From 29c8bbbd6e21daa0997d1c3ee886b897ee7ad652 Mon Sep 17 00:00:00 2001 From: David Howells Date: Thu, 16 Mar 2017 16:27:43 +0000 Subject: afs: Fix missing put_page() In afs_writepages_region(), inside the loop where we find dirty pages to deal with, one of the if-statements is missing a put_page(). Signed-off-by: David Howells --- fs/afs/write.c | 1 + 1 file changed, 1 insertion(+) diff --git a/fs/afs/write.c b/fs/afs/write.c index c83c1a0e851f..e919e64cd4e0 100644 --- a/fs/afs/write.c +++ b/fs/afs/write.c @@ -513,6 +513,7 @@ static int afs_writepages_region(struct address_space *mapping, if (PageWriteback(page) || !PageDirty(page)) { unlock_page(page); + put_page(page); continue; } -- cgit From 5611ef280d814042825ee17688f5751266fc538b Mon Sep 17 00:00:00 2001 From: David Howells Date: Thu, 16 Mar 2017 16:27:43 +0000 Subject: afs: Fix page overput in afs_fill_page() afs_fill_page() loads the page it wants to fill into the afs_read request without incrementing its refcount - but then calls afs_put_read() to clean up afterwards, which then releases a ref on the page. Fix this by getting a ref on the page before calling afs_vnode_fetch_data(). This causes sync after a write to hang in afs_writepages_region() because find_get_pages_tag() gets confused and doesn't return. Fixes: 196ee9cd2d04 ("afs: Make afs_fs_fetch_data() take a list of pages") Reported-by: Marc Dionne Signed-off-by: David Howells Tested-by: Marc Dionne --- fs/afs/write.c | 1 + 1 file changed, 1 insertion(+) diff --git a/fs/afs/write.c b/fs/afs/write.c index e919e64cd4e0..3ac52f6a96ff 100644 --- a/fs/afs/write.c +++ b/fs/afs/write.c @@ -101,6 +101,7 @@ static int afs_fill_page(struct afs_vnode *vnode, struct key *key, req->pos = pos; req->nr_pages = 1; req->pages[0] = page; + get_page(page); i_size = i_size_read(&vnode->vfs_inode); if (pos + PAGE_SIZE > i_size) -- cgit From 6186f0788b31f44affceeedc7b48eb10faea120d Mon Sep 17 00:00:00 2001 From: Marc Dionne Date: Thu, 16 Mar 2017 16:27:43 +0000 Subject: afs: Populate group ID from vnode status The group was hard coded to GLOBAL_ROOT_GID; use the group ID that was received from the server. Signed-off-by: Marc Dionne Signed-off-by: David Howells --- fs/afs/inode.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/fs/afs/inode.c b/fs/afs/inode.c index 1e4897a048d2..299dbaeb2e2a 100644 --- a/fs/afs/inode.c +++ b/fs/afs/inode.c @@ -70,7 +70,7 @@ static int afs_inode_map_status(struct afs_vnode *vnode, struct key *key) set_nlink(inode, vnode->status.nlink); inode->i_uid = vnode->status.owner; - inode->i_gid = GLOBAL_ROOT_GID; + inode->i_gid = vnode->status.group; inode->i_size = vnode->status.size; inode->i_ctime.tv_sec = vnode->status.mtime_server; inode->i_ctime.tv_nsec = 0; -- cgit From 627f46943ff90bcc32ddeb675d881c043c6fa2ae Mon Sep 17 00:00:00 2001 From: Marc Dionne Date: Thu, 16 Mar 2017 16:27:44 +0000 Subject: afs: Adjust mode bits processing Mode bits for an afs file should not be enforced in the usual way. For files, the absence of user bits can restrict file access with respect to what is granted by the server. These bits apply regardless of the owner or the current uid; the rest of the mode bits (group, other) are ignored. Signed-off-by: Marc Dionne Signed-off-by: David Howells --- fs/afs/security.c | 7 ++++++- 1 file changed, 6 insertions(+), 1 deletion(-) diff --git a/fs/afs/security.c b/fs/afs/security.c index 8d010422dc89..bfa9d3428383 100644 --- a/fs/afs/security.c +++ b/fs/afs/security.c @@ -340,17 +340,22 @@ int afs_permission(struct inode *inode, int mask) } else { if (!(access & AFS_ACE_LOOKUP)) goto permission_denied; + if ((mask & MAY_EXEC) && !(inode->i_mode & S_IXUSR)) + goto permission_denied; if (mask & (MAY_EXEC | MAY_READ)) { if (!(access & AFS_ACE_READ)) goto permission_denied; + if (!(inode->i_mode & S_IRUSR)) + goto permission_denied; } else if (mask & MAY_WRITE) { if (!(access & AFS_ACE_WRITE)) goto permission_denied; + if (!(inode->i_mode & S_IWUSR)) + goto permission_denied; } } key_put(key); - ret = generic_permission(inode, mask); _leave(" = %d", ret); return ret; -- cgit From bcd89270d93b7edebb5de5e5e7dca1a77a33496e Mon Sep 17 00:00:00 2001 From: Marc Dionne Date: Thu, 16 Mar 2017 16:27:44 +0000 Subject: afs: Deal with an empty callback array Servers may send a callback array that is the same size as the FID array, or an empty array. If the callback count is 0, the code would attempt to read (fid_count * 12) bytes of data, which would fail and result in an unmarshalling error. This would lead to stale data for remotely modified files or directories. Store the callback array size in the internal afs_call structure and use that to determine the amount of data to read. Signed-off-by: Marc Dionne --- fs/afs/cmservice.c | 11 +++++------ fs/afs/internal.h | 5 ++++- 2 files changed, 9 insertions(+), 7 deletions(-) diff --git a/fs/afs/cmservice.c b/fs/afs/cmservice.c index 2edbdcbf6432..3062cceb5c2a 100644 --- a/fs/afs/cmservice.c +++ b/fs/afs/cmservice.c @@ -187,7 +187,6 @@ static int afs_deliver_cb_callback(struct afs_call *call) struct afs_callback *cb; struct afs_server *server; __be32 *bp; - u32 tmp; int ret, loop; _enter("{%u}", call->unmarshall); @@ -249,9 +248,9 @@ static int afs_deliver_cb_callback(struct afs_call *call) if (ret < 0) return ret; - tmp = ntohl(call->tmp); - _debug("CB count: %u", tmp); - if (tmp != call->count && tmp != 0) + call->count2 = ntohl(call->tmp); + _debug("CB count: %u", call->count2); + if (call->count2 != call->count && call->count2 != 0) return -EBADMSG; call->offset = 0; call->unmarshall++; @@ -259,14 +258,14 @@ static int afs_deliver_cb_callback(struct afs_call *call) case 4: _debug("extract CB array"); ret = afs_extract_data(call, call->buffer, - call->count * 3 * 4, false); + call->count2 * 3 * 4, false); if (ret < 0) return ret; _debug("unmarshall CB array"); cb = call->request; bp = call->buffer; - for (loop = call->count; loop > 0; loop--, cb++) { + for (loop = call->count2; loop > 0; loop--, cb++) { cb->version = ntohl(*bp++); cb->expiry = ntohl(*bp++); cb->type = ntohl(*bp++); diff --git a/fs/afs/internal.h b/fs/afs/internal.h index 