex); if (check) ceph_check_caps(ci, 0, NULL); @@ -1473,16 +1481,7 @@ static void ceph_vmtruncate_work(struct work_struct *work) struct inode *inode = &ci->vfs_inode; dout("vmtruncate_work %p\n", inode); - if (!mutex_trylock(&inode->i_mutex)) { - /* - * the i_mutex can be hold by a writer who is waiting for - * caps. wake up waiters, they will do pending vmtruncate. - */ - wake_up_all(&ci->i_cap_wq); - mutex_lock(&inode->i_mutex); - } __ceph_do_pending_vmtruncate(inode); - mutex_unlock(&inode->i_mutex); iput(inode); } @@ -1515,11 +1514,13 @@ void __ceph_do_pending_vmtruncate(struct inode *inode) u64 to; int wrbuffer_refs, finish = 0; + mutex_lock(&ci->i_truncate_mutex); retry: spin_lock(&ci->i_ceph_lock); if (ci->i_truncate_pending == 0) { dout("__do_pending_vmtruncate %p none pending\n", inode); spin_unlock(&ci->i_ceph_lock); + mutex_unlock(&ci->i_truncate_mutex); return; } @@ -1536,6 +1537,9 @@ retry: goto retry; } + /* there should be no reader or writer */ + WARN_ON_ONCE(ci->i_rd_ref || ci->i_wr_ref); + to = ci->i_truncate_size; wrbuffer_refs = ci->i_wrbuffer_ref; dout("__do_pending_vmtruncate %p (%d) to %lld\n", inode, @@ -1553,6 +1557,8 @@ retry: if (!finish) goto retry; + mutex_unlock(&ci->i_truncate_mutex); + if (wrbuffer_refs == 0) ceph_check_caps(ci, CHECK_CAPS_AUTHONLY, NULL); @@ -1601,8 +1607,6 @@ int ceph_setattr(struct dentry *dentry, struct iattr *attr) if (ceph_snap(inode) != CEPH_NOSNAP) return -EROFS; - __ceph_do_pending_vmtruncate(inode); - err = inode_change_ok(inode, attr); if (err != 0) return err; @@ -1783,7 +1787,8 @@ int ceph_setattr(struct dentry *dentry, struct iattr *attr) ceph_cap_string(dirtied), mask); ceph_mdsc_put_request(req); - __ceph_do_pending_vmtruncate(inode); + if (mask & CEPH_SETATTR_SIZE) + __ceph_do_pending_vmtruncate(inode); return err; out: spin_unlock(&ci->i_ceph_lock); diff --git a/fs/ceph/super.h b/fs/ceph/super.h index afcd62a68916..f1e4e4766ea2 100644 --- a/fs/ceph/super.h +++ b/fs/ceph/super.h @@ -288,6 +288,7 @@ struct ceph_inode_info { int i_nr_by_mode[CEPH_FILE_MODE_NUM]; /* open file counts */ + struct mutex i_truncate_mutex; u32 i_truncate_seq; /* last truncate to smaller size */ u64 i_truncate_size; /* and the size we last truncated down to */ int i_truncate_pending; /* still need to call vmtruncate */ -- cgit v1.2.3-59-g8ed1b From 3871cbb9a41b1371dc13fc619e3ab4e0a1e29b4a Mon Sep 17 00:00:00 2001 From: "Yan, Zheng" Date: Mon, 5 Aug 2013 14:10:29 +0800 Subject: ceph: fix request max size ceph_check_caps() requests new max size only when there is Fw cap. If we call check_max_size() while there is no Fw cap. It updates i_wanted_max_size and calls ceph_check_caps(), but ceph_check_caps() does nothing. Later when Fw cap is issued, we call check_max_size() again. But i_wanted_max_size is equal to 'endoff' at this time, so check_max_size() doesn't call ceph_check_caps() and we end up with waiting for the new max size forever. The fix is duplicate ceph_check_caps()'s "request max size" code in check_max_size(), and make try_get_cap_refs() wait for the Fw cap before retry requesting new max size. This patch also removes the "endoff > (inode->i_size << 1)" check in check_max_size(). It's useless because there is no corresponding logic in ceph_check_caps(). Reviewed-by: Sage Weil Signed-off-by: Yan, Zheng --- fs/ceph/caps.c | 18 +++++++++++------- 1 file changed, 11 insertions(+), 7 deletions(-) (limited to 'fs') diff --git a/fs/ceph/caps.c b/fs/ceph/caps.c index 0e94d27fa284..165ebbeab1c3 100644 --- a/fs/ceph/caps.c +++ b/fs/ceph/caps.c @@ -2076,11 +2076,13 @@ static int try_get_cap_refs(struct ceph_inode_info *ci, int need, int want, spin_lock(&ci->i_ceph_lock); } - if (need & CEPH_CAP_FILE_WR) { + have = __ceph_caps_issued(ci, &implemented); + + if (have & need & CEPH_CAP_FILE_WR) { if (endoff >= 0 && endoff > (loff_t)ci->i_max_size) { dout("get_cap_refs %p endoff %llu > maxsize %llu\n", inode, endoff, ci->i_max_size); - if (endoff > ci->i_wanted_max_size) { + if (endoff > ci->i_requested_max_size) { *check_max = 1; ret = 1; } @@ -2095,7 +2097,6 @@ static int try_get_cap_refs(struct ceph_inode_info *ci, int need, int want, goto out; } } - have = __ceph_caps_issued(ci, &implemented); if ((have & need) == need) { /* @@ -2137,14 +2138,17 @@ static void check_max_size(struct inode *inode, loff_t endoff) /* do we need to explicitly request a larger max_size? */ spin_lock(&ci->i_ceph_lock); - if ((endoff >= ci->i_max_size || - endoff > (inode->i_size << 1)) && - endoff > ci->i_wanted_max_size) { + if (endoff >= ci->i_max_size && endoff > ci->i_wanted_max_size) { dout("write %p at large endoff %llu, req max_size\n", inode, endoff); ci->i_wanted_max_size = endoff; - check = 1; } + /* duplicate ceph_check_caps()'s logic */ + if (ci->i_auth_cap && + (ci->i_auth_cap->issued & CEPH_CAP_FILE_WR) && + ci->i_wanted_max_size > ci->i_max_size && + ci->i_wanted_max_size > ci->i_requested_max_size) + check = 1; spin_unlock(&ci->i_ceph_lock); if (check) ceph_check_caps(ci, CHECK_CAPS_AUTHONLY, NULL); -- cgit v1.2.3-59-g8ed1b From ad7a60de882aca31afb58721db166f7e77afcd92 Mon Sep 17 00:00:00 2001 From: Li Wang Date: Thu, 15 Aug 2013 11:51:44 +0800 Subject: ceph: punch hole support This patch implements fallocate and punch hole support for Ceph kernel client. Signed-off-by: Li Wang Signed-off-by: Yunchuan Wen --- fs/ceph/file.c | 196 ++++++++++++++++++++++++++++++++++++++++++++++++++ net/ceph/osd_client.c | 11 ++- 2 files changed, 205 insertions(+), 2 deletions(-) (limited to 'fs') diff --git a/fs/ceph/file.c b/fs/ceph/file.c index abc0e0759bdc..68af489c2abd 100644 --- a/fs/ceph/file.c +++ b/fs/ceph/file.c @@ -8,6 +8,7 @@ #include #include #include +#include #include "super.h" #include "mds_client.h" @@ -874,6 +875,200 @@ out: return offset; } +static inline void ceph_zero_partial_page( + struct inode *inode, loff_t offset, unsigned size) +{ + struct page *page; + pgoff_t index = offset >> PAGE_CACHE_SHIFT; + + page = find_lock_page(inode->i_mapping, index); + if (page) { + wait_on_page_writeback(page); + zero_user(page, offset & (PAGE_CACHE_SIZE - 1), size); + unlock_page(page); + page_cache_release(page); + } +} + +static void ceph_zero_pagecache_range(struct inode *inode, loff_t offset, + loff_t length) +{ + loff_t nearly = round_up(offset, PAGE_CACHE_SIZE); + if (offset < nearly) { + loff_t size = nearly - offset; + if (length < size) + size = length; + ceph_zero_partial_page(inode, offset, size); + offset += size; + length -= size; + } + if (length >= PAGE_CACHE_SIZE) { + loff_t size = round_down(length, PAGE_CACHE_SIZE); + truncate_pagecache_range(inode, offset, offset + size - 1); + offset += size; + length -= size; + } + if (length) + ceph_zero_partial_page(inode, offset, length); +} + +static int ceph_zero_partial_object(struct inode *inode, + loff_t offset, loff_t *length) +{ + struct ceph_inode_info *ci = ceph_inode(inode); + struct ceph_fs_client *fsc = ceph_inode_to_client(inode); + struct ceph_osd_request *req; + int ret = 0; + loff_t zero = 0; + int op; + + if (!length) { + op = offset ? CEPH_OSD_OP_DELETE : CEPH_OSD_OP_TRUNCATE; + length = &zero; + } else { + op = CEPH_OSD_OP_ZERO; + } + + req = ceph_osdc_new_request(&fsc->client->osdc, &ci->i_layout, + ceph_vino(inode), + offset, length, + 1, op, + CEPH_OSD_FLAG_WRITE | + CEPH_OSD_FLAG_ONDISK, + NULL, 0, 0, false); + if (IS_ERR(req)) { + ret = PTR_ERR(req); + goto out; + } + + ceph_osdc_build_request(req, offset, NULL, ceph_vino(inode).snap, + &inode->i_mtime); + + ret = ceph_osdc_start_request(&fsc->client->osdc, req, false); + if (!ret) { + ret = ceph_osdc_wait_request(&fsc->client->osdc, req); + if (ret == -ENOENT) + ret = 0; + } + ceph_osdc_put_request(req); + +out: + return ret; +} + +static int ceph_zero_objects(struct inode *inode, loff_t offset, loff_t length) +{ + int ret = 0; + struct ceph_inode_info *ci = ceph_inode(inode); + __s32 stripe_unit = ceph_file_layout_su(ci->i_layout); + __s32 stripe_count = ceph_file_layout_stripe_count(ci->i_layout); + __s32 object_size = ceph_file_layout_object_size(ci->i_layout); + loff_t object_set_size = (loff_t)object_size * stripe_count; + + loff_t nearly = (offset + object_set_size - 1) + / object_set_size * object_set_size; + while (length && offset < nearly) { + loff_t size = length; + ret = ceph_zero_partial_object(inode, offset, &size); + if (ret < 0) + return ret; + offset += size; + length -= size; + } + while (length >= object_set_size) { + int i; + loff_t pos = offset; + for (i = 0; i < stripe_count; ++i) { + ret = ceph_zero_partial_object(inode, pos, NULL); + if (ret < 0) + return ret; + pos += stripe_unit; + } + offset += object_set_size; + length -= object_set_size; + } + while (length) { + loff_t size = length; + ret = ceph_zero_partial_object(inode, offset, &size); + if (ret < 0) + return ret; + offset += size; + length -= size; + } + return ret; +} + +static long ceph_fallocate(struct file *file, int mode, + loff_t offset, loff_t length) +{ + struct ceph_file_info *fi = file->private_data; + struct inode *inode = file->f_dentry->d_inode; + struct ceph_inode_info *ci = ceph_inode(inode); + struct ceph_osd_client *osdc = + &ceph_inode_to_client(inode)->client->osdc; + int want, got = 0; + int dirty; + int ret = 0; + loff_t endoff = 0; + loff_t size; + + if (!S_ISREG(inode->i_mode)) + return -EOPNOTSUPP; + + if (IS_SWAPFILE(inode)) + return -ETXTBSY; + + mutex_lock(&inode->i_mutex); + + if (ceph_snap(inode) != CEPH_NOSNAP) { + ret = -EROFS; + goto unlock; + } + + if (ceph_osdmap_flag(osdc->osdmap, CEPH_OSDMAP_FULL) && + !(mode & FALLOC_FL_PUNCH_HOLE)) { + ret = -ENOSPC; + goto unlock; + } + + size = i_size_read(inode); + if (!