5dfa56903a2d..8499870147ef 100644 --- a/fs/afs/internal.h +++ b/fs/afs/internal.h @@ -90,7 +90,10 @@ struct afs_call { unsigned request_size; /* size of request data */ unsigned reply_max; /* maximum size of reply */ unsigned first_offset; /* offset into mapping[first] */ - unsigned last_to; /* amount of mapping[last] */ + union { + unsigned last_to; /* amount of mapping[last] */ + unsigned count2; /* count used in unmarshalling */ + }; unsigned char unmarshall; /* unmarshalling phase */ bool incoming; /* T if incoming call */ bool send_pages; /* T if data from mapping should be sent */ -- cgit From 6db3ac3c4bc552837d232ec794559a2fae2815a0 Mon Sep 17 00:00:00 2001 From: David Howells Date: Thu, 16 Mar 2017 16:27:44 +0000 Subject: afs: Handle better the server returning excess or short data When an AFS server is given an FS.FetchData{,64} request to read data from a file, it is permitted by the protocol to return more or less than was requested. kafs currently relies on the latter behaviour in readpage{,s} to handle a partial page at the end of the file (we just ask for a whole page and clear space beyond the short read). However, we don't handle all cases. Add: (1) Handle excess data by discarding it rather than aborting. Note that we use a common static buffer to discard into so that the decryption algorithm advances the PCBC state. (2) Handle a short read that affects more than just the last page. Note that if a read comes up unexpectedly short of long, it's possible that the server's copy of the file changed - in which case the data version number will have been incremented and the callback will have been broken - in which case all the pages currently attached to the inode will be zapped anyway at some point. Signed-off-by: David Howells --- fs/afs/file.c | 7 +++++-- fs/afs/fsclient.c | 49 +++++++++++++++++++++++++++++++++++-------------- 2 files changed, 40 insertions(+), 16 deletions(-) diff --git a/fs/afs/file.c b/fs/afs/file.c index ba7b71fba34b..a38e1c30d110 100644 --- a/fs/afs/file.c +++ b/fs/afs/file.c @@ -184,10 +184,13 @@ int afs_page_filler(void *data, struct page *page) if (!req) goto enomem; + /* We request a full page. If the page is a partial one at the + * end of the file, the server will return a short read and the + * unmarshalling code will clear the unfilled space. + */ atomic_set(&req->usage, 1); req->pos = (loff_t)page->index << PAGE_SHIFT; - req->len = min_t(size_t, i_size_read(inode) - req->pos, - PAGE_SIZE); + req->len = PAGE_SIZE; req->nr_pages = 1; req->pages[0] = page; get_page(page); diff --git a/fs/afs/fsclient.c b/fs/afs/fsclient.c index ac8e766978dc..bf8904a1a58f 100644 --- a/fs/afs/fsclient.c +++ b/fs/afs/fsclient.c @@ -16,6 +16,12 @@ #include "internal.h" #include "afs_fs.h" +/* + * We need somewhere to discard into in case the server helpfully returns more + * than we asked for in FS.FetchData{,64}. + */ +static u8 afs_discard_buffer[64]; + /* * decode an AFSFid block */ @@ -353,12 +359,6 @@ static int afs_deliver_fs_fetch_data(struct afs_call *call) req->actual_len |= ntohl(call->tmp); _debug("DATA length: %llu", req->actual_len); - /* Check that the server didn't want to send us extra. We - * might want to just discard instead, but that requires - * cooperation from AF_RXRPC. - */ - if (req->actual_len > req->len) - return -EBADMSG; req->remain = req->actual_len; call->offset = req->pos & (PAGE_SIZE - 1); @@ -368,6 +368,7 @@ static int afs_deliver_fs_fetch_data(struct afs_call *call) call->unmarshall++; begin_page: + ASSERTCMP(req->index, <, req->nr_pages); if (req->remain > PAGE_SIZE - call->offset) size = PAGE_SIZE - call->offset; else @@ -390,18 +391,37 @@ static int afs_deliver_fs_fetch_data(struct afs_call *call) if (req->page_done) req->page_done(call, req); if (req->remain > 0) { - req->index++; call->offset = 0; + req->index++; + if (req->index >= req->nr_pages) + goto begin_discard; goto begin_page; } } + goto no_more_data; + + /* Discard any excess data the server gave us */ + begin_discard: + case 4: + size = min_t(size_t, sizeof(afs_discard_buffer), req->remain); + call->count = size; + _debug("extract discard %u/%llu %zu/%u", + req->remain, req->actual_len, call->offset, call->count); + + call->offset = 0; + ret = afs_extract_data(call, afs_discard_buffer, call->count, true); + req->remain -= call->offset; + if (ret < 0) + return ret; + if (req->remain > 0) + goto begin_discard; no_more_data: call->offset = 0; - call->unmarshall++; + call->unmarshall = 5; /* extract the metadata */ - case 4: + case 5: ret = afs_extract_data(call, call->buffer, (21 + 3 + 6) * 4, false); if (ret < 0) @@ -416,16 +436,17 @@ static int afs_deliver_fs_fetch_data(struct afs_call *call) call->offset = 0; call->unmarshall++; - case 5: + case 6: break; } - if (call->count < PAGE_SIZE) { - buffer = kmap(req->pages[req->index]); - memset(buffer + call->count, 0, PAGE_SIZE - call->count); - kunmap(req->pages[req->index]); + for (; req->index < req->nr_pages; req->index++) { + if (call->count < PAGE_SIZE) + zero_user_segment(req->pages[req->index], + call->count, PAGE_SIZE); if (req->page_done) req->page_done(call, req); + call->count = 0; } _leave(" = 0 [done]"); -- cgit From 3448e6521755862446aed28e29abf12565d8844e Mon Sep 17 00:00:00 2001 From: David Howells Date: Thu, 16 Mar 2017 16:27:44 +0000 Subject: afs: Kill struct afs_read::pg_offset Kill struct afs_read::pg_offset as nothing uses it. It's unnecessary as pos can be masked off. Signed-off-by: David Howells --- fs/afs/internal.h | 1 - 1 file changed, 1 deletion(-) diff --git a/fs/afs/internal.h b/fs/afs/internal.h index 8499870147ef..7784a8bc375c 100644 --- a/fs/afs/internal.h +++ b/fs/afs/internal.h @@ -135,7 +135,6 @@ struct afs_read { atomic_t usage; unsigned int remain; /* Amount remaining */ unsigned int index; /* Which page we're reading into */ - unsigned int pg_offset; /* Offset in page we're at */ unsigned int nr_pages; void (*page_done)(struct afs_call *, struct afs_read *); struct page *pages[]; -- cgit From e8e581a88c5f5fc7cf1f636d122b77fbcfc8c2f6 Mon Sep 17 00:00:00 2001 From: David Howells Date: Thu, 16 Mar 2017 16:27:44 +0000 Subject: afs: Handle a short write to an AFS page Handle the situation where afs_write_begin() is told to expect that a full-page write will be made, but this doesn't happen (EFAULT, CTRL-C, etc.), and so afs_write_end() sees a partial write took place. Currently, no attempt is to deal with the discrepency. Fix this by loading the gap from the server. Reported-by: Al Viro Signed-off-by: David Howells --- fs/afs/fsclient.c | 4 +++- fs/afs/internal.h | 2 +- fs/afs/write.c | 28 +++++++++++++++++++--------- 3 files changed, 23 insertions(+), 11 deletions(-) diff --git a/fs/afs/fsclient.c b/fs/afs/fsclient.c index bf8904a1a58f..6f917dd1238c 100644 --- a/fs/afs/fsclient.c +++ b/fs/afs/fsclient.c @@ -393,8 +393,10 @@ static int afs_deliver_fs_fetch_data(struct afs_call *call) if (req->remain > 0) { call->offset = 0; req->index++; - if (req->index >= req->nr_pages) + if (req->index >= req->nr_pages) { + call->unmarshall = 4; goto begin_discard; + } goto begin_page; } } diff --git a/fs/afs/internal.h b/fs/afs/internal.h index 7784a8bc375c..dc2cb486e127 100644 --- a/fs/afs/internal.h +++ b/fs/afs/internal.h @@ -130,7 +130,7 @@ struct afs_call_type { */ struct afs_read { loff_t pos; /* Where to start reading */ - loff_t len; /* How much to read */ + loff_t len; /* How much we're asking for */ loff_t actual_len; /* How much we're actually getting */ atomic_t usage; unsigned int remain; /* Amount remaining */ diff --git a/fs/afs/write.c b/fs/afs/write.c index 3ac52f6a96ff..ea66890fc188 100644 --- a/fs/afs/write.c +++ b/fs/afs/write.c @@ -84,10 +84,9 @@ void afs_put_writeback(struct afs_writeback *wb) * partly or wholly fill a page that's under preparation for writing */ static int afs_fill_page(struct afs_vnode *vnode, struct key *key, - loff_t pos, struct page *page) + loff_t pos, unsigned int len, struct page *page) { struct afs_read *req; - loff_t i_size; int ret; _enter(",,%llu", (unsigned long long)pos); @@ -99,16 +98,11 @@ static int afs_fill_page(struct afs_vnode *vnode, struct key *key, atomic_set(&req->usage, 1); req->pos = pos; + req->len = len; req->nr_pages = 1; req->pages[0] = page; get_page(page); - i_size = i_size_read(&vnode->vfs_inode); - if (pos + PAGE_SIZE > i_size) - req->len = i_size - pos; - else - req->len = PAGE_SIZE; - ret = afs_vnode_fetch_data(vnode, key, req); afs_put_read(req); if (ret < 0) { @@ -164,7 +158,7 @@ int afs_write_begin(struct file *file, struct address_space *mapping, /* page won't leak in error case: it eventually gets cleaned off LRU */ if (!PageUptodate(page) && len != PAGE_SIZE) { - ret = afs_fill_page(vnode, key, index << PAGE_SHIFT, page); + ret = afs_fill_page(vnode, key, pos & PAGE_MASK, PAGE_SIZE, page); if (ret < 0) { kfree(candidate); _leave(" = %d [prep]", ret); @@ -258,7 +252,9 @@ int afs_write_end(struct file *file, struct address_space *mapping, struct page *page, void *fsdata) { struct afs_vnode *vnode = AFS_FS_I(file_inode(file)); + struct key *key = file->private_data; loff_t i_size, maybe_i_size; + int ret; _enter("{%x:%u},{%lx}", vnode->fid.vid, vnode->fid.vnode, page->index); @@ -274,6 +270,20 @@ int afs_write_end(struct file *file, struct address_space *mapping, spin_unlock(&vnode->writeback_lock); } + if (!PageUptodate(page)) { + if (copied < len) { + /* Try and load any missing data from the server. The + * unmarshalling routine will take care of clearing any + * bits that are beyond the EOF. + */ + ret = afs_fill_page(vnode, key, pos + copied, + len - copied, page); + if (ret < 0) + return ret; + } + SetPageUptodate(page); + } + set_page_dirty(page); if (PageDirty(page)) _debug("dirtied"); -- cgit From 58fed94dfb17e89556b5705f20f90e5b2971b6a1 Mon Sep 17 00:00:00 2001 From: David Howells Date: Thu, 16 Mar 2017 16:27:45 +0000 Subject: afs: Flush outstanding writes when an fd is closed Flush outstanding writes in afs when an fd is closed. This is what NFS and CIFS do. Reported-by: Marc Dionne Signed-off-by: David Howells --- fs/afs/file.c | 1 + fs/afs/internal.h | 1 + fs/afs/write.c | 14 ++++++++++++++ 3 files changed, 16 insertions(+) diff --git a/fs/afs/file.c b/fs/afs/file.c index a38e1c30d110..b5829443ff69 100644 --- a/fs/afs/file.c +++ b/fs/afs/file.c @@ -30,6 +30,7 @@ static int afs_readpages(struct file *filp, struct address_space *mapping, const struct file_operations afs_file_operations = { .open = afs_open, + .flush = afs_flush, .release = afs_release, .llseek = generic_file_llseek, .read_iter = generic_file_read_iter, diff --git a/fs/afs/internal.h b/fs/afs/internal.h index dc2cb486e127..af1d91ec7f2c 100644 --- a/fs/afs/internal.h +++ b/fs/afs/internal.h @@ -720,6 +720,7 @@ extern int afs_writepages(struct address_space *, struct writeback_control *); extern void afs_pages_written_back(struct afs_vnode *, struct afs_call *); extern ssize_t afs_file_write(struct kiocb *, struct iov_iter *); extern int afs_writeback_all(struct afs_vnode *); +extern int afs_flush(struct file *, fl_owner_t); extern int afs_fsync(struct file *, loff_t, loff_t, int); diff --git a/fs/afs/write.c b/fs/afs/write.c index ea66890fc188..f1450ea09406 100644 --- a/fs/afs/write.c +++ b/fs/afs/write.c @@ -757,6 +757,20 @@ out: return ret; } +/* + * Flush out all outstanding writes on a file opened for writing when it is + * closed. + */ +int afs_flush(struct file *file, fl_owner_t id) +{ + _enter(""); + + if ((file->f_mode & FMODE_WRITE) == 0) + return 0; + + return vfs_fsync(file, 0); +} + /* * notification that a previously read-only page is about to become writable * - if it returns an error, the caller will deliver a bus error signal -- cgit From 944c74f472f926785b1948efa0e73e2f1b3b539b Mon Sep 17 00:00:00 2001 From: David Howells Date: Thu, 16 Mar 2017 16:27:45 +0000 Subject: afs: Distinguish mountpoints from symlinks by file mode alone In AFS, mountpoints appear as symlinks with mode 0644 and normal symlinks have mode 0777, so use this to distinguish them rather than reading the content and parsing it. In the case of a mountpoint, the symlink body is a formatted string indicating the location of the target volume. Note that with this, kAFS no longer 'pre-fetches' the contents