(mode & FALLOC_FL_KEEP_SIZE)) + endoff = offset + length; + + if (fi->fmode & CEPH_FILE_MODE_LAZY) + want = CEPH_CAP_FILE_BUFFER | CEPH_CAP_FILE_LAZYIO; + else + want = CEPH_CAP_FILE_BUFFER; + + ret = ceph_get_caps(ci, CEPH_CAP_FILE_WR, want, &got, endoff); + if (ret < 0) + goto unlock; + + if (mode & FALLOC_FL_PUNCH_HOLE) { + if (offset < size) + ceph_zero_pagecache_range(inode, offset, length); + ret = ceph_zero_objects(inode, offset, length); + } else if (endoff > size) { + truncate_pagecache_range(inode, size, -1); + if (ceph_inode_set_size(inode, endoff)) + ceph_check_caps(ceph_inode(inode), + CHECK_CAPS_AUTHONLY, NULL); + } + + if (!ret) { + spin_lock(&ci->i_ceph_lock); + dirty = __ceph_mark_dirty_caps(ci, CEPH_CAP_FILE_WR); + spin_unlock(&ci->i_ceph_lock); + if (dirty) + __mark_inode_dirty(inode, dirty); + } + + ceph_put_cap_refs(ci, got); +unlock: + mutex_unlock(&inode->i_mutex); + return ret; +} + const struct file_operations ceph_file_fops = { .open = ceph_open, .release = ceph_release, @@ -890,5 +1085,6 @@ const struct file_operations ceph_file_fops = { .splice_write = generic_file_splice_write, .unlocked_ioctl = ceph_ioctl, .compat_ioctl = ceph_ioctl, + .fallocate = ceph_fallocate, }; diff --git a/net/ceph/osd_client.c b/net/ceph/osd_client.c index dbc0a7392d67..8ec65bc11c71 100644 --- a/net/ceph/osd_client.c +++ b/net/ceph/osd_client.c @@ -503,7 +503,9 @@ void osd_req_op_extent_init(struct ceph_osd_request *osd_req, struct ceph_osd_req_op *op = _osd_req_op_init(osd_req, which, opcode); size_t payload_len = 0; - BUG_ON(opcode != CEPH_OSD_OP_READ && opcode != CEPH_OSD_OP_WRITE); + BUG_ON(opcode != CEPH_OSD_OP_READ && opcode != CEPH_OSD_OP_WRITE && + opcode != CEPH_OSD_OP_DELETE && opcode != CEPH_OSD_OP_ZERO && + opcode != CEPH_OSD_OP_TRUNCATE); op->extent.offset = offset; op->extent.length = length; @@ -631,6 +633,9 @@ static u64 osd_req_encode_op(struct ceph_osd_request *req, break; case CEPH_OSD_OP_READ: case CEPH_OSD_OP_WRITE: + case CEPH_OSD_OP_ZERO: + case CEPH_OSD_OP_DELETE: + case CEPH_OSD_OP_TRUNCATE: if (src->op == CEPH_OSD_OP_WRITE) request_data_len = src->extent.length; dst->extent.offset = cpu_to_le64(src->extent.offset); @@ -715,7 +720,9 @@ struct ceph_osd_request *ceph_osdc_new_request(struct ceph_osd_client *osdc, u64 object_base; int r; - BUG_ON(opcode != CEPH_OSD_OP_READ && opcode != CEPH_OSD_OP_WRITE); + BUG_ON(opcode != CEPH_OSD_OP_READ && opcode != CEPH_OSD_OP_WRITE && + opcode != CEPH_OSD_OP_DELETE && opcode != CEPH_OSD_OP_ZERO && + opcode != CEPH_OSD_OP_TRUNCATE); req = ceph_osdc_alloc_request(osdc, snapc, num_ops, use_mempool, GFP_NOFS); -- cgit v1.2.3-59-g8ed1b From b314a90d8f3f1d16ec45744e5e2141ea6e14e034 Mon Sep 17 00:00:00 2001 From: Sage Weil Date: Tue, 27 Aug 2013 12:15:16 -0700 Subject: ceph: fix fallocate division We need to use do_div to divide by a 64-bit value. Signed-off-by: Sage Weil Reviewed-by: Josh Durgin --- fs/ceph/file.c | 16 ++++++++++------ 1 file changed, 10 insertions(+), 6 deletions(-) (limited to 'fs') diff --git a/fs/ceph/file.c b/fs/ceph/file.c index 68af489c2abd..d5e12f580671 100644 --- a/fs/ceph/file.c +++ b/fs/ceph/file.c @@ -960,13 +960,17 @@ static int ceph_zero_objects(struct inode *inode, loff_t offset, loff_t length) { int ret = 0; struct ceph_inode_info *ci = ceph_inode(inode); - __s32 stripe_unit = ceph_file_layout_su(ci->i_layout); - __s32 stripe_count = ceph_file_layout_stripe_count(ci->i_layout); - __s32 object_size = ceph_file_layout_object_size(ci->i_layout); - loff_t object_set_size = (loff_t)object_size * stripe_count; + s32 stripe_unit = ceph_file_layout_su(ci->i_layout); + s32 stripe_count = ceph_file_layout_stripe_count(ci->i_layout); + s32 object_size = ceph_file_layout_object_size(ci->i_layout); + u64 object_set_size = object_size * stripe_count; + u64 nearly, t; + + /* round offset up to next period boundary */ + nearly = offset + object_set_size - 1; + t = nearly; + nearly -= do_div(t, object_set_size); - loff_t nearly = (offset + object_set_size - 1) - / object_set_size * object_set_size; while (length && offset < nearly) { loff_t size = length; ret = ceph_zero_partial_object(inode, offset, &size); -- cgit v1.2.3-59-g8ed1b From e90757432361bb8b3ad3c3fd866324ed47875693 Mon Sep 17 00:00:00 2001 From: Li Wang Date: Thu, 15 Aug 2013 22:00:25 -0700 Subject: ceph: remove useless variable revoked_rdcache Cleanup in handle_cap_grant(). Signed-off-by: Li Wang Reviewed-by: Sage Weil --- fs/ceph/caps.c | 5 +---- 1 file changed, 1 insertion(+), 4 deletions(-) (limited to 'fs') diff --git a/fs/ceph/caps.c b/fs/ceph/caps.c index 165ebbeab1c3..5a26bc1dd799 100644 --- a/fs/ceph/caps.c +++ b/fs/ceph/caps.c @@ -2393,7 +2393,6 @@ static void handle_cap_grant(struct inode *inode, struct ceph_mds_caps *grant, int check_caps = 0; int wake = 0; int writeback = 0; - int revoked_rdcache = 0; int queue_invalidate = 0; int deleted_inode = 0; @@ -2410,9 +2409,7 @@ static void handle_cap_grant(struct inode *inode, struct ceph_mds_caps *grant, if (((cap->issued & ~newcaps) & CEPH_CAP_FILE_CACHE) && (newcaps & CEPH_CAP_FILE_LAZYIO) == 0 && !ci->i_wrbuffer_ref) { - if (try_nonblocking_invalidate(inode) == 0) { - revoked_rdcache = 1; - } else { + if (try_nonblocking_invalidate(inode)) { /* there were locked pages.. invalidate later in a separate thread. */ if (ci->i_rdcache_revoking != ci->i_rdcache_gen) { -- cgit v1.2.3-59-g8ed1b From 02ae66d8b229708fd94b764f6c17ead1c7741fcf Mon Sep 17 00:00:00 2001 From: majianpeng Date: Tue, 6 Aug 2013 16:20:38 +0800 Subject: ceph: fix bugs about handling short-read for sync read mode. cephfs . show_layout >layyout.data_pool: 0 >layout.object_size: 4194304 >layout.stripe_unit: 4194304 >layout.stripe_count: 1 TestA: >dd if=/dev/urandom of=test bs=1M count=2 oflag=direct >dd if=/dev/urandom of=test bs=1M count=2 seek=4 oflag=direct >dd if=test of=/dev/null bs=6M count=1 iflag=direct The messages from func striped_read are: ceph: file.c:350 : striped_read 0~6291456 (read 0) got 2097152 HITSTRIPE SHORT ceph: file.c:350 : striped_read 2097152~4194304 (read 2097152) got 0 HITSTRIPE SHORT ceph: file.c:381 : zero tail 4194304 ceph: file.c:390 : striped_read returns 6291456 The hole of file is from 2M--4M.But actualy it zero the last 4M include the last 2M area which isn't a hole. Using this patch, the messages are: ceph: file.c:350 : striped_read 0~6291456 (read 0) got 2097152 HITSTRIPE SHORT ceph: file.c:358 : zero gap 2097152 to 4194304 ceph: file.c:350 : striped_read 4194304~2097152 (read 4194304) got 2097152 ceph: file.c:384 : striped_read returns 6291456 TestB: >echo majianpeng > test >dd if=test of=/dev/null bs=2M count=1 iflag=direct The messages are: ceph: file.c:350 : striped_read 0~6291456 (read 0) got 11 HITSTRIPE SHORT ceph: file.c:350 : striped_read 11~6291445 (read 11) got 0 HITSTRIPE SHORT ceph: file.c:390 : striped_read returns 11 For this case,it did once more striped_read.It's no meaningless. Using this patch, the message are: ceph: file.c:350 : striped_read 0~6291456 (read 0) got 11 HITSTRIPE SHORT ceph: file.c:384 : striped_read returns 11 Big thanks to Yan Zheng for the patch. Reviewed-by: Yan, Zheng Signed-off-by: Jianpeng Ma --- fs/ceph/file.c | 39 ++++++++++++++++----------------------- 1 file changed, 16 insertions(+), 23 deletions(-) (limited to 'fs') diff --git a/fs/ceph/file.c b/fs/ceph/file.c index d5e12f580671..98b9035b2e81 100644 --- a/fs/ceph/file.c +++ b/fs/ceph/file.c @@ -350,44 +350,37 @@ more: dout("striped_read %llu~%llu (read %u) got %d%s%s\n", pos, left, read, ret, hit_stripe ? " HITSTRIPE" : "", was_short ? " SHORT" : ""); - if (ret > 0) { - int didpages = (page_align + ret) >> PAGE_CACHE_SHIFT; - - if (read < pos - off) { - dout(" zero gap %llu to %llu\n", off + read, pos); - ceph_zero_page_vector_range(page_align + read, - pos - off - read, pages); + if (ret >= 0) { + int didpages; + if (was_short && (pos + ret < inode->i_size)) { + u64 tmp = min(this_len - ret, + inode->i_size - pos - ret); + dout(" zero gap %llu to %llu\n", + pos + ret, pos + ret + tmp); + ceph_zero_page_vector_range(page_align + read + ret, + tmp, pages); + ret += tmp; } + + didpages = (page_align + ret) >> PAGE_CACHE_SHIFT; pos += ret; read = pos - off; left -= ret; page_pos += didpages; pages_left -= didpages; - /* hit stripe? */ - if (left && hit_stripe) + /* hit stripe and need continue*/ + if (left && hit_stripe && pos < inode->i_size) goto more; } - if (was_short) { + if (ret >= 0) { + ret = read; /* did we bounce off eof? */ if (pos + left > inode->i_size) *checkeof = 1; - - /* zero trailing bytes (inside i_size) */ - if (left > 0 && pos < inode->i_size) { - if (pos + left > inode->i_size) - left = inode->i_size - pos; - - dout("zero tail %llu\n", left); - ceph_zero_page_vector_range(page_align + read, left, - pages); - read += left; - } } - if (ret >= 0) - ret = read; dout("striped_read returns %d\n", ret); return ret; } -- cgit v1.2.3-59-g8ed1b From ee7289bfadda5f4ef60884547ebc9989c8fb314a Mon Sep 17 00:00:00 2001 From: majianpeng Date: Wed, 21 Aug 2013 15:02:51 +0800 Subject: ceph: allow sync_read/write return partial successed size of read/write. For sync_read/write, it may do multi stripe operations.If one of those met erro, we return the former successed size rather than a error value. There is a exception for write-operation met -EOLDSNAPC.If this occur,we retry the whole write again. Signed-off-by: Jianpeng Ma --- fs/ceph/file.c | 4 +++- 1 file changed, 3 insertions(+), 1 deletion(-) (limited to 'fs') diff --git a/fs/ceph/file.c b/fs/ceph/file.c index 98b9035b2e81..20d0222c2e76 100644 --- a/fs/ceph/file.c +++ b/fs/ceph/file.c @@ -374,7 +374,7 @@ more: goto more; } - if (ret >= 0) { + if (read > 0) { ret = read; /* did we bounce off eof? */ if (pos + left > inode->i_size) @@ -612,6 +612,8 @@ out: if (check_caps) ceph_check_caps(ceph_inode(inode), CHECK_CAPS_AUTHONLY, NULL); + } else if (ret != -EOLDSNAPC && written > 0) { + ret = written; } return ret; } -- cgit v1.2.3-59-g8ed1b From 7d6e1f5461d0c16eb6aa8d226976995856d85e4e Mon Sep 17 00:00:00 2001 From: Sha Zhengju Date: Wed, 21 Aug 2013 16:27:34 +0800 Subject: ceph: use vfs __set_page_dirty_nobuffers interface instead of doing it inside filesystem Following we will begin to add memcg dirty page accounting around __set_page_dirty_{buffers,nobuffers} in vfs layer, so we'd better use vfs interface to avoid exporting those details to filesystems. Since vfs set_page_dirty() should be called under page lock, here we don't need elaborate codes to handle racy anymore, and two WARN_ON() are added to detect such exceptions. Thanks very much for Sage and Yan Zheng's coaching! I tested it in a two server's ceph environment that one is client and the other is mds/osd/mon, and run the following fsx test from xfstests: ./fsx 1MB -N 50000 -p 10000 -l 1048576 ./fsx 10MB -N 50000 -p 10000 -l 10485760 ./fsx 100MB -N 50000 -p 10000 -l 104857600 The fsx does lots of mmap-read/mmap-write/truncate operations and the tests completed successfully without triggering any of WARN_ON. Signed-off-by: Sha Zhengju Reviewed-by: Sage Weil --- fs/ceph/addr.c | 43 ++++++++++++++----------------------------- 1 file changed, 14 insertions(+), 29 deletions(-) (limited to 'fs') diff --git a/fs/ceph/addr.c b/fs/ceph/addr.c index cb78ce81d6a6..3bed7da38326 100644 --- a/fs/ceph/addr.c +++ b/fs/ceph/addr.c @@ -70,15 +70,16 @@ static int ceph_set_page_dirty(struct page *page) struct address_space *mapping = page->mapping; struct inode *inode; struct ceph_inode_info *ci; - int undo = 0; struct ceph_snap_context *snapc; + int ret; if (unlikely(!mapping)) return !TestSetPageDirty(page); - if (TestSetPageDirty(page)) { + if (PageDirty(page)) { dout("%p set_page_dirty %p idx %lu -- already dirty\n", mapping->host, page, page->index); + BUG_ON(!PagePrivate(page)); return 0; } @@ -107,35 +108,19 @@ static int ceph_set_page_dirty(struct page *page) snapc, snapc->seq, snapc->num_snaps); spin_unlock(&ci->i_ceph_lock); - /* now adjust page */ - spin_lock_irq(&mapping->tree_lock); - if (page->mapping) { /* Race with truncate? */ - WARN_ON_ONCE(!PageUptodate(page)); - account_page_dirtied(page, page->mapping); - radix_tree_tag_set(&mapping->page_tree, - page_index(page), PAGECACHE_TAG_DIRTY); - - /* - * Reference snap context in page->private. Also set - * PagePrivate so that we get invalidatepage callback. - */ - page->private = (unsigned long)snapc; - SetPagePrivate(page); - } else { - dout("ANON set_page_dirty %p (raced truncate?)