of symlinks, so afs_readpage() may fail with an access-denial because when the VFS calls d_automount(), it wraps the call in an credentials override that sets the initial creds - thereby preventing access to the caller's keyrings and the authentication keys held therein. To this end, a patch reverting that change to the VFS is required also. Reported-by: Jeffrey Altman Signed-off-by: David Howells --- fs/afs/inode.c | 29 +++++++++++++++-------------- fs/afs/internal.h | 1 - fs/afs/mntpt.c | 53 ----------------------------------------------------- 3 files changed, 15 insertions(+), 68 deletions(-) diff --git a/fs/afs/inode.c b/fs/afs/inode.c index 299dbaeb2e2a..ade6ec3873cf 100644 --- a/fs/afs/inode.c +++ b/fs/afs/inode.c @@ -54,8 +54,21 @@ static int afs_inode_map_status(struct afs_vnode *vnode, struct key *key) inode->i_fop = &afs_dir_file_operations; break; case AFS_FTYPE_SYMLINK: - inode->i_mode = S_IFLNK | vnode->status.mode; - inode->i_op = &page_symlink_inode_operations; + /* Symlinks with a mode of 0644 are actually mountpoints. */ + if ((vnode->status.mode & 0777) == 0644) { + inode->i_flags |= S_AUTOMOUNT; + + spin_lock(&vnode->lock); + set_bit(AFS_VNODE_MOUNTPOINT, &vnode->flags); + spin_unlock(&vnode->lock); + + inode->i_mode = S_IFDIR | 0555; + inode->i_op = &afs_mntpt_inode_operations; + inode->i_fop = &afs_mntpt_file_operations; + } else { + inode->i_mode = S_IFLNK | vnode->status.mode; + inode->i_op = &page_symlink_inode_operations; + } inode_nohighmem(inode); break; default: @@ -79,18 +92,6 @@ static int afs_inode_map_status(struct afs_vnode *vnode, struct key *key) inode->i_generation = vnode->fid.unique; inode->i_version = vnode->status.data_version; inode->i_mapping->a_ops = &afs_fs_aops; - - /* check to see whether a symbolic link is really a mountpoint */ - if (vnode->status.type == AFS_FTYPE_SYMLINK) { - afs_mntpt_check_symlink(vnode, key); - - if (test_bit(AFS_VNODE_MOUNTPOINT, &vnode->flags)) { - inode->i_mode = S_IFDIR | vnode->status.mode; - inode->i_op = &afs_mntpt_inode_operations; - inode->i_fop = &afs_mntpt_file_operations; - } - } - return 0; } diff --git a/fs/afs/internal.h b/fs/afs/internal.h index af1d91ec7f2c..39de154fb42e 100644 --- a/fs/afs/internal.h +++ b/fs/afs/internal.h @@ -559,7 +559,6 @@ extern const struct inode_operations afs_autocell_inode_operations; extern const struct file_operations afs_mntpt_file_operations; extern struct vfsmount *afs_d_automount(struct path *); -extern int afs_mntpt_check_symlink(struct afs_vnode *, struct key *); extern void afs_mntpt_kill_timer(void); /* diff --git a/fs/afs/mntpt.c b/fs/afs/mntpt.c index d4fb0afc0097..bd3b65cde282 100644 --- a/fs/afs/mntpt.c +++ b/fs/afs/mntpt.c @@ -46,59 +46,6 @@ static DECLARE_DELAYED_WORK(afs_mntpt_expiry_timer, afs_mntpt_expiry_timed_out); static unsigned long afs_mntpt_expiry_timeout = 10 * 60; -/* - * check a symbolic link to see whether it actually encodes a mountpoint - * - sets the AFS_VNODE_MOUNTPOINT flag on the vnode appropriately - */ -int afs_mntpt_check_symlink(struct afs_vnode *vnode, struct key *key) -{ - struct page *page; - size_t size; - char *buf; - int ret; - - _enter("{%x:%u,%u}", - vnode->fid.vid, vnode->fid.vnode, vnode->fid.unique); - - /* read the contents of the symlink into the pagecache */ - page = read_cache_page(AFS_VNODE_TO_I(vnode)->i_mapping, 0, - afs_page_filler, key); - if (IS_ERR(page)) { - ret = PTR_ERR(page); - goto out; - } - - ret = -EIO; - if (PageError(page)) - goto out_free; - - buf = kmap(page); - - /* examine the symlink's contents */ - size = vnode->status.size; - _debug("symlink to %*.*s", (int) size, (int) size, buf); - - if (size > 2 && - (buf[0] == '%' || buf[0] == '#') && - buf[size - 1] == '.' - ) { - _debug("symlink is a mountpoint"); - spin_lock(&vnode->lock); - set_bit(AFS_VNODE_MOUNTPOINT, &vnode->flags); - vnode->vfs_inode.i_flags |= S_AUTOMOUNT; - spin_unlock(&vnode->lock); - } - - ret = 0; - - kunmap(page); -out_free: - put_page(page); -out: - _leave(" = %d", ret); - return ret; -} - /* * no valid lookup procedure on this sort of dir */ -- cgit From 1d7e4ebf291d3103466006926329e39aa355944f Mon Sep 17 00:00:00 2001 From: Andreea-Cristina Bernat Date: Thu, 16 Mar 2017 16:27:45 +0000 Subject: afs: inode: Replace rcu_assign_pointer() with RCU_INIT_POINTER() The use of "rcu_assign_pointer()" is NULLing out the pointer. According to RCU_INIT_POINTER()'s block comment: "1. This use of RCU_INIT_POINTER() is NULLing out the pointer" it is better to use it instead of rcu_assign_pointer() because it has a smaller overhead. The following Coccinelle semantic patch was used: @@ @@ - rcu_assign_pointer + RCU_INIT_POINTER (..., NULL) Signed-off-by: Andreea-Cristina Bernat Signed-off-by: David Howells --- fs/afs/inode.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/fs/afs/inode.c b/fs/afs/inode.c index ade6ec3873cf..e083e086b7ca 100644 --- a/fs/afs/inode.c +++ b/fs/afs/inode.c @@ -445,7 +445,7 @@ void afs_evict_inode(struct inode *inode) mutex_lock(&vnode->permits_lock); permits = vnode->permits; - rcu_assign_pointer(vnode->permits, NULL); + RCU_INIT_POINTER(vnode->permits, NULL); mutex_unlock(&vnode->permits_lock); if (permits) call_rcu(&permits->rcu, afs_zap_permits); -- cgit From df8a09d1b8f9e693ec3f6b7e0162fc817f2cf0db Mon Sep 17 00:00:00 2001 From: Andreea-Cristina Bernat Date: Thu, 16 Mar 2017 16:27:45 +0000 Subject: afs: security: Replace rcu_assign_pointer() with RCU_INIT_POINTER() The use of "rcu_assign_pointer()" is NULLing out the pointer. According to RCU_INIT_POINTER()'s block comment: "1. This use of RCU_INIT_POINTER() is NULLing out the pointer" it is better to use it instead of rcu_assign_pointer() because it has a smaller overhead. The following Coccinelle semantic patch was used: @@ @@ - rcu_assign_pointer + RCU_INIT_POINTER (..., NULL) Signed-off-by: Andreea-Cristina Bernat Signed-off-by: David Howells --- fs/afs/security.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/fs/afs/security.c b/fs/afs/security.c index bfa9d3428383..ecb86a670180 100644 --- a/fs/afs/security.c +++ b/fs/afs/security.c @@ -114,7 +114,7 @@ void afs_clear_permits(struct afs_vnode *vnode) mutex_lock(&vnode->permits_lock); permits = vnode->permits; - rcu_assign_pointer(vnode->permits, NULL); + RCU_INIT_POINTER(vnode->permits, NULL); mutex_unlock(&vnode->permits_lock); if (permits) -- cgit From 8a79790bf0b7da216627ffb85f52cfb4adbf1e4e Mon Sep 17 00:00:00 2001 From: Tina Ruchandani Date: Thu, 16 Mar 2017 16:27:46 +0000 Subject: afs: Migrate vlocation fields to 64-bit get_seconds() returns real wall-clock seconds. On 32-bit systems this value will overflow in year 2038 and beyond. This patch changes afs's vlocation record to use ktime_get_real_seconds() instead, for the fields time_of_death and update_at. Signed-off-by: Tina Ruchandani Signed-off-by: David Howells --- fs/afs/callback.c | 7 ++++--- fs/afs/internal.h | 7 ++++--- fs/afs/server.c | 6 +++--- fs/afs/vlocation.c | 16 +++++++++------- 4 files changed, 20 insertions(+), 16 deletions(-) diff --git a/fs/afs/callback.c b/fs/afs/callback.c index b29447e03ede..25d404d22cae 100644 --- a/fs/afs/callback.c +++ b/fs/afs/callback.c @@ -362,7 +362,7 @@ static void afs_callback_updater(struct work_struct *work) { struct afs_server *server; struct afs_vnode *vnode, *xvnode; - time_t now; + time64_t now; long timeout; int ret; @@ -370,7 +370,7 @@ static void afs_callback_updater(struct work_struct *work) _enter(""); - now = get_seconds(); + now = ktime_get_real_seconds(); /* find the first vnode to update */ spin_lock(&server->cb_lock); @@ -424,7 +424,8 @@ static void afs_callback_updater(struct work_struct *work) /* and then reschedule */ _debug("reschedule"); - vnode->update_at = get_seconds() + afs_vnode_update_timeout; + vnode->update_at = ktime_get_real_seconds() + + afs_vnode_update_timeout; spin_lock(&server->cb_lock); diff --git a/fs/afs/internal.h b/fs/afs/internal.h index 39de154fb42e..97a16ce200be 100644 --- a/fs/afs/internal.h +++ b/fs/afs/internal.h @@ -11,6 +11,7 @@ #include #include +#include #include #include #include @@ -249,7 +250,7 @@ struct afs_cache_vhash { */ struct afs_vlocation { atomic_t usage; - time_t time_of_death; /* time at which put reduced usage to 0 */ + time64_t time_of_death; /* time at which put reduced usage to 0 */ struct list_head link; /* link in cell volume location list */ struct list_head grave; /* link in master graveyard list */ struct list_head update; /* link in master update list */ @@ -260,7 +261,7 @@ struct afs_vlocation { struct afs_cache_vlocation vldb; /* volume information DB record */ struct afs_volume *vols[3]; /* volume access record pointer (index by type) */ wait_queue_head_t waitq; /* status change waitqueue */ - time_t update_at; /* time at which record should be updated */ + time64_t update_at; /* time at which record should be updated */ spinlock_t lock; /* access lock */ afs_vlocation_state_t state; /* volume location state */ unsigned short upd_rej_cnt; /* ENOMEDIUM count during update */ @@ -273,7 +274,7 @@ struct afs_vlocation { */ struct afs_server { atomic_t usage; - time_t time_of_death; /* time at which put reduced usage to 0 */ + time64_t time_of_death; /* time at which put reduced usage to 0 */ struct in_addr addr; /* server address */ struct afs_cell *cell; /* cell in which server resides */ struct list_head link; /* link in cell's server list */ diff --git a/fs/afs/server.c b/fs/afs/server.c index d4066ab7dd55..c001b1f2455f 100644 --- a/fs/afs/server.c +++ b/fs/afs/server.c @@ -242,7 +242,7 @@ void afs_put_server(struct afs_server *server) spin_lock(&afs_server_graveyard_lock); if (atomic_read(&server->usage) == 0) { list_move_tail(&server->grave, &afs_server_graveyard); - server->time_of_death = get_seconds(); + server->time_of_death = ktime_get_real_seconds(); queue_delayed_work(afs_wq, &afs_server_reaper, afs_server_timeout * HZ); } @@ -277,9 +277,9 @@ static void afs_reap_server(struct work_struct *work) LIST_HEAD(corpses); struct afs_server *server; unsigned long delay, expiry; - time_t now; + time64_t now; - now = get_seconds(); + now = ktime_get_real_seconds(); spin_lock(&afs_server_graveyard_lock); while (!list_empty(&afs_server_graveyard)) { diff --git a/fs/afs/vlocation.c b/fs/afs/vlocation.c index d7d8dd8c0b31..37b7c3b342a6 100644 --- a/fs/afs/vlocation.c +++ b/fs/afs/vlocation.c @@ -340,7 +340,8 @@ static void afs_vlocation_queue_for_updates(struct afs_vlocation *vl) struct afs_vlocation *xvl; /* wait at least 10 minutes before updating... */ - vl->update_at = get_seconds() + afs_vlocation_update_timeout; + vl->update_at = ktime_get_real_seconds() + + afs_vlocation_update_timeout; spin_lock(&afs_vlocation_updates_lock); @@ -506,7 +507,7 @@ void afs_put_vlocation(struct afs_vlocation *vl) if (atomic_read(&vl->usage) == 0) { _debug("buried"); list_move_tail(&vl->grave, &afs_vlocation_graveyard); - vl->time_of_death = get_seconds(); + vl->time_of_death = ktime_get_real_seconds(); queue_delayed_work(afs_wq, &afs_vlocation_reap, afs_vlocation_timeout * HZ); @@ -543,11 +544,11 @@ static void afs_vlocation_reaper(struct work_struct *work) LIST_HEAD(corpses); struct afs_vlocation *vl; unsigned long delay, expiry; - time_t now; + time64_t now; _enter(""); - now = get_seconds(); + now = ktime_get_real_seconds(); spin_lock(&afs_vlocation_graveyard_lock); while (!list_empty(&afs_vlocation_graveyard)) { @@ -622,13 +623,13 @@ static void afs_vlocation_updater(struct work_struct *work) { struct afs_cache_vlocation vldb; struct afs_vlocation *vl, *xvl; - time_t now; + time64_t now; long timeout; int ret; _enter(""); - now = get_seconds(); + now = ktime_get_real_seconds(); /* find a record to update */ spin_lock(&afs_vlocation_updates_lock); @@ -684,7 +685,8 @@ static void afs_vlocation_updater(struct work_struct *work) /* and then reschedule */ _debug("reschedule"); - vl->update_at = get_seconds() + afs_vlocation_update_timeout; + vl->update_at = ktime_get_real_seconds() + + afs_vlocation_update_timeout; spin_lock(&afs_vlocation_updates_lock); -- cgit From 56e714312e7dbd6bb83b2f78d3ec19a404c7649f Mon Sep 17 00:00:00 2001 From: Tina Ruchandani Date: Thu, 16 Mar 2017 16:27:46 +0000 Subject: afs: Prevent callback expiry timer overflow get_seconds() returns real wall-clock