\n", page); - undo = 1; - } - - spin_unlock_irq(&mapping->tree_lock); - - if (undo) - /* whoops, we failed to dirty the page */ - ceph_put_wrbuffer_cap_refs(ci, 1, snapc); + /* + * Reference snap context in page->private. Also set + * PagePrivate so that we get invalidatepage callback. + */ + BUG_ON(PagePrivate(page)); + page->private = (unsigned long)snapc; + SetPagePrivate(page); - __mark_inode_dirty(mapping->host, I_DIRTY_PAGES); + ret = __set_page_dirty_nobuffers(page); + WARN_ON(!PageLocked(page)); + WARN_ON(!page->mapping); - BUG_ON(!PageDirty(page)); - return 1; + return ret; } /* -- cgit v1.2.3-59-g8ed1b From da9803bc8812f5bd3b26baaa90e515b843c65ff7 Mon Sep 17 00:00:00 2001 From: David Howells Date: Wed, 21 Aug 2013 17:29:38 -0400 Subject: FS-Cache: Add interface to check consistency of a cached object Extend the fscache netfs API so that the netfs can ask as to whether a cache object is up to date with respect to its corresponding netfs object: int fscache_check_consistency(struct fscache_cookie *cookie) This will call back to the netfs to check whether the auxiliary data associated with a cookie is correct. It returns 0 if it is and -ESTALE if it isn't; it may also return -ENOMEM and -ERESTARTSYS. The backends now have to implement a mandatory operation pointer: int (*check_consistency)(struct fscache_object *object) that corresponds to the above API call. FS-Cache takes care of pinning the object and the cookie in memory and managing this call with respect to the object state. Original-author: Hongyi Jia Signed-off-by: David Howells cc: Hongyi Jia cc: Milosz Tanski --- Documentation/filesystems/caching/backend-api.txt | 9 +++ Documentation/filesystems/caching/netfs-api.txt | 17 ++++-- fs/fscache/cookie.c | 71 +++++++++++++++++++++++ fs/fscache/internal.h | 6 ++ fs/fscache/page.c | 55 ++++++++++-------- include/linux/fscache-cache.h | 4 ++ include/linux/fscache.h | 20 +++++++ 7 files changed, 154 insertions(+), 28 deletions(-) (limited to 'fs') diff --git a/Documentation/filesystems/caching/backend-api.txt b/Documentation/filesystems/caching/backend-api.txt index d78bab9622c6..277d1e810670 100644 --- a/Documentation/filesystems/caching/backend-api.txt +++ b/Documentation/filesystems/caching/backend-api.txt @@ -299,6 +299,15 @@ performed on the denizens of the cache. These are held in a structure of type: enough space in the cache to permit this. + (*) Check coherency state of an object [mandatory]: + + int (*check_consistency)(struct fscache_object *object) + + This method is called to have the cache check the saved auxiliary data of + the object against the netfs's idea of the state. 0 should be returned + if they're consistent and -ESTALE otherwise. -ENOMEM and -ERESTARTSYS + may also be returned. + (*) Update object [mandatory]: int (*update_object)(struct fscache_object *object) diff --git a/Documentation/filesystems/caching/netfs-api.txt b/Documentation/filesystems/caching/netfs-api.txt index 97e6c0ecc5ef..12b344251523 100644 --- a/Documentation/filesystems/caching/netfs-api.txt +++ b/Documentation/filesystems/caching/netfs-api.txt @@ -32,7 +32,7 @@ This document contains the following sections: (9) Setting the data file size (10) Page alloc/read/write (11) Page uncaching - (12) Index and data file update + (12) Index and data file consistency (13) Miscellaneous cookie operations (14) Cookie unregistration (15) Index invalidation @@ -690,9 +690,18 @@ written to the cache and for the cache to finish with the page generally. No error is returned. -========================== -INDEX AND DATA FILE UPDATE -========================== +=============================== +INDEX AND DATA FILE CONSISTENCY +=============================== + +To find out whether auxiliary data for an object is up to data within the +cache, the following function can be called: + + int fscache_check_consistency(struct fscache_cookie *cookie) + +This will call back to the netfs to check whether the auxiliary data associated +with a cookie is correct. It returns 0 if it is and -ESTALE if it isn't; it +may also return -ENOMEM and -ERESTARTSYS. To request an update of the index data for an index or other object, the following function should be called: diff --git a/fs/fscache/cookie.c b/fs/fscache/cookie.c index 0e91a3c9fdb2..318e8433527c 100644 --- a/fs/fscache/cookie.c +++ b/fs/fscache/cookie.c @@ -558,3 +558,74 @@ void __fscache_cookie_put(struct fscache_cookie *cookie) _leave(""); } + +/* + * check the consistency between the netfs inode and the backing cache + * + * NOTE: it only serves no-index type + */ +int __fscache_check_consistency(struct fscache_cookie *cookie) +{ + struct fscache_operation *op; + struct fscache_object *object; + int ret; + + _enter("%p,", cookie); + + ASSERTCMP(cookie->def->type, ==, FSCACHE_COOKIE_TYPE_DATAFILE); + + if (fscache_wait_for_deferred_lookup(cookie) < 0) + return -ERESTARTSYS; + + if (hlist_empty(&cookie->backing_objects)) + return 0; + + op = kzalloc(sizeof(*op), GFP_NOIO | __GFP_NOMEMALLOC | __GFP_NORETRY); + if (!op) + return -ENOMEM; + + fscache_operation_init(op, NULL, NULL); + op->flags = FSCACHE_OP_MYTHREAD | + (1 << FSCACHE_OP_WAITING); + + spin_lock(&cookie->lock); + + if (hlist_empty(&cookie->backing_objects)) + goto inconsistent; + object = hlist_entry(cookie->backing_objects.first, + struct fscache_object, cookie_link); + if (test_bit(FSCACHE_IOERROR, &object->cache->flags)) + goto inconsistent; + + op->debug_id = atomic_inc_return(&fscache_op_debug_id); + + atomic_inc(&cookie->n_active); + if (fscache_submit_op(object, op) < 0) + goto submit_failed; + + /* the work queue now carries its own ref on the object */ + spin_unlock(&cookie->lock); + + ret = fscache_wait_for_operation_activation(object, op, + NULL, NULL, NULL); + if (ret == 0) { + /* ask the cache to honour the operation */ + ret = object->cache->ops->check_consistency(op); + fscache_op_complete(op, false); + } else if (ret == -ENOBUFS) { + ret = 0; + } + + fscache_put_operation(op); + _leave(" = %d", ret); + return ret; + +submit_failed: + atomic_dec(&cookie->n_active); +inconsistent: + spin_unlock(&cookie->lock); + kfree(op); + _leave(" = -ESTALE"); + return -ESTALE; +} +EXPORT_SYMBOL(__fscache_check_consistency); diff --git a/fs/fscache/internal.h b/fs/fscache/internal.h index 12d505bedb5c..4226f6680b06 100644 --- a/fs/fscache/internal.h +++ b/fs/fscache/internal.h @@ -130,6 +130,12 @@ extern void fscache_operation_gc(struct work_struct *); /* * page.c */ +extern int fscache_wait_for_deferred_lookup(struct fscache_cookie *); +extern int fscache_wait_for_operation_activation(struct fscache_object *, + struct fscache_operation *, + atomic_t *, + atomic_t *, + void (*)(struct fscache_operation *)); extern void fscache_invalidate_writes(struct fscache_cookie *); /* diff --git a/fs/fscache/page.c b/fs/fscache/page.c index d479ab3c63e4..793e3d5ca4b5 100644 --- a/fs/fscache/page.c +++ b/fs/fscache/page.c @@ -278,7 +278,7 @@ static struct fscache_retrieval *fscache_alloc_retrieval( /* * wait for a deferred lookup to complete */ -static int fscache_wait_for_deferred_lookup(struct fscache_cookie *cookie) +int fscache_wait_for_deferred_lookup(struct fscache_cookie *cookie) { unsigned long jif; @@ -322,42 +322,46 @@ static void fscache_do_cancel_retrieval(struct fscache_operation *_op) /* * wait for an object to become active (or dead) */ -static int fscache_wait_for_retrieval_activation(struct fscache_object *object, - struct fscache_retrieval *op, - atomic_t *stat_op_waits, - atomic_t *stat_object_dead) +int fscache_wait_for_operation_activation(struct fscache_object *object, + struct fscache_operation *op, + atomic_t *stat_op_waits, + atomic_t *stat_object_dead, + void (*do_cancel)(struct fscache_operation *)) { int ret; - if (!test_bit(FSCACHE_OP_WAITING, &op->op.flags)) + if (!test_bit(FSCACHE_OP_WAITING, &op->flags)) goto check_if_dead; _debug(">>> WT"); - fscache_stat(stat_op_waits); - if (wait_on_bit(&op->op.flags, FSCACHE_OP_WAITING, + if (stat_op_waits) + fscache_stat(stat_op_waits); + if (wait_on_bit(&op->flags, FSCACHE_OP_WAITING, fscache_wait_bit_interruptible, TASK_INTERRUPTIBLE) != 0) { - ret = fscache_cancel_op(&op->op, fscache_do_cancel_retrieval); + ret = fscache_cancel_op(op, do_cancel); if (ret == 0) return -ERESTARTSYS; /* it's been removed from the pending queue by another party, * so we should get to run shortly */ - wait_on_bit(&op->op.flags, FSCACHE_OP_WAITING, + wait_on_bit(&op->flags, FSCACHE_OP_WAITING, fscache_wait_bit, TASK_UNINTERRUPTIBLE); } _debug("<<< GO"); check_if_dead: - if (op->op.state == FSCACHE_OP_ST_CANCELLED) { - fscache_stat(stat_object_dead); + if (op->state == FSCACHE_OP_ST_CANCELLED) { + if (stat_object_dead) + fscache_stat(stat_object_dead); _leave(" = -ENOBUFS [cancelled]"); return -ENOBUFS; } if (unlikely(fscache_object_is_dead(object))) { - pr_err("%s() = -ENOBUFS [obj dead %d]\n", __func__, op->op.state); - fscache_cancel_op(&op->op, fscache_do_cancel_retrieval); - fscache_stat(stat_object_dead); + pr_err("%s() = -ENOBUFS [obj dead %d]\n", __func__, op->state); + fscache_cancel_op(op, do_cancel); + if (stat_object_dead) + fscache_stat(stat_object_dead); return -ENOBUFS; } return 0; @@ -432,10 +436,11 @@ int __fscache_read_or_alloc_page(struct fscache_cookie *cookie, /* we wait for the operation to become active, and then process it * *here*, in this thread, and not in the thread pool */ - ret = fscache_wait_for_retrieval_activation( - object, op, + ret = fscache_wait_for_operation_activation( + object, &op->op, __fscache_stat(&fscache_n_retrieval_op_waits), - __fscache_stat(&fscache_n_retrievals_object_dead)); + __fscache_stat(&fscache_n_retrievals_object_dead), + fscache_do_cancel_retrieval); if (ret < 0) goto error; @@ -557,10 +562,11 @@ int __fscache_read_or_alloc_pages(struct fscache_cookie *cookie, /* we wait for the operation to become active, and then process it * *here*, in this thread, and not in the thread pool */ - ret = fscache_wait_for_retrieval_activation( - object, op, + ret = fscache_wait_for_operation_activation( + object, &op->op, __fscache_stat(&fscache_n_retrieval_op_waits), - __fscache_stat(&fscache_n_retrievals_object_dead)); + __fscache_stat(&fscache_n_retrievals_object_dead), + fscache_do_cancel_retrieval); if (ret < 0) goto error; @@ -658,10 +664,11 @@ int __fscache_alloc_page(struct fscache_cookie *cookie, fscache_stat(&fscache_n_alloc_ops); - ret = fscache_wait_for_retrieval_activation( - object, op, + ret = fscache_wait_for_operation_activation( + object, &op->op, __fscache_stat(&fscache_n_alloc_op_waits), - __fscache_stat(&fscache_n_allocs_object_dead)); + __fscache_stat(&fscache_n_allocs_object_dead), + fscache_do_cancel_retrieval); if (ret < 0) goto error; diff --git a/include/linux/fscache-cache.h b/include/linux/fscache-cache.h index a9ff9a36b86d..7823e9ef995e 100644 --- a/include/linux/fscache-cache.h +++ b/include/linux/fscache-cache.h @@ -251,6 +251,10 @@ struct fscache_cache_ops { /* unpin an object in the cache */ void (*unpin_object)(struct fscache_object *object); + /* check the consistency between the backing cache and the FS-Cache + * cookie */ + bool (*check_consistency)(struct fscache_operation *op); + /* store the updated auxiliary data on an object */ void (*update_object)(struct fscache_object *object); diff --git a/include/linux/fscache.h b/include/linux/fscache.h index 7a086235da4b..d984aff32a11 100644 --- a/include/linux/fscache.h +++ b/include/linux/fscache.h @@ -183,6 +183,7 @@ extern struct fscache_cookie *__fscache_acquire_cookie( const struct fscache_cookie_def *, void *); extern void __fscache_relinquish_cookie(struct fscache_cookie *, int); +extern int __fscache_check_consistency(struct fscache_cookie *); extern void __fscache_update_cookie(struct fscache_cookie *); extern int __fscache_attr_changed(struct fscache_cookie *); extern void __fscache_invalidate(struct fscache_cookie *); @@ -325,6 +326,25 @@ void fscache_relinquish_cookie(struct fscache_cookie *cookie, int retire) __fscache_relinquish_cookie(cookie, retire); } +/** + * fscache_check_consistency - Request that if the cache is updated + * @cookie: The cookie representing the cache object + * + * Request an consistency check from fscache, which passes the request + * to the backing cache. + * + * Returns 0 if consistent and -ESTALE if inconsistent. May also + * return -ENOMEM and -ERESTARTSYS. + */ +static inline +int fscache_check_consistency(struct fscache_cookie *cookie) +{ + if (fscache_cookie_valid(cookie)) + return __fscache_check_consistency(cookie); + else + return 0; +} + /** * fscache_update_cookie - Request that a cache object be updated * @cookie: The cookie representing the cache object -- cgit v1.2.3-59-g8ed1b From 5002d7bef81c9646bbb06fb57db4a100aa5a57c5 Mon Sep 17 00:00:00 2001 From: David Howells Date: Wed, 21 Aug 2013 17:29:21 -0400 Subject: CacheFiles: Implement interface to check cache consistency Implement the FS-Cache interface to check the consistency of a cache object in CacheFiles. Original-author: Hongyi Jia Signed-off-by: David Howells cc: Hongyi Jia cc: Milosz Tanski --- fs/cachefiles/interface.c | 26 ++++++++++++++++++++++++++ fs/cachefiles/internal.h | 1 + fs/cachefiles/xattr.c | 36 ++++++++++++++++++++++++++++++++++++ 3 files changed, 63 insertions(+) (limited to 'fs') diff --git a/fs/cachefiles/interface.c b/fs/cachefiles/interface.c index d4c1206af9fc..43eb5592cdea 100644 --- a/fs/cachefiles/interface.c +++ b/fs/cachefiles/interface.c @@ -377,6 +377,31 @@ static void cachefiles_sync_cache(struct fscache_cache *_cache) ret); } +/* + * check if the backing cache is updated to FS-Cache + * - called by FS-Cache when evaluates if need to invalidate the cache + */ +static bool cachefiles_check_consistency(struct fscache_operation *op) +{ + struct cachefiles_object *object; + struct cachefiles_cache *cache; + const struct cred *saved_cred; + int ret; + + _enter("{OBJ%x}", op->object->debug_id); + + object = container_of(op->object, struct cachefiles_object, fscache); + cache = container_of(object->fscache.cache, + struct cachefiles_cache, cache); + + cachefiles_begin_secure(cache, &saved_cred); + ret = cachefiles_check_auxdata(object); + cachefiles_end_secure(cache, saved_cred); + + _leave(" = %d", ret); + return ret; +} + /* * notification the attributes on an object have changed * - called with reads/writes excluded by FS-Cache @@ -522,4 +547,5 @@ const struct fscache_cache_ops cachefiles_cache_ops = { .write_page = cachefiles_write_page, .uncache_page = cachefiles_uncache_page, .dissociate_pages = cachefiles_dissociate_pages, + .check_consistency = cachefiles_check_consistency, }; diff --git a/fs/cachefiles/internal.h b/fs/cachefiles/internal.h index 49382519907a..5349473df1b1 100644 --- a/fs/cachefiles/internal.h +++ b/fs/cachefiles/internal.h @@ -235,6 +235,7 @@ extern int cachefiles_set_object_xattr(struct cachefiles_object *object, struct cachefiles_xattr *auxdata); extern int cachefiles_update_object_xattr(struct cachefiles_object *object, struct cachefiles_xattr *auxdata); +extern int cachefiles_check_auxdata(struct cachefiles_object *object); extern int cachefiles_check_object_xattr(struct cachefiles_object *object, struct cachefiles_xattr *auxdata); extern int cachefiles_remove_object_xattr(struct cachefiles_cache *cache, diff --git a/fs/cachefiles/xattr.c b/fs/cachefiles/xattr.c index 2476e5162609..34c88b83e39f 100644 --- a/fs/cachefiles/xattr.c +++ b/fs/cachefiles/xattr.c @@ -156,6 +156,42 @@ int cachefiles_update_object_xattr(struct cachefiles_object *object, return ret; } +/* + * check the consistency between the backing cache and the FS-Cache cookie + */ +int cachefiles_check_auxdata(struct cachefiles_object *object) +{ + struct cachefiles_xattr *auxbuf; + struct dentry *dentry = object->dentry; + unsigned int dlen; + int ret; + + ASSERT(dentry); + ASSERT(dentry->d_inode); + ASSERT(object->fscache.cookie->def->check_aux); + + auxbuf = kmalloc(sizeof(struct cachefiles_xattr) + 512, GFP_KERNEL); + if (!auxbuf) + return -ENOMEM; + + auxbuf->len = vfs_getxattr(dentry, cachefiles_xattr_cache, + &auxbuf->type, 512 + 1); + if (auxbuf->len < 1) + return -ESTALE; + + if (auxbuf->type != object->fscache.cookie->def->type) + return -ESTALE; + + dlen = auxbuf->len - 1; + ret = fscache_check_aux(&object->fscache, &auxbuf->data, dlen); + + kfree(auxbuf); + if (ret != FSCACHE_CHECKAUX_OKAY) + return -ESTALE; + + return 0; +} + /* * check the state xattr on a cache file * - return -ESTALE if the object should be deleted -- cgit v1.2.3-59-g8ed1b From 5a6f282a2052bb13171b53f03b34501cf72c33f1 Mon Sep 17 00:00:00 2001 From: Milosz Tanski Date: Wed, 21 Aug 2013 17:30:11 -0400 Subject: fscache: Netfs function for cleanup post readpages Currently the fscache code expect the netfs to call fscache_readpages_or_alloc inside the aops readpages callback. It marks all the pages in the list provided by readahead with PG_private_2. In the cases that the netfs fails to read all the pages (which is legal) it ends up returning to the readahead and triggering a BUG. This happens because the page list still contains marked pages. This patch implements a simple fscache_readpages_cancel function that the netfs should call before returning from readpages. It will revoke the pages from the underlying cache backend and unmark them. The problem was originally worked out in the Ceph devel tree, but it also occurs in CIFS. It appears that NFS, AFS and 9P are okay as read_cache_pages() will clean up the unprocessed pages in the case of an error. This can be used to address the following oops: [12410647.597278] BUG: Bad page state in process petabucket pfn:3d504e [12410647.597292] page:ffffea000f541380 count:0 mapcount:0 mapping: (null) index:0x0 [12410647.597298] page flags: 0x200000000001000(private_2) ... [12410647.597334] Call Trace: [12410647.597345] [] dump_stack+0x19/0x1b [12410647.597356] [] bad_page+0xc7/0x120 [12410647.597359] [] free_pages_prepare+0x10e/0x120 [12410647.597361] [] free_hot_cold_page+0x40/0x170 [12410647.597363] [] __put_single_page+0x27/0x30 [12410647.597365] [] put_page+0x25/0x40 [12410647.597376] [] ceph_readpages+0x2e9/0x6e0 [ceph] [12410647.597379] [] __do_page_cache_readahead+0x1af/0x260 [12410647.597382] [] ra_submit+0x21/0x30 [12410647.597384] [] filemap_fault+0x254/0x490 [12410647.597387] [] __do_fault+0x6f/0x4e0 [12410647.597391] [] ? __switch_to+0x16d/0x4a0 [12410647.597395] [] ? finish_task_switch+0x5a/0xc0 [12410647.597398] [] handle_pte_fault+0xf6/0x930 [12410647.597401] [] ? pte_mfn_to_pfn+0x93/0x110 [12410647.597403] [] ? xen_pmd_val+0xe/0x10 [12410647.597405] [] ? __raw_callee_save_xen_pmd_val+0x11/0x1e [12410647.597407] [] handle_mm_fault+0x251/0x370 [12410647.597411] [] ? call_rwsem_down_read_failed+0x14/0x30 [12410647.597414] [] __do_page_fault+0x1aa/0x550 [12410647.597418] [] ? up_write+0x1d/0x20 [12410647.597422] [] ? vm_mmap_pgoff+0xbc/0xe0 [12410647.597425] [] ? SyS_mmap_pgoff+0xd8/0x240 [12410647.597427] [] do_page_fault+0xe/0x10 [12410647.597431] [] page_fault+0x28/0x30 Signed-off-by: Milosz Tanski Signed-off-by: David Howells --- Documentation/filesystems/caching/netfs-api.txt | 18 +++++++++++++++++- fs/fscache/page.c | 16 ++++++++++++++++ include/linux/fscache.h | 22 ++++++++++++++++++++++ 3 files changed, 55 insertions(+), 1 deletion(-) (limited to 'fs') diff --git a/Documentation/filesystems/caching/netfs-api.txt b/Documentation/filesystems/caching/netfs-api.txt index 26c1dd5a6a21..11a0a40ce445 100644 --- a/Documentation/filesystems/caching/netfs-api.txt +++ b/Documentation/filesystems/caching/netfs-api.txt @@ -499,7 +499,7 @@ Else if there's a copy of the page resident in the cache: (*) An argument that's 0 on success or negative for an error code. If an error occurs, it should be assumed that the page contains no usable - data. + data. fscache_readpages_cancel() may need to be called. end_io_func() will be called in process context if the read is results in an error, but it might be called in interrupt context if the read is @@ -623,6 +623,22 @@ some of the pages being read and some being allocated. Those pages will have been marked appropriately and will need uncaching. +CANCELLATION OF UNREAD PAGES +---------------------------- + +If one or more pages are passed to fscache_read_or_alloc_pages() but not then +read from the cache and also not read from the underlying filesystem then +those pages will need to have any marks and reservations removed. This can be +done by calling: + + void fscache_readpages_cancel(struct fscache_cookie *cookie, + struct list_head *pages); + +prior to returning to the caller. The cookie argument should be as passed to +fscache_read_or_alloc_pages(). Every page in the pages list will be examined +and any that have PG_fscache set will be uncached. + + ============== PAGE UNCACHING ============== diff --git a/fs/fscache/page.c b/fs/fscache/page.c index 793e3d5ca4b5..8702b732109a 100644 --- a/fs/fscache/page.c +++ b/fs/fscache/page.c @@ -700,6 +700,22 @@ nobufs: } EXPORT_SYMBOL(__fscache_alloc_page); +/* + * Unmark pages allocate in the readahead code path (via: + * fscache_readpages_or_alloc) after delegating to the base filesystem + */ +void __fscache_readpages_cancel(struct fscache_cookie *cookie, + struct list_head *pages) +{ + struct page *page; + + list_for_each_entry(page, pages, lru) { + if (PageFsCache(page)) + __fscache_uncache_page(cookie, page); + } +} +EXPORT_SYMBOL(__fscache_readpages_cancel); + /* * release a write op reference */ diff --git a/include/linux/fscache.h b/include/linux/fscache.h index d984aff32a11..19b46458e4e8 100644 --- a/include/linux/fscache.h +++ b/include/linux/fscache.h @@ -209,6 +209,8 @@ extern bool __fscache_maybe_release_page(struct fscache_cookie *, struct page *, gfp_t); extern void __fscache_uncache_all_inode_pages(struct fscache_cookie *, struct inode *); +extern void __fscache_readpages_cancel(struct fscache_cookie *cookie, + struct list_head *pages); /** * fscache_register_netfs - Register a filesystem as desiring caching services @@ -589,6 +591,26 @@ int fscache_alloc_page(struct fscache_cookie *cookie, return -ENOBUFS; } +/** + * fscache_readpages_cancel - Cancel read/alloc on pages + * @cookie: The cookie representing the inode's cache object. + * @pages: The netfs pages that we canceled write on in readpages() + * + * Uncache/unreserve the pages reserved earlier in readpages() via + * fscache_readpages_or_alloc() and similar. In most successful caches in + * readpages() this doesn't do anything. In cases when the underlying netfs's + * readahead failed we need to clean up the pagelist (unmark and uncache). + * + * This function may sleep as it may have to clean up disk state. + */ +static inline +void fscache_readpages_cancel(struct fscache_cookie *cookie, + struct list_head *pages) +{ + if (fscache_cookie_valid(cookie)) + __fscache_readpages_cancel(cookie, pages); +} + /** * fscache_write_page - Request storage of a page in the cache * @cookie: The cookie representing the cache object -- cgit v1.2.3-59-g8ed1b From 99ccbd229cf7453206bc858e795ec1f0345ff258 Mon Sep 17 00:00:00 2001 From: Milosz Tanski Date: Wed, 21 Aug 2013 17:29:54 -0400 Subject: ceph: use fscache as a local presisent cache Adding support for fscache to the Ceph filesystem. This would bring it to on par with some of the other network filesystems in Linux (like NFS, AFS, etc...) In order to mount the filesystem with fscache the 'fsc' mount option must be passed. Signed-off-by: Milosz Tanski Signed-off-by: Sage Weil --- fs/ceph/Kconfig | 9 ++ fs/ceph/Makefile | 1 + fs/ceph/addr.c | 37 +++++- fs/ceph/cache.c | 393 +++++++++++++++++++++++++++++++++++++++++++++++++++++++ fs/ceph/cache.h | 138 +++++++++++++++++++ fs/ceph/caps.c | 19 ++- fs/ceph/file.c | 17 +++ fs/ceph/inode.c | 14 +- fs/ceph/super.c | 35 ++++- fs/ceph/super.h | 16 +++ 10 files changed, 666 insertions(+), 13 deletions(-) create mode 100644 fs/ceph/cache.c create mode 100644 fs/ceph/cache.h (limited to 'fs') diff --git a/fs/ceph/Kconfig b/fs/ceph/Kconfig index 49bc78243db9..ac9a2ef5bb9b 100644 --- a/fs/ceph/Kconfig +++ b/fs/ceph/Kconfig @@ -16,3 +16,12 @@ config CEPH_FS If unsure, say N. +if CEPH_FS +config CEPH_FSCACHE + bool "Enable Ceph client caching support" + depends on CEPH_FS=m && FSCACHE || CEPH_FS=y && FSCACHE=y + help + Choose Y here to enable persistent, read-only local + caching support for Ceph clients using FS-Cache + +endif diff --git a/fs/ceph/Makefile b/fs/ceph/Makefile index bd352125e829..32e30106a2f0 100644 --- a/fs/ceph/Makefile +++ b/fs/ceph/Makefile @@ -9,3 +9,4 @@ ceph-y := super.o inode.o dir.o file.o locks.o addr.o ioctl.o \ mds_client.o mdsmap.o strings.o ceph_frag.o \ debugfs.o +ceph-$(CONFIG_CEPH_FSCACHE) += cache.o diff --git a/fs/ceph/addr.c b/fs/ceph/addr.c index 3bed7da38326..3a21a7cbc21c 100644 --- a/fs/ceph/addr.c +++ b/fs/ceph/addr.c @@ -11,6 +11,7 @@ #include "super.h" #include "mds_client.h" +#include "cache.h" #include /* @@ -144,6 +145,11 @@ static void ceph_invalidatepage(struct page *page, unsigned int offset, return; } + ceph_invalidate_fscache_page(inode, page); + + if (!PagePrivate(page)) + return; + /* * We can get non-dirty pages here due to races between * set_page_dirty and truncate_complete_page; just spit out a @@ -163,14 +169,17 @@ static void ceph_invalidatepage(struct page *page, unsigned int offset, ClearPagePrivate(page); } -/* just a sanity check */ static int ceph_releasepage(struct page *page, gfp_t g) { struct inode *inode = page->mapping ? page->mapping->host : NULL; dout("%p releasepage %p idx %lu\n", inode, page, page->index); WARN_ON(PageDirty(page)); - WARN_ON(PagePrivate(page)); - return 0; + + /* Can we release the page from the cache? */ + if (!ceph_release_fscache_page(page, g)) + return 0; + + return !PagePrivate(page); } /* @@ -180,11 +189,16 @@ static int readpage_nounlock(struct file *filp, struct page *page) { struct inode *inode = file_inode(filp); struct ceph_inode_info *ci = ceph_inode(inode); - struct ceph_osd_client *osdc = + struct ceph_osd_client *osdc = &ceph_inode_to_client(inode)->client->osdc; int err = 0; u64 len = PAGE_CACHE_SIZE; + err = ceph_readpage_from_fscache(inode, page); + + if (err == 0) + goto out; + dout("readpage inode %p file %p page %p index %lu\n", inode, filp, page, page->index); err = ceph_osdc_readpages(osdc, ceph_vino(inode), &ci->i_layout, @@ -202,6 +216,9 @@ static int readpage_nounlock(struct file *filp, struct page *page) } SetPageUptodate(page); + if (err == 0) + ceph_readpage_to_fscache(inode, page); + out: return err < 0 ? err : 0; } @@ -244,6 +261,7 @@ static void finish_read(struct ceph_osd_request *req, struct ceph_msg *msg) page->index); flush_dcache_page(page); SetPageUptodate(page); + ceph_readpage_to_fscache(inode, page); unlock_page(page); page_cache_release(page); bytes -= PAGE_CACHE_SIZE; @@ -313,7 +331,7 @@ static int start_read(struct inode *inode, struct list_head *page_list, int max) page = list_entry(page_list->prev, struct page, lru); BUG_ON(PageLocked(page)); list_del(&page->lru); - + dout("start_read %p adding %p idx %lu\n", inode, page, page->index); if (add_to_page_cache_lru(page, &inode->i_data, page->index, @@ -360,6 +378,12 @@ static int ceph_readpages(struct file *file, struct address_space *mapping, int rc = 0; int max = 0; + rc = ceph_readpages_from_fscache(mapping->host, mapping, page_list, + &nr_pages); + + if (rc == 0) + goto out; + if (fsc->mount_options->rsize >= PAGE_CACHE_SIZE) max = (fsc->mount_options->rsize + PAGE_CACHE_SIZE - 1) >> PAGE_SHIFT; @@ -479,6 +503,8 @@ static int writepage_nounlock(struct page *page, struct writeback_control *wbc) CONGESTION_ON_THRESH(fsc->mount_options->congestion_kb)) set_bdi_congested(&fsc->backing_dev_info, BLK_RW_ASYNC); + ceph_readpage_to_fscache(inode, page); + set_page_writeback(page); err = ceph_osdc_writepages(osdc, ceph_vino(inode), &ci->i_layout, snapc, @@ -534,7 +560,6 @@ static void ceph_release_pages(struct page **pages, int num) pagevec_release(&pvec); } - /* * async writeback completion handler. * diff --git a/fs/ceph/cache.c b/fs/ceph/cache.c new file mode 100644 index 000000000000..5c413ecf1f15 --- /dev/null +++ b/fs/ceph/cache.c @@ -0,0 +1,393 @@ +/* + * Ceph cache definitions. + * + * Copyright (C) 2013 by Adfin Solutions, Inc. All Rights Reserved. + * Written by Milosz Tanski (milosz@adfin.com) + * + * This program is free software; you can redistribute it and/or modify + * it under the terms of the GNU General Public License version 2 + * as published by the Free Software Foundation. + * + * This program is distributed in the hope that it will be useful, + * but WITHOUT ANY WARRANTY; without even the implied warranty of + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the + * GNU General Public License for more details. + * + * You should have received a copy of the GNU General Public License + * along with this program; if not, write to: + * Free Software Foundation + * 51 Franklin Street, Fifth Floor + * Boston, MA 02111-1301 USA + * + */ + +#include + +#include "super.h" +#include "cache.h" + +struct ceph_aux_inode { + struct timespec mtime; + loff_t size; +}; + +struct fscache_netfs ceph_cache_netfs = { + .name = "ceph", + .version = 0, +}; + +static uint16_t ceph_fscache_session_get_key(const void *cookie_netfs_data, + void *buffer, uint16_t maxbuf) +{ + const struct ceph_fs_client* fsc = cookie_netfs_data; + uint16_t klen; + + klen = sizeof(fsc->client->fsid); + if (klen > maxbuf) + return 0; + + memcpy(buffer, &fsc->client->fsid, klen); + return klen; +} + +static const struct fscache_cookie_def ceph_fscache_fsid_object_def = { + .name = "CEPH.fsid", + .type = FSCACHE_COOKIE_TYPE_INDEX, + .get_key = ceph_fscache_session_get_key, +}; + +int ceph_fscache_register() +{ + return fscache_register_netfs(&ceph_cache_netfs); +} + +void ceph_fscache_unregister() +{ + fscache_unregister_netfs(&ceph_cache_netfs); +} + +int ceph_fscache_register_fs(struct ceph_fs_client* fsc) +{ + fsc->fscache = fscache_acquire_cookie(ceph_cache_netfs.primary_index, + &ceph_fscache_fsid_object_def, + fsc); + + if (fsc->fscache == NULL) { + pr_err("Unable to resgister fsid: %p fscache cookie", fsc); + return 0; + } + + fsc->revalidate_wq = alloc_workqueue("ceph-revalidate", 0, 1); + if (fsc->revalidate_wq == NULL) + return -ENOMEM; + + return 0; +} + +static uint16_t ceph_fscache_inode_get_key(const void *cookie_netfs_data, + void *buffer, uint16_t maxbuf) +{ + const struct ceph_inode_info* ci = cookie_netfs_data; + uint16_t klen; + + /* use ceph virtual inode (id + snaphot) */ + klen = sizeof(ci->i_vino); + if (klen > maxbuf) + return 0; + + memcpy(buffer, &ci->i_vino, klen); + return klen; +} + +static uint16_t ceph_fscache_inode_get_aux(const void *cookie_netfs_data, + void *buffer, uint16_t