seconds. On 32-bit systems this value will overflow in year 2038 and beyond. This patch changes afs_vnode record to use ktime_get_real_seconds() instead, for the fields cb_expires and cb_expires_at. Signed-off-by: Tina Ruchandani Signed-off-by: David Howells --- fs/afs/fsclient.c | 2 +- fs/afs/inode.c | 7 ++++--- fs/afs/internal.h | 4 ++-- 3 files changed, 7 insertions(+), 6 deletions(-) diff --git a/fs/afs/fsclient.c b/fs/afs/fsclient.c index 6f917dd1238c..c05452a09398 100644 --- a/fs/afs/fsclient.c +++ b/fs/afs/fsclient.c @@ -145,7 +145,7 @@ static void xdr_decode_AFSCallBack(const __be32 **_bp, struct afs_vnode *vnode) vnode->cb_version = ntohl(*bp++); vnode->cb_expiry = ntohl(*bp++); vnode->cb_type = ntohl(*bp++); - vnode->cb_expires = vnode->cb_expiry + get_seconds(); + vnode->cb_expires = vnode->cb_expiry + ktime_get_real_seconds(); *_bp = bp; } diff --git a/fs/afs/inode.c b/fs/afs/inode.c index e083e086b7ca..4079c832ff27 100644 --- a/fs/afs/inode.c +++ b/fs/afs/inode.c @@ -246,12 +246,13 @@ struct inode *afs_iget(struct super_block *sb, struct key *key, vnode->cb_version = 0; vnode->cb_expiry = 0; vnode->cb_type = 0; - vnode->cb_expires = get_seconds(); + vnode->cb_expires = ktime_get_real_seconds(); } else { vnode->cb_version = cb->version; vnode->cb_expiry = cb->expiry; vnode->cb_type = cb->type; - vnode->cb_expires = vnode->cb_expiry + get_seconds(); + vnode->cb_expires = vnode->cb_expiry + + ktime_get_real_seconds(); } } @@ -324,7 +325,7 @@ int afs_validate(struct afs_vnode *vnode, struct key *key) !test_bit(AFS_VNODE_CB_BROKEN, &vnode->flags) && !test_bit(AFS_VNODE_MODIFIED, &vnode->flags) && !test_bit(AFS_VNODE_ZAP_DATA, &vnode->flags)) { - if (vnode->cb_expires < get_seconds() + 10) { + if (vnode->cb_expires < ktime_get_real_seconds() + 10) { _debug("callback expired"); set_bit(AFS_VNODE_CB_BROKEN, &vnode->flags); } else { diff --git a/fs/afs/internal.h b/fs/afs/internal.h index 97a16ce200be..832555003d03 100644 --- a/fs/afs/internal.h +++ b/fs/afs/internal.h @@ -377,8 +377,8 @@ struct afs_vnode { struct rb_node server_rb; /* link in server->fs_vnodes */ struct rb_node cb_promise; /* link in server->cb_promises */ struct work_struct cb_broken_work; /* work to be done on callback break */ - time_t cb_expires; /* time at which callback expires */ - time_t cb_expires_at; /* time used to order cb_promise */ + time64_t cb_expires; /* time at which callback expires */ + time64_t cb_expires_at; /* time used to order cb_promise */ unsigned cb_version; /* callback version */ unsigned cb_expiry; /* callback expiry time */ afs_callback_type_t cb_type; /* type of callback */ -- cgit From 29f069853287dcb46eaf45a50dbf1232c1444ac6 Mon Sep 17 00:00:00 2001 From: David Howells Date: Thu, 16 Mar 2017 16:27:46 +0000 Subject: afs: Fix AFS read bug Fix a bug in AFS read whereby the request page afs_read::index isn't incremented after calling ->page_done() if ->remain reaches 0, indicating that the data read is complete. Without this a NULL pointer exception happens when ->page_done() is called twice for the last page because the page clearing loop will call it also and afs_readpages_page_done() clears the current entry in the page list. BUG: unable to handle kernel NULL pointer dereference at (null) IP: afs_readpages_page_done+0x21/0xa4 [kafs] PGD 0 Oops: 0002 [#1] SMP Modules linked in: kafs(E) CPU: 2 PID: 3002 Comm: md5sum Tainted: G E 4.10.0-fscache #485 Hardware name: ASUS All Series/H97-PLUS, BIOS 2306 10/09/2014 task: ffff8804017d86c0 task.stack: ffff8803fc1d8000 RIP: 0010:afs_readpages_page_done+0x21/0xa4 [kafs] RSP: 0018:ffff8803fc1db978 EFLAGS: 00010282 RAX: ffff880405d39af8 RBX: 0000000000000000 RCX: ffff880407d83ed4 RDX: 0000000000000000 RSI: ffff880405d39a00 RDI: ffff880405c6f400 RBP: ffff8803fc1db988 R08: 0000000000000000 R09: 0000000000000001 R10: ffff8803fc1db820 R11: ffff88040cf56000 R12: ffff8804088f1780 R13: ffff8804017d86c0 R14: ffff8804088f1780 R15: 0000000000003840 FS: 00007f8154469700(0000) GS:ffff88041fb00000(0000) knlGS:0000000000000000 CS: 0010 DS: 0000 ES: 0000 CR0: 0000000080050033 CR2: 0000000000000000 CR3: 00000004016ec000 CR4: 00000000001406e0 Call Trace: afs_deliver_fs_fetch_data+0x5b9/0x60e [kafs] ? afs_make_call+0x316/0x4e8 [kafs] ? afs_make_call+0x359/0x4e8 [kafs] afs_deliver_to_call+0x173/0x2e8 [kafs] ? afs_make_call+0x316/0x4e8 [kafs] afs_make_call+0x37a/0x4e8 [kafs] ? wake_up_q+0x4f/0x4f ? __init_waitqueue_head+0x36/0x49 afs_fs_fetch_data+0x21c/0x227 [kafs] ? afs_fs_fetch_data+0x21c/0x227 [kafs] afs_vnode_fetch_data+0xf3/0x1d2 [kafs] afs_readpages+0x314/0x3fd [kafs] __do_page_cache_readahead+0x208/0x2c5 ondemand_readahead+0x3a2/0x3b7 ? ondemand_readahead+0x3a2/0x3b7 page_cache_async_readahead+0x5e/0x67 generic_file_read_iter+0x23b/0x70c ? __inode_security_revalidate+0x2f/0x62 __vfs_read+0xc4/0xe8 vfs_read+0xd1/0x15a SyS_read+0x4c/0x89 do_syscall_64+0x80/0x191 entry_SYSCALL64_slow_path+0x25/0x25 Reported-by: Marc Dionne Signed-off-by: David Howells Tested-by: Marc Dionne --- fs/afs/fsclient.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/fs/afs/fsclient.c b/fs/afs/fsclient.c index c05452a09398..4314f9e63a2c 100644 --- a/fs/afs/fsclient.c +++ b/fs/afs/fsclient.c @@ -390,9 +390,9 @@ static int afs_deliver_fs_fetch_data(struct afs_call *call) if (call->offset == PAGE_SIZE) { if (req->page_done) req->page_done(call, req); + req->index++; if (req->remain > 0) { call->offset = 0; - req->index++; if (req->index >= req->nr_pages) { call->unmarshall = 4; goto begin_discard; -- cgit From 6a0e3999e5cb3daa0468073fcdee0767422a4056 Mon Sep 17 00:00:00 2001 From: David Howells Date: Thu, 16 Mar 2017 16:27:46 +0000 Subject: afs: Make struct afs_read::remain 64-bit Make struct afs_read::remain 64-bit so that it can handle huge transfers if we ever request them or the server decides to give us a bit extra data (the other fields there are already 64-bit). Signed-off-by: David Howells Tested-by: Marc Dionne --- fs/afs/fsclient.c | 8 ++++---- fs/afs/internal.h | 2 +- 2 files changed, 5 insertions(+), 5 deletions(-) diff --git a/fs/afs/fsclient.c b/fs/afs/fsclient.c index 4314f9e63a2c..0778c5b6b59b 