bufmax) +{ + struct ceph_aux_inode aux; + const struct ceph_inode_info* ci = cookie_netfs_data; + const struct inode* inode = &ci->vfs_inode; + + memset(&aux, 0, sizeof(aux)); + aux.mtime = inode->i_mtime; + aux.size = inode->i_size; + + memcpy(buffer, &aux, sizeof(aux)); + + return sizeof(aux); +} + +static void ceph_fscache_inode_get_attr(const void *cookie_netfs_data, + uint64_t *size) +{ + const struct ceph_inode_info* ci = cookie_netfs_data; + const struct inode* inode = &ci->vfs_inode; + + *size = inode->i_size; +} + +static enum fscache_checkaux ceph_fscache_inode_check_aux( + void *cookie_netfs_data, const void *data, uint16_t dlen) +{ + struct ceph_aux_inode aux; + struct ceph_inode_info* ci = cookie_netfs_data; + struct inode* inode = &ci->vfs_inode; + + if (dlen != sizeof(aux)) + return FSCACHE_CHECKAUX_OBSOLETE; + + memset(&aux, 0, sizeof(aux)); + aux.mtime = inode->i_mtime; + aux.size = inode->i_size; + + if (memcmp(data, &aux, sizeof(aux)) != 0) + return FSCACHE_CHECKAUX_OBSOLETE; + + dout("ceph inode 0x%p cached okay", ci); + return FSCACHE_CHECKAUX_OKAY; +} + +static void ceph_fscache_inode_now_uncached(void* cookie_netfs_data) +{ + struct ceph_inode_info* ci = cookie_netfs_data; + struct pagevec pvec; + pgoff_t first; + int loop, nr_pages; + + pagevec_init(&pvec, 0); + first = 0; + + dout("ceph inode 0x%p now uncached", ci); + + while (1) { + nr_pages = pagevec_lookup(&pvec, ci->vfs_inode.i_mapping, first, + PAGEVEC_SIZE - pagevec_count(&pvec)); + + if (!nr_pages) + break; + + for (loop = 0; loop < nr_pages; loop++) + ClearPageFsCache(pvec.pages[loop]); + + first = pvec.pages[nr_pages - 1]->index + 1; + + pvec.nr = nr_pages; + pagevec_release(&pvec); + cond_resched(); + } +} + +static const struct fscache_cookie_def ceph_fscache_inode_object_def = { + .name = "CEPH.inode", + .type = FSCACHE_COOKIE_TYPE_DATAFILE, + .get_key = ceph_fscache_inode_get_key, + .get_attr = ceph_fscache_inode_get_attr, + .get_aux = ceph_fscache_inode_get_aux, + .check_aux = ceph_fscache_inode_check_aux, + .now_uncached = ceph_fscache_inode_now_uncached, +}; + +void ceph_fscache_register_inode_cookie(struct ceph_fs_client* fsc, + struct ceph_inode_info* ci) +{ + struct inode* inode = &ci->vfs_inode; + + /* No caching for filesystem */ + if (fsc->fscache == NULL) + return; + + /* Only cache for regular files that are read only */ + if ((ci->vfs_inode.i_mode & S_IFREG) == 0) + return; + + /* Avoid multiple racing open requests */ + mutex_lock(&inode->i_mutex); + + if (ci->fscache) + goto done; + + ci->fscache = fscache_acquire_cookie(fsc->fscache, + &ceph_fscache_inode_object_def, + ci); +done: + mutex_unlock(&inode->i_mutex); + +} + +void ceph_fscache_unregister_inode_cookie(struct ceph_inode_info* ci) +{ + struct fscache_cookie* cookie; + + if ((cookie = ci->fscache) == NULL) + return; + + ci->fscache = NULL; + + fscache_uncache_all_inode_pages(cookie, &ci->vfs_inode); + fscache_relinquish_cookie(cookie, 0); +} + +static void ceph_vfs_readpage_complete(struct page *page, void *data, int error) +{ + if (!error) + SetPageUptodate(page); +} + +static void ceph_vfs_readpage_complete_unlock(struct page *page, void *data, int error) +{ + if (!error) + SetPageUptodate(page); + + unlock_page(page); +} + +static inline int cache_valid(struct ceph_inode_info *ci) +{ + return ((ceph_caps_issued(ci) & CEPH_CAP_FILE_CACHE) && + (ci->i_fscache_gen == ci->i_rdcache_gen)); +} + + +/* Atempt to read from the fscache, + * + * This function is called from the readpage_nounlock context. DO NOT attempt to + * unlock the page here (or in the callback). + */ +int ceph_readpage_from_fscache(struct inode *inode, struct page *page) +{ + struct ceph_inode_info *ci = ceph_inode(inode); + int ret; + + if (!cache_valid(ci)) + return -ENOBUFS; + + ret = fscache_read_or_alloc_page(ci->fscache, page, + ceph_vfs_readpage_complete, NULL, + GFP_KERNEL); + + switch (ret) { + case 0: /* Page found */ + dout("page read submitted\n"); + return 0; + case -ENOBUFS: /* Pages were not found, and can't be */ + case -ENODATA: /* Pages were not found */ + dout("page/inode not in cache\n"); + return ret; + default: + dout("%s: unknown error ret = %i\n", __func__, ret); + return ret; + } +} + +int ceph_readpages_from_fscache(struct inode *inode, + struct address_space *mapping, + struct list_head *pages, + unsigned *nr_pages) +{ + struct ceph_inode_info *ci = ceph_inode(inode); + int ret; + + if (!cache_valid(ci)) + return -ENOBUFS; + + ret = fscache_read_or_alloc_pages(ci->fscache, mapping, pages, nr_pages, + ceph_vfs_readpage_complete_unlock, + NULL, mapping_gfp_mask(mapping)); + + switch (ret) { + case 0: /* All pages found */ + dout("all-page read submitted\n"); + return 0; + case -ENOBUFS: /* Some pages were not found, and can't be */ + case -ENODATA: /* some pages were not found */ + dout("page/inode not in cache\n"); + return ret; + default: + dout("%s: unknown error ret = %i\n", __func__, ret); + return ret; + } +} + +void ceph_readpage_to_fscache(struct inode *inode, struct page *page) +{ + struct ceph_inode_info *ci = ceph_inode(inode); + int ret; + + if (!cache_valid(ci)) + return; + + ret = fscache_write_page(ci->fscache, page, GFP_KERNEL); + if (ret) + fscache_uncache_page(ci->fscache, page); +} + +void ceph_invalidate_fscache_page(struct inode* inode, struct page *page) +{ + struct ceph_inode_info *ci = ceph_inode(inode); + + fscache_wait_on_page_write(ci->fscache, page); + fscache_uncache_page(ci->fscache, page); +} + +void ceph_fscache_unregister_fs(struct ceph_fs_client* fsc) +{ + if (fsc->revalidate_wq) + destroy_workqueue(fsc->revalidate_wq); + + fscache_relinquish_cookie(fsc->fscache, 0); + fsc->fscache = NULL; +} + +static void ceph_revalidate_work(struct work_struct *work) +{ + int issued; + u32 orig_gen; + struct ceph_inode_info *ci = container_of(work, struct ceph_inode_info, + i_revalidate_work); + struct inode *inode = &ci->vfs_inode; + + spin_lock(&ci->i_ceph_lock); + issued = __ceph_caps_issued(ci, NULL); + orig_gen = ci->i_rdcache_gen; + spin_unlock(&ci->i_ceph_lock); + + if (!(issued & CEPH_CAP_FILE_CACHE)) { + dout("revalidate_work lost cache before validation %p\n", + inode); + goto out; + } + + if (!fscache_check_consistency(ci->fscache)) + fscache_invalidate(ci->fscache); + + spin_lock(&ci->i_ceph_lock); + /* Update the new valid generation (backwards sanity check too) */ + if (orig_gen > ci->i_fscache_gen) { + ci->i_fscache_gen = orig_gen; + } + spin_unlock(&ci->i_ceph_lock); + +out: + iput(&ci->vfs_inode); +} + +void ceph_queue_revalidate(struct inode *inode) +{ + struct ceph_inode_info *ci = ceph_inode(inode); + + ihold(inode); + + if (queue_work(ceph_sb_to_client(inode->i_sb)->revalidate_wq, + &ci->i_revalidate_work)) { + dout("ceph_queue_revalidate %p\n", inode); + } else { + dout("ceph_queue_revalidate %p failed\n)", inode); + iput(inode); + } +} + +void ceph_fscache_inode_init(struct ceph_inode_info *ci) +{ + ci->fscache = NULL; + /* The first load is verifed cookie open time */ + ci->i_fscache_gen = 1; + INIT_WORK(&ci->i_revalidate_work, ceph_revalidate_work); +} diff --git a/fs/ceph/cache.h b/fs/ceph/cache.h new file mode 100644 index 000000000000..0ea95cb7f389 --- /dev/null +++ b/fs/ceph/cache.h @@ -0,0 +1,138 @@ +/* + * Ceph cache definitions. + * + * Copyright (C) 2013 by Adfin Solutions, Inc. All Rights Reserved. + * Written by Milosz Tanski (milosz@adfin.com) + * + * This program is free software; you can redistribute it and/or modify + * it under the terms of the GNU General Public License version 2 + * as published by the Free Software Foundation. + * + * This program is distributed in the hope that it will be useful, + * but WITHOUT ANY WARRANTY; without even the implied warranty of + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the + * GNU General Public License for more details. + * + * You should have received a copy of the GNU General Public License + * along with this program; if not, write to: + * Free Software Foundation + * 51 Franklin Street, Fifth Floor + * Boston, MA 02111-1301 USA + * + */ + +#ifndef _CEPH_CACHE_H +#define _CEPH_CACHE_H + +#ifdef CONFIG_CEPH_FSCACHE + +int ceph_fscache_register(void); +void ceph_fscache_unregister(void); + +int ceph_fscache_register_fs(struct ceph_fs_client* fsc); +void ceph_fscache_unregister_fs(struct ceph_fs_client* fsc); + +void ceph_fscache_inode_init(struct ceph_inode_info *ci); +void ceph_fscache_register_inode_cookie(struct ceph_fs_client* fsc, + struct ceph_inode_info* ci); +void ceph_fscache_unregister_inode_cookie(struct ceph_inode_info* ci); + +int ceph_readpage_from_fscache(struct inode *inode, struct page *page); +int ceph_readpages_from_fscache(struct inode *inode, + struct address_space *mapping, + struct list_head *pages, + unsigned *nr_pages); +void ceph_readpage_to_fscache(struct inode *inode, struct page *page); +void ceph_invalidate_fscache_page(struct inode* inode, struct page *page); +void ceph_queue_revalidate(struct inode *inode); + +static inline void ceph_fscache_invalidate(struct inode *inode) +{ + fscache_invalidate(ceph_inode(inode)->fscache); +} + +static inline int ceph_release_fscache_page(struct page *page, gfp_t gfp) +{ + struct inode* inode = page->mapping->host; + struct ceph_inode_info *ci = ceph_inode(inode); + return fscache_maybe_release_page(ci->fscache, page, gfp); +} + +#else + +static inline int ceph_fscache_register(void) +{ + return 0; +} + +static inline void ceph_fscache_unregister(void) +{ +} + +static inline int ceph_fscache_register_fs(struct ceph_fs_client* fsc) +{ + return 0; +} + +static inline void ceph_fscache_unregister_fs(struct ceph_fs_client* fsc) +{ +} + +static inline void ceph_fscache_inode_init(struct ceph_inode_info *ci) +{ +} + +static inline void ceph_fscache_register_inode_cookie(struct ceph_fs_client* parent_fsc, + struct ceph_inode_info* ci) +{ +} + +static inline void ceph_fscache_unregister_inode_cookie(struct ceph_inode_info* ci) +{ +} + +static inline int ceph_readpage_from_fscache(struct inode* inode, + struct page *page) +{ + return -ENOBUFS; +} + +static inline int ceph_readpages_from_fscache(struct inode *inode, + struct address_space *mapping, + struct list_head *pages, + unsigned *nr_pages) +{ + return -ENOBUFS; +} + +static inline void ceph_readpage_to_fscache(struct inode *inode, + struct page *page) +{ +} + +static inline void ceph_fscache_invalidate(struct inode *inode) +{ +} + +static inline void ceph_invalidate_fscache_page(struct inode *inode, + struct page *page) +{ +} + +static inline int ceph_release_fscache_page(struct page *page, gfp_t gfp) +{ + return 1; +} + +static inline void ceph_fscache_readpages_cancel(struct inode *inode, + struct list_head *pages) +{ +} + +static inline void ceph_queue_revalidate(struct inode *inode) +{ +} + +#endif + +#endif diff --git a/fs/ceph/caps.c b/fs/ceph/caps.c index 5a26bc1dd799..7b451eb7d123 100644 --- a/fs/ceph/caps.c +++ b/fs/ceph/caps.c @@ -10,6 +10,7 @@ #include "super.h" #include "mds_client.h" +#include "cache.h" #include #include @@ -479,8 +480,9 @@ static void __check_cap_issue(struct ceph_inode_info *ci, struct ceph_cap *cap, * i_rdcache_gen. */ if ((issued & (CEPH_CAP_FILE_CACHE|CEPH_CAP_FILE_LAZYIO)) && - (had & (CEPH_CAP_FILE_CACHE|CEPH_CAP_FILE_LAZYIO)) == 0) + (had & (CEPH_CAP_FILE_CACHE|CEPH_CAP_FILE_LAZYIO)) == 0) { ci->i_rdcache_gen++; + } /* * if we are newly issued FILE_SHARED, mark dir not complete; we @@ -2395,6 +2397,7 @@ static void handle_cap_grant(struct inode *inode, struct ceph_mds_caps *grant, int writeback = 0; int queue_invalidate = 0; int deleted_inode = 0; + int queue_revalidate = 0; dout("handle_cap_grant inode %p cap %p mds%d seq %d %s\n", inode, cap, mds, seq, ceph_cap_string(newcaps)); @@ -2417,6 +2420,8 @@ static void handle_cap_grant(struct inode *inode, struct ceph_mds_caps *grant, ci->i_rdcache_revoking = ci->i_rdcache_gen; } } + + ceph_fscache_invalidate(inode); } /* side effects now are allowed */ @@ -2458,6 +2463,11 @@ static void handle_cap_grant(struct inode *inode, struct ceph_mds_caps *grant, } } + /* Do we need to revalidate our fscache cookie. Don't bother on the + * first cache cap as we already validate at cookie creation time. */ + if ((issued & CEPH_CAP_FILE_CACHE) && ci->i_rdcache_gen > 1) + queue_revalidate = 1; + /* size/ctime/mtime/atime? */ ceph_fill_file_size(inode, issued, le32_to_cpu(grant->truncate_seq), @@ -2542,6 +2552,7 @@ static void handle_cap_grant(struct inode *inode, struct ceph_mds_caps *grant, BUG_ON(cap->issued & ~cap->implemented); spin_unlock(&ci->i_ceph_lock); + if (writeback) /* * queue inode for writeback: we can't actually call @@ -2553,6 +2564,8 @@ static void handle_cap_grant(struct inode *inode, struct ceph_mds_caps *grant, ceph_queue_invalidate(inode); if (deleted_inode) invalidate_aliases(inode); + if (queue_revalidate) + ceph_queue_revalidate(inode); if (wake) wake_up_all(&ci->i_cap_wq); @@ -2709,8 +2722,10 @@ static void handle_cap_trunc(struct inode *inode, truncate_seq, truncate_size, size); spin_unlock(&ci->i_ceph_lock); - if (queue_trunc) + if (queue_trunc) { ceph_queue_vmtruncate(inode); + ceph_fscache_invalidate(inode); + } } /* diff --git a/fs/ceph/file.c b/fs/ceph/file.c index 20d0222c2e76..3de89829e2a1 100644 --- a/fs/ceph/file.c +++ b/fs/ceph/file.c @@ -12,6 +12,7 @@ #include "super.h" #include "mds_client.h" +#include "cache.h" /* * Ceph file operations @@ -69,9 +70,23 @@ static int ceph_init_file(struct inode *inode, struct file *file, int fmode) { struct ceph_file_info *cf; int ret = 0; + struct ceph_inode_info *ci = ceph_inode(inode); + struct ceph_fs_client *fsc = ceph_sb_to_client(inode->i_sb); + struct ceph_mds_client *mdsc = fsc->mdsc; switch (inode->i_mode & S_IFMT) { case S_IFREG: + /* First file open request creates the cookie, we want to keep + * this cookie around for the filetime of the inode as not to + * have to worry about fscache register / revoke / operation + * races. + * + * Also, if we know the operation is going to invalidate data + * (non readonly) just nuke the cache right away. + */ + ceph_fscache_register_inode_cookie(mdsc->fsc, ci); + if ((fmode & CEPH_FILE_MODE_WR)) + ceph_fscache_invalidate(inode); case S_IFDIR: dout("init_file %p %p 0%o (regular)\n", inode, file, inode->i_mode); @@ -182,6 +197,7 @@ int ceph_open(struct inode *inode, struct file *file) spin_unlock(&ci->i_ceph_lock); return ceph_init_file(inode, file, fmode); } + spin_unlock(&ci->i_ceph_lock); dout("open fmode %d wants %s\n", fmode, ceph_cap_string(wanted)); @@ -192,6 +208,7 @@ int ceph_open(struct inode *inode, struct file *file) } req->r_inode = inode; ihold(inode); + req->r_num_caps = 1; if (flags & (O_CREAT|O_TRUNC)) parent_inode = ceph_get_dentry_parent_inode(file->f_dentry); diff --git a/fs/ceph/inode.c b/fs/ceph/inode.c index 602ccd8e06b7..eae41cd73276 100644 --- a/fs/ceph/inode.c +++ b/fs/ceph/inode.c @@ -12,6 +12,7 @@ #include "super.h" #include "mds_client.h" +#include "cache.h" #include /* @@ -386,6 +387,8 @@ struct inode *ceph_alloc_inode(struct super_block *sb) INIT_WORK(&ci->i_vmtruncate_work, ceph_vmtruncate_work); + ceph_fscache_inode_init(ci); + return &ci->vfs_inode; } @@ -405,6 +408,8 @@ void ceph_destroy_inode(struct inode *inode) dout("destroy_inode %p ino %llx.%llx\n", inode, ceph_vinop(inode)); + ceph_fscache_unregister_inode_cookie(ci); + ceph_queue_caps_release(inode); /* @@ -439,7 +444,6 @@ void ceph_destroy_inode(struct inode *inode) call_rcu(&inode->i_rcu, ceph_i_callback); } - /* * Helpers to fill in size, ctime, mtime, and atime. We have to be * careful because either the client or MDS may have more up to date @@ -491,6 +495,10 @@ int ceph_fill_file_size(struct inode *inode, int issued, truncate_size); ci->i_truncate_size = truncate_size; } + + if (queue_trunc) + ceph_fscache_invalidate(inode); + return queue_trunc; } @@ -1079,7 +1087,7 @@ int ceph_fill_trace(struct super_block *sb, struct ceph_mds_request *req, * complete. */ ceph_set_dentry_offset(req->r_old_dentry); - dout("dn %p gets new offset %lld\n", req->r_old_dentry, + dout("dn %p gets new offset %lld\n", req->r_old_dentry, ceph_dentry(req->r_old_dentry)->offset); dn = req->r_old_dentry; /* use old_dentry */ @@ -1494,6 +1502,7 @@ void ceph_queue_vmtruncate(struct inode *inode) struct ceph_inode_info *ci = ceph_inode(inode); ihold(inode); + if (queue_work(ceph_sb_to_client(inode->i_sb)->trunc_wq, &ci->i_vmtruncate_work)) { dout("ceph_queue_vmtruncate %p\n", inode); @@ -1565,7 +1574,6 @@ retry: wake_up_all(&ci->i_cap_wq); } - /* * symlinks */ diff --git a/fs/ceph/super.c b/fs/ceph/super.c index 6627b26a800c..6a0951e43044 100644 --- a/fs/ceph/super.c +++ b/fs/ceph/super.c @@ -17,6 +17,7 @@ #include "super.h" #include "mds_client.h" +#include "cache.h" #include #include @@ -142,6 +143,8 @@ enum { Opt_nodcache, Opt_ino32, Opt_noino32, + Opt_fscache, + Opt_nofscache }; static match_table_t fsopt_tokens = { @@ -167,6 +170,8 @@ static match_table_t fsopt_tokens = { {Opt_nodcache, "nodcache"}, {Opt_ino32, "ino32"}, {Opt_noino32, "noino32"}, + {Opt_fscache, "fsc"}, + {Opt_nofscache, "nofsc"}, {-1, NULL} }; @@ -260,6 +265,12 @@ static int parse_fsopt_token(char *c, void *private) case Opt_noino32: fsopt->flags &= ~CEPH_MOUNT_OPT_INO32; break; + case Opt_fscache: + fsopt->flags |= CEPH_MOUNT_OPT_FSCACHE; + break; + case Opt_nofscache: + fsopt->flags &= ~CEPH_MOUNT_OPT_FSCACHE; + break; default: BUG_ON(token); } @@ -422,6 +433,10 @@ static int ceph_show_options(struct seq_file *m, struct dentry *root) seq_puts(m, ",dcache"); else seq_puts(m, ",nodcache"); + if (fsopt->flags & CEPH_MOUNT_OPT_FSCACHE) + seq_puts(m, ",fsc"); + else + seq_puts(m, ",nofsc"); if (fsopt->wsize) seq_printf(m, ",wsize=%d", fsopt->wsize); @@ -530,11 +545,18 @@ static struct ceph_fs_client *create_fs_client(struct ceph_mount_options *fsopt, if (!fsc->wb_pagevec_pool) goto fail_trunc_wq; + /* setup fscache */ + if ((fsopt->flags & CEPH_MOUNT_OPT_FSCACHE) && + (ceph_fscache_register_fs(fsc) != 0)) + goto fail_fscache; + /* caps */ fsc->min_caps = fsopt->max_readdir; return fsc; +fail_fscache: + ceph_fscache_unregister_fs(fsc); fail_trunc_wq: destroy_workqueue(fsc->trunc_wq); fail_pg_inv_wq: @@ -554,6 +576,8 @@ static void destroy_fs_client(struct ceph_fs_client *fsc) { dout("destroy_fs_client %p\n", fsc); + ceph_fscache_unregister_fs(fsc); + destroy_workqueue(fsc->wb_wq); destroy_workqueue(fsc->pg_inv_wq); destroy_workqueue(fsc->trunc_wq); @@ -588,6 +612,8 @@ static void ceph_inode_init_once(void *foo) static int __init init_caches(void) { + int error = -ENOMEM; + ceph_inode_cachep = kmem_cache_create("ceph_inode_info", sizeof(struct ceph_inode_info), __alignof__(struct ceph_inode_info), @@ -611,15 +637,17 @@ static int __init init_caches(void) if (ceph_file_cachep == NULL) goto bad_file; - return 0; + if ((error = ceph_fscache_register())) + goto bad_file; + return 0; bad_file: kmem_cache_destroy(ceph_dentry_cachep); bad_dentry: kmem_cache_destroy(ceph_cap_cachep); bad_cap: kmem_cache_destroy(ceph_inode_cachep); - return -ENOMEM; + return error; } static void destroy_caches(void) @@ -629,10 +657,13 @@ static void destroy_caches(void) * destroy cache. */ rcu_barrier(); + kmem_cache_destroy(ceph_inode_cachep); kmem_cache_destroy(ceph_cap_cachep); kmem_cache_destroy(ceph_dentry_cachep); kmem_cache_destroy(ceph_file_cachep); + + ceph_fscache_unregister(); } diff --git a/fs/ceph/super.h b/fs/ceph/super.h index f1e4e4766ea2..bb23ef636177 100644 --- a/fs/ceph/super.h +++ b/fs/ceph/super.h @@ -16,6 +16,10 @@ #include +#ifdef CONFIG_CEPH_FSCACHE +#include +#endif + /* f_type in struct statfs */ #define CEPH_SUPER_MAGIC 0x00c36400 @@ -29,6 +33,7 @@ #define CEPH_MOUNT_OPT_NOASYNCREADDIR (1<<7) /* no dcache readdir */ #define CEPH_MOUNT_OPT_INO32 (1<<8) /* 32 bit inos */ #define CEPH_MOUNT_OPT_DCACHE (1<<9) /* use dcache for readdir etc */ +#define CEPH_MOUNT_OPT_FSCACHE (1<<10) /* use fscache */ #define CEPH_MOUNT_OPT_DEFAULT (CEPH_MOUNT_OPT_RBYTES) @@ -90,6 +95,11 @@ struct ceph_fs_client { struct dentry *debugfs_bdi; struct dentry *debugfs_mdsc, *debugfs_mdsmap; #endif + +#ifdef CONFIG_CEPH_FSCACHE + struct fscache_cookie *fscache; + struct workqueue_struct *revalidate_wq; +#endif }; @@ -320,6 +330,12 @@ struct ceph_inode_info { struct work_struct i_vmtruncate_work; +#ifdef CONFIG_CEPH_FSCACHE + struct fscache_cookie *fscache; + u32 i_fscache_gen; /* sequence, for delayed fscache validate */ + struct work_struct i_revalidate_work; +#endif + struct inode vfs_inode; /* at end */ }; -- cgit v1.2.3-59-g8ed1b From 76be778b3a4eae63ee4dcb22ff2045d3a0fe863b Mon Sep 17 00:00:00 2001 From: Milosz Tanski Date: Wed, 21 Aug 2013 17:30:27 -0400 Subject: ceph: clean PgPrivate2 on returning from readpages In some cases the ceph readapages code code bails without filling all the pages already marked by fscache. When we return back to readahead code this causes a BUG. Signed-off-by: Milosz Tanski --- fs/ceph/addr.c | 2 ++ fs/ceph/cache.h | 7 +++++++ 2 files changed, 9 insertions(+) (limited to 'fs') diff --git a/fs/ceph/addr.c b/fs/ceph/addr.c index 3a21a7cbc21c..1fda9cf04a81 100644 --- a/fs/ceph/addr.c +++ b/fs/ceph/addr.c @@ -398,6 +398,8 @@ static int ceph_readpages(struct