100644 --- a/fs/afs/fsclient.c +++ b/fs/afs/fsclient.c @@ -321,7 +321,7 @@ static int afs_deliver_fs_fetch_data(struct afs_call *call) void *buffer; int ret; - _enter("{%u,%zu/%u;%u/%llu}", + _enter("{%u,%zu/%u;%llu/%llu}", call->unmarshall, call->offset, call->count, req->remain, req->actual_len); @@ -379,7 +379,7 @@ static int afs_deliver_fs_fetch_data(struct afs_call *call) /* extract the returned data */ case 3: - _debug("extract data %u/%llu %zu/%u", + _debug("extract data %llu/%llu %zu/%u", req->remain, req->actual_len, call->offset, call->count); buffer = kmap(req->pages[req->index]); @@ -405,9 +405,9 @@ static int afs_deliver_fs_fetch_data(struct afs_call *call) /* Discard any excess data the server gave us */ begin_discard: case 4: - size = min_t(size_t, sizeof(afs_discard_buffer), req->remain); + size = min_t(loff_t, sizeof(afs_discard_buffer), req->remain); call->count = size; - _debug("extract discard %u/%llu %zu/%u", + _debug("extract discard %llu/%llu %zu/%u", req->remain, req->actual_len, call->offset, call->count); call->offset = 0; diff --git a/fs/afs/internal.h b/fs/afs/internal.h index 832555003d03..a6901360fb81 100644 --- a/fs/afs/internal.h +++ b/fs/afs/internal.h @@ -133,8 +133,8 @@ struct afs_read { loff_t pos; /* Where to start reading */ loff_t len; /* How much we're asking for */ loff_t actual_len; /* How much we're actually getting */ + loff_t remain; /* Amount remaining */ atomic_t usage; - unsigned int remain; /* Amount remaining */ unsigned int index; /* Which page we're reading into */ unsigned int nr_pages; void (*page_done)(struct afs_call *, struct afs_read *); -- cgit From 2f5705a5c805e7f761f2228820656bb9363a3d8c Mon Sep 17 00:00:00 2001 From: David Howells Date: Thu, 16 Mar 2017 16:27:46 +0000 Subject: afs: Use a bvec rather than a kvec in afs_send_pages() Use a bvec rather than a kvec in afs_send_pages() as we don't then have to call kmap() in advance. This allows us to pass the array of contiguous pages that we extracted through to rxrpc in one go rather than passing a single page at a time. Signed-off-by: David Howells --- fs/afs/rxrpc.c | 97 +++++++++++++++++++++++++++++++--------------------------- 1 file changed, 52 insertions(+), 45 deletions(-) diff --git a/fs/afs/rxrpc.c b/fs/afs/rxrpc.c index 419ef05dcb5e..bf45307ff201 100644 --- a/fs/afs/rxrpc.c +++ b/fs/afs/rxrpc.c @@ -259,67 +259,74 @@ void afs_flat_call_destructor(struct afs_call *call) call->buffer = NULL; } +#define AFS_BVEC_MAX 8 + +/* + * Load the given bvec with the next few pages. + */ +static void afs_load_bvec(struct afs_call *call, struct msghdr *msg, + struct bio_vec *bv, pgoff_t first, pgoff_t last, + unsigned offset) +{ + struct page *pages[AFS_BVEC_MAX]; + unsigned int nr, n, i, to, bytes = 0; + + nr = min_t(pgoff_t, last - first + 1, AFS_BVEC_MAX); + n = find_get_pages_contig(call->mapping, first, nr, pages); + ASSERTCMP(n, ==, nr); + + msg->msg_flags |= MSG_MORE; + for (i = 0; i < nr; i++) { + to = PAGE_SIZE; + if (first + i >= last) { + to = call->last_to; + msg->msg_flags &= ~MSG_MORE; + } + bv[i].bv_page = pages[i]; + bv[i].bv_len = to - offset; + bv[i].bv_offset = offset; + bytes += to - offset; + offset = 0; + } + + iov_iter_bvec(&msg->msg_iter, WRITE | ITER_BVEC, bv, nr, bytes); +} + /* * attach the data from a bunch of pages on an inode to a call */ static int afs_send_pages(struct afs_call *call, struct msghdr *msg) { - struct page *pages[8]; - unsigned count, n, loop, offset, to; + struct bio_vec bv[AFS_BVEC_MAX]; + unsigned int bytes, nr, loop, offset; pgoff_t first = call->first, last = call->last; int ret; - _enter(""); - offset = call->first_offset; call->first_offset = 0; do { - _debug("attach %lx-%lx", first, last); - - count = last - first + 1; - if (count > ARRAY_SIZE(pages)) - count = ARRAY_SIZE(pages); - n = find_get_pages_contig(call->mapping, first, count, pages); - ASSERTCMP(n, ==, count); - - loop = 0; - do { - struct bio_vec bvec = {.bv_page = pages[loop], - .bv_offset = offset}; - msg->msg_flags = 0; - to = PAGE_SIZE; - if (first + loop >= last) - to = call->last_to; - else - msg->msg_flags = MSG_MORE; - bvec.bv_len = to - offset; - offset = 0; - - _debug("- range %u-%u%s", - offset, to, msg->msg_flags ? " [more]" : ""); - iov_iter_bvec(&msg->msg_iter, WRITE | ITER_BVEC, - &bvec, 1, to - offset); - - /* have to change the state *before* sending the last - * packet as RxRPC might give us the reply before it - * returns from sending the request */ - if (first + loop >= last) - call->state = AFS_CALL_AWAIT_REPLY; - ret = rxrpc_kernel_send_data(afs_socket, call->rxcall, - msg, to - offset); - if (ret < 0) - break; - } while (++loop < count); - first += count; - - for (loop = 0; loop < count; loop++) - put_page(pages[loop]); + afs_load_bvec(call, msg, bv, first, last, offset); + offset = 0; + bytes = msg->msg_iter.count; + nr = msg->msg_iter.nr_segs; + + /* Have to change the state *before* sending the last + * packet as RxRPC might give us the reply before it + * returns from sending the request. + */ + if (first + nr >= last) + call->state = AFS_CALL_AWAIT_REPLY; + ret = rxrpc_kernel_send_data(afs_socket, call->rxcall, + msg, bytes); + for (loop = 0; loop < nr; loop++) + put_page(bv[loop].bv_page); if (ret < 0) break; + + first += nr; } while (first <= last); - _leave(" = %d", ret); return ret; } -- cgit From 146a1192783697810b63a1e41c4d59fc93387340 Mon Sep 17 00:00:00 2001 From: David Howells Date: Thu, 16 Mar 2017 16:27:47 +0000 Subject: afs: Fix the maths in afs_fs_store_data() afs_fs_store_data() works out of the size of the write it's going to make, but it uses 32-bit unsigned subtraction in one place that gets automatically cast to loff_t. However, if to < offset, then the number goes negative, but as the result isn't signed, this doesn't get sign-extended to 64-bits when placed in a loff_t. Fix by casting the operands to loff_t. Signed-off-by: David Howells --- fs/afs/fsclient.