file *file, struct address_space *mapping, BUG_ON(rc == 0); } out: + ceph_fscache_readpages_cancel(inode, page_list); + dout("readpages %p file %p ret %d\n", inode, file, rc); return rc; } diff --git a/fs/ceph/cache.h b/fs/ceph/cache.h index 0ea95cb7f389..fb326fd33251 100644 --- a/fs/ceph/cache.h +++ b/fs/ceph/cache.h @@ -58,6 +58,13 @@ static inline int ceph_release_fscache_page(struct page *page, gfp_t gfp) return fscache_maybe_release_page(ci->fscache, page, gfp); } +static inline void ceph_fscache_readpages_cancel(struct inode *inode, + struct list_head *pages) +{ + struct ceph_inode_info *ci = ceph_inode(inode); + return fscache_readpages_cancel(ci->fscache, pages); +} + #else static inline int ceph_fscache_register(void) -- cgit v1.2.3-59-g8ed1b From 9b8dd1e8a55a12b67240b6b28160ac8c1dec0172 Mon Sep 17 00:00:00 2001 From: Milosz Tanski Date: Tue, 3 Sep 2013 19:11:01 -0400 Subject: ceph: ceph_readpage_to_fscache didn't check if marked Previously ceph_readpage_to_fscache did not call if page was marked as cached before calling fscache_write_page resulting in a BUG inside of fscache. FS-Cache: Assertion failed ------------[ cut here ]------------ kernel BUG at fs/fscache/page.c:874! invalid opcode: 0000 [#1] SMP Call Trace: [] __ceph_readpage_to_fscache+0x66/0x80 [ceph] [] readpage_nounlock+0x124/0x210 [ceph] [] ceph_readpage+0x1d/0x40 [ceph] [] generic_file_aio_read+0x1f6/0x700 [] ceph_aio_read+0x5fc/0xab0 [ceph] Signed-off-by: Milosz Tanski Signed-off-by: Sage Weil --- fs/ceph/cache.c | 3 +++ 1 file changed, 3 insertions(+) (limited to 'fs') diff --git a/fs/ceph/cache.c b/fs/ceph/cache.c index 5c413ecf1f15..c737ae9893ef 100644 --- a/fs/ceph/cache.c +++ b/fs/ceph/cache.c @@ -311,6 +311,9 @@ void ceph_readpage_to_fscache(struct inode *inode, struct page *page) struct ceph_inode_info *ci = ceph_inode(inode); int ret; + if (!PageFsCache(page)) + return; + if (!cache_valid(ci)) return; -- cgit v1.2.3-59-g8ed1b From d4d3aa38d66d0313401534bff2e4647df0a6d538 Mon Sep 17 00:00:00 2001 From: Milosz Tanski Date: Tue, 3 Sep 2013 19:11:17 -0400 Subject: ceph: page still marked private_2 Previous patch that allowed us to cleanup most of the issues with pages marked as private_2 when calling ceph_readpages. However, there seams to be a case in the error case clean up in start read that still trigers this from time to time. I've only seen this one a couple times. BUG: Bad page state in process petabucket pfn:335b82 page:ffffea000cd6e080 count:0 mapcount:0 mapping: (null) index:0x0 page flags: 0x200000000001000(private_2) Call Trace: [] dump_stack+0x46/0x58 [] bad_page+0xc7/0x120 [] free_pages_prepare+0x10e/0x120 [] free_hot_cold_page+0x40/0x160 [] __put_single_page+0x27/0x30 [] put_page+0x25/0x40 [] ceph_readpages+0x2e9/0x6f0 [ceph] [] __do_page_cache_readahead+0x1af/0x260 Signed-off-by: Milosz Tanski Signed-off-by: Sage Weil --- fs/ceph/addr.c | 1 + fs/ceph/cache.h | 14 +++++++++++++- 2 files changed, 14 insertions(+), 1 deletion(-) (limited to 'fs') diff --git a/fs/ceph/addr.c b/fs/ceph/addr.c index 1fda9cf04a81..6df8bd481425 100644 --- a/fs/ceph/addr.c +++ b/fs/ceph/addr.c @@ -336,6 +336,7 @@ static int start_read(struct inode *inode, struct list_head *page_list, int max) page->index); if (add_to_page_cache_lru(page, &inode->i_data, page->index, GFP_NOFS)) { + ceph_fscache_uncache_page(inode, page); page_cache_release(page); dout("start_read %p add_to_page_cache failed %p\n", inode, page); diff --git a/fs/ceph/cache.h b/fs/ceph/cache.h index fb326fd33251..bf4869547291 100644 --- a/fs/ceph/cache.h +++ b/fs/ceph/cache.h @@ -51,6 +51,13 @@ static inline void ceph_fscache_invalidate(struct inode *inode) fscache_invalidate(ceph_inode(inode)->fscache); } +static inline void ceph_fscache_uncache_page(struct inode *inode, + struct page *page) +{ + struct ceph_inode_info *ci = ceph_inode(inode); + return fscache_uncache_page(ci->fscache, page); +} + static inline int ceph_release_fscache_page(struct page *page, gfp_t gfp) { struct inode* inode = page->mapping->host; @@ -94,7 +101,8 @@ static inline void ceph_fscache_register_inode_cookie(struct ceph_fs_client* par { } -static inline void ceph_fscache_unregister_inode_cookie(struct ceph_inode_info* ci) +static inline void ceph_fscache_uncache_page(struct inode *inode, + struct page *pages) { } @@ -126,6 +134,10 @@ static inline void ceph_invalidate_fscache_page(struct inode *inode, { } +static inline void ceph_fscache_unregister_inode_cookie(struct ceph_inode_info* ci) +{ +} + static inline int ceph_release_fscache_page(struct page *page, gfp_t gfp) { return 1; -- cgit v1.2.3-59-g8ed1b From e81568eb1819af1391ac27ab28ac851410315a9f Mon Sep 17 00:00:00 2001 From: Milosz Tanski Date: Thu, 5 Sep 2013 18:29:03 +0000 Subject: ceph: Do not do invalidate if the filesystem is mounted nofsc Previously we would always try to enqueue work even if the filesystem is not mounted with fscache enabled (or the file has no cookie). In the case of the filesystem mouned nofsc (but with fscache compiled in) this would lead to a crash. Signed-off-by: Milosz Tanski --- fs/ceph/cache.c | 4 ++++ 1 file changed, 4 insertions(+) (limited to 'fs') diff --git a/fs/ceph/cache.c b/fs/ceph/cache.c index c737ae9893ef..d3b88c7518d0 100644 --- a/fs/ceph/cache.c +++ b/fs/ceph/cache.c @@ -374,8 +374,12 @@ out: void ceph_queue_revalidate(struct inode *inode) { + struct ceph_fs_client *fsc = ceph_sb_to_client(inode->i_sb); struct ceph_inode_info *ci = ceph_inode(inode); + if (fsc->revalidate_wq == NULL || ci->fscache == NULL) + return; + ihold(inode); if (queue_work(ceph_sb_to_client(inode->i_sb)->revalidate_wq, -- cgit v1.2.3-59-g8ed1b From 971f0bdeaabac4fcc335dace2f98e79157db4302 Mon Sep 17 00:00:00 2001 From: Milosz Tanski Date: Fri, 6 Sep 2013 15:13:18 +0000 Subject: ceph: trivial buildbot warnings fix The linux-next build bot found a three of warnings, this addresses all of them. * non-ANSI function declaration of function 'ceph_fscache_register' and 'ceph_fscache_unregister' * symbol 'ceph_cache_netfs' was not declared, now it's extern in the header. * warning: "pr_fmt" redefined Signed-off-by: Milosz Tanski --- fs/ceph/cache.c | 6 ++---- fs/ceph/cache.h | 2 ++ 2 files changed, 4 insertions(+), 4 deletions(-) (limited to 'fs') diff --git a/fs/ceph/cache.c b/fs/ceph/cache.c index d3b88c7518d0..6bfe65e0b038 100644 --- a/fs/ceph/cache.c +++ b/fs/ceph/cache.c @@ -21,8 +21,6 @@ * */ -#include - #include "super.h" #include "cache.h" @@ -56,12 +54,12 @@ static const struct fscache_cookie_def ceph_fscache_fsid_object_def = { .get_key = ceph_fscache_session_get_key, }; -int ceph_fscache_register() +int ceph_fscache_register(void) { return fscache_register_netfs(&ceph_cache_netfs); } -void ceph_fscache_unregister() +void ceph_fscache_unregister(void) { fscache_unregister_netfs(&ceph_cache_netfs); } diff --git a/fs/ceph/cache.h b/fs/ceph/cache.h index bf4869547291..ba949408a336 100644 --- a/fs/ceph/cache.h +++ b/fs/ceph/cache.h @@ -26,6 +26,8 @@ #ifdef CONFIG_CEPH_FSCACHE +extern struct fscache_netfs ceph_cache_netfs; + int ceph_fscache_register(void); void ceph_fscache_unregister(void); -- cgit v1.2.3-59-g8ed1b From ed284c49f61165c3ba1b4e6969d1cc30a769c31b Mon Sep 17 00:00:00 2001 From: "Yan, Zheng" Date: Mon, 2 Sep 2013 15:19:53 +0800 Subject: ceph: remove ceph_lookup_inode() commit 6f60f889 (ceph: fix freeing inode vs removing session caps race) introduced ceph_lookup_inode(). But there is already a ceph_find_inode() which provides similar function. So remove ceph_lookup_inode(), use ceph_find_inode() instead. Signed-off-by: Yan, Zheng Reviewed-by: Alex Elder Reviewed-by: Sage Weil --- fs/ceph/inode.c | 8 -------- fs/ceph/mds_client.c | 2 +- fs/ceph/super.h | 2 -- 3 files changed, 1 insertion(+), 11 deletions(-) (limited to 'fs') diff --git a/fs/ceph/inode.c b/fs/ceph/inode.c index eae41cd73276..8549a48115f7 100644 --- a/fs/ceph/inode.c +++ b/fs/ceph/inode.c @@ -62,14 +62,6 @@ struct inode *ceph_get_inode(struct super_block *sb, struct ceph_vino vino) return inode; } -struct inode *ceph_lookup_inode(struct super_block *sb, struct ceph_vino vino) -{ - struct inode *inode; - ino_t t = ceph_vino_to_ino(vino); - inode = ilookup5_nowait(sb, t, ceph_ino_compare, &vino); - return inode; -} - /* * get/constuct snapdir inode for a given directory */ diff --git a/fs/ceph/mds_client.c b/fs/ceph/mds_client.c index 603786b564be..b7bda5d9611d 100644 --- a/fs/ceph/mds_client.c +++ b/fs/ceph/mds_client.c @@ -1054,7 +1054,7 @@ static void remove_session_caps(struct ceph_mds_session *session) vino = cap->ci->i_vino; spin_unlock(&session->s_cap_lock); - inode = ceph_lookup_inode(sb, vino); + inode = ceph_find_inode(sb, vino); iput(inode); spin_lock(&session->s_cap_lock); diff --git a/fs/ceph/super.h b/fs/ceph/super.h index bb23ef636177..6014b0a3c405 100644 --- a/fs/ceph/super.h +++ b/fs/ceph/super.h @@ -694,8 +694,6 @@ extern void ceph_destroy_inode(struct inode *inode); extern struct inode *ceph_get_inode(struct super_block *sb, struct ceph_vino vino); -extern struct inode *ceph_lookup_inode(struct super_block *sb, - struct ceph_vino vino); extern struct inode *ceph_get_snapdir(struct inode *parent); extern int ceph_fill_file_size(struct inode *inode, int issued, u32 truncate_seq, u64 truncate_size, u64 size); -- cgit v1.2.3-59-g8ed1b From a8d436f015b627a55ec3b1d15f13d6ab92dd892b Mon Sep 17 00:00:00 2001 From: "Yan, Zheng" Date: Mon, 2 Sep 2013 15:19:54 +0800 Subject: ceph: use d_invalidate() to invalidate aliases d_invalidate() is the standard VFS method to invalidate dentry. compare to d_delete(), it also try shrinking children dentries. Signed-off-by: Yan, Zheng Reviewed-by: Sage Weil --- fs/ceph/caps.c | 8 ++++---- 1 file changed, 4 insertions(+), 4 deletions(-) (limited to 'fs') diff --git a/fs/ceph/caps.c b/fs/ceph/caps.c index 7b451eb7d123..13976c33332e 100644 --- a/fs/ceph/caps.c +++ b/fs/ceph/caps.c @@ -2346,10 +2346,10 @@ static void invalidate_aliases(struct inode *inode) d_prune_aliases(inode); /* * For non-directory inode, d_find_alias() only returns - * connected dentry. After calling d_delete(), the dentry - * become disconnected. + * connected dentry. After calling d_invalidate(), the + * dentry become disconnected. * - * For directory inode, d_find_alias() only can return + * For directory inode, d_find_alias() can return * disconnected dentry. But directory inode should have * one alias at most. */ @@ -2358,7 +2358,7 @@ static void invalidate_aliases(struct inode *inode) dput(dn); break; } - d_delete(dn); + d_invalidate(dn); if (prev) dput(prev); prev = dn; -- cgit v1.2.3-59-g8ed1b