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/fs/afs/fsclient.c b/fs/afs/fsclient.c index 0778c5b6b59b..d9234b767287 100644 --- a/fs/afs/fsclient.c +++ b/fs/afs/fsclient.c @@ -1236,7 +1236,7 @@ int afs_fs_store_data(struct afs_server *server, struct afs_writeback *wb, _enter(",%x,{%x:%u},,", key_serial(wb->key), vnode->fid.vid, vnode->fid.vnode); - size = to - offset; + size = (loff_t)to - (loff_t)offset; if (first != last) size += (loff_t)(last - first) << PAGE_SHIFT; pos = (loff_t)first << PAGE_SHIFT; -- cgit From 1157f153f37a8586765034470e4f00a4a6c4ce6f Mon Sep 17 00:00:00 2001 From: David Howells Date: Thu, 16 Mar 2017 16:27:47 +0000 Subject: afs: Invalid op ID should abort with RXGEN_OPCODE When we are given an invalid operation ID, we should abort that with RXGEN_OPCODE rather than RX_INVALID_OPERATION. Also map RXGEN_OPCODE to -ENOTSUPP. Signed-off-by: David Howells --- fs/afs/misc.c | 2 ++ fs/afs/rxrpc.c | 2 +- 2 files changed, 3 insertions(+), 1 deletion(-) diff --git a/fs/afs/misc.c b/fs/afs/misc.c index 91ea1aa0d8b3..100b207efc9e 100644 --- a/fs/afs/misc.c +++ b/fs/afs/misc.c @@ -84,6 +84,8 @@ int afs_abort_to_error(u32 abort_code) case RXKADDATALEN: return -EKEYREJECTED; case RXKADILLEGALLEVEL: return -EKEYREJECTED; + case RXGEN_OPCODE: return -ENOTSUPP; + default: return -EREMOTEIO; } } diff --git a/fs/afs/rxrpc.c b/fs/afs/rxrpc.c index bf45307ff201..bf7761fe6ef5 100644 --- a/fs/afs/rxrpc.c +++ b/fs/afs/rxrpc.c @@ -465,7 +465,7 @@ static void afs_deliver_to_call(struct afs_call *call) abort_code, -ret, "KNC"); goto do_abort; case -ENOTSUPP: - abort_code = RX_INVALID_OPERATION; + abort_code = RXGEN_OPCODE; rxrpc_kernel_abort_call(afs_socket, call->rxcall, abort_code, -ret, "KIV"); goto do_abort; -- cgit From 70af0e3bd65142f9e674961c975451638a7ce1d5 Mon Sep 17 00:00:00 2001 From: David Howells Date: Thu, 16 Mar 2017 16:27:47 +0000 Subject: afs: Better abort and net error handling If we receive a network error, a remote abort or a protocol error whilst we're still transmitting data, make sure we return an appropriate error to the caller rather than ESHUTDOWN or ECONNABORTED. Signed-off-by: David Howells --- fs/afs/rxrpc.c | 35 +++++++++++++++++++++++++++-------- 1 file changed, 27 insertions(+), 8 deletions(-) diff --git a/fs/afs/rxrpc.c b/fs/afs/rxrpc.c index bf7761fe6ef5..22d26b369070 100644 --- a/fs/afs/rxrpc.c +++ b/fs/afs/rxrpc.c @@ -340,6 +340,8 @@ int afs_make_call(struct in_addr *addr, struct afs_call *call, gfp_t gfp, struct rxrpc_call *rxcall; struct msghdr msg; struct kvec iov[1]; + size_t offset; + u32 abort_code; int ret; _enter("%x,{%d},", addr->s_addr, ntohs(call->port)); @@ -388,9 +390,11 @@ int afs_make_call(struct in_addr *addr, struct afs_call *call, gfp_t gfp, msg.msg_controllen = 0; msg.msg_flags = (call->send_pages ? MSG_MORE : 0); - /* have to change the state *before* sending the last packet as RxRPC - * might give us the reply before it returns from sending the - * request */ + /* We have to change the state *before* sending the last packet as + * rxrpc might give us the reply before it returns from sending the + * request. Further, if the send fails, we may already have been given + * a notification and may have collected it. + */ if (!call->send_pages) call->state = AFS_CALL_AWAIT_REPLY; ret = rxrpc_kernel_send_data(afs_socket, rxcall, @@ -412,7 +416,17 @@ int afs_make_call(struct in_addr *addr, struct afs_call *call, gfp_t gfp, return afs_wait_for_call_to_complete(call); error_do_abort: - rxrpc_kernel_abort_call(afs_socket, rxcall, RX_USER_ABORT, -ret, "KSD"); + call->state = AFS_CALL_COMPLETE; + if (ret != -ECONNABORTED) { + rxrpc_kernel_abort_call(afs_socket, rxcall, RX_USER_ABORT, + -ret, "KSD"); + } else { + abort_code = 0; + offset = 0; + rxrpc_kernel_recv_data(afs_socket, rxcall, NULL, 0, &offset, + false, &abort_code); + ret = call->type->abort_to_error(abort_code); + } error_kill_call: afs_put_call(call); _leave(" = %d", ret); @@ -459,16 +473,18 @@ static void afs_deliver_to_call(struct afs_call *call) case -EINPROGRESS: case -EAGAIN: goto out; + case -ECONNABORTED: + goto call_complete; case -ENOTCONN: abort_code = RX_CALL_DEAD; rxrpc_kernel_abort_call(afs_socket, call->rxcall, abort_code, -ret, "KNC"); - goto do_abort; + goto save_error; case -ENOTSUPP: abort_code = RXGEN_OPCODE; rxrpc_kernel_abort_call(afs_socket, call->rxcall, abort_code, -ret, "KIV"); - goto do_abort; + goto save_error; case -ENODATA: case -EBADMSG: case -EMSGSIZE: @@ -478,7 +494,7 @@ static void afs_deliver_to_call(struct afs_call *call) abort_code = RXGEN_SS_UNMARSHAL; rxrpc_kernel_abort_call(afs_socket, call->rxcall, abort_code, EBADMSG, "KUM"); - goto do_abort; + goto save_error; } } @@ -489,8 +505,9 @@ out: _leave(""); return; -do_abort: +save_error: call->error = ret; +call_complete: call->state = AFS_CALL_COMPLETE; goto done; } @@ -538,6 +555,8 @@ static int afs_wait_for_call_to_complete(struct afs_call *call) _debug("call incomplete"); rxrpc_kernel_abort_call(afs_socket, call->rxcall, RX_CALL_DEAD, -ret, abort_why); + } else if (call->error < 0) { + ret = call->error; } _debug("call complete"); -- cgit From ab94f5d0dd6fd82e7eeca5e7c8096eaea0a0261f Mon Sep 17 00:00:00 2001 From: Marc Dionne Date: Thu, 16 Mar 2017 16:27:47 +0000 Subject: afs: Populate and use client modification time The inode timestamps should be set from the client time in the status received from the server, rather than the server time which is meant for internal server use. Set AFS_SET_MTIME and populate the mtime for operations that take an input status, such as file/dir creation and StoreData. If an input time is not provided the server will set the vnode times based on the current server time. In a situation where the server has some skew with the client, this could lead to the client seeing a timestamp in the future for a file that it just created or wrote. Signed-off-by: Marc Dionne Signed-off-by: David Howells --- fs/afs/fsclient.c | 18 +++++++++--------- fs/afs/inode.c | 2 +- 2 files changed, 10 insertions(+), 10 deletions(-) diff --git a/fs/afs/fsclient.c b/fs/afs/fsclient.c index d9234b767287..19f76ae36982 100644 --- a/fs/afs/fsclient.c +++ b/fs/afs/fsclient.c @@ -111,7 +111,7 @@ static void xdr_decode_AFSFetchStatus(const __be32 **_bp, vnode->vfs_inode.i_mode = mode; } - vnode->vfs_inode.i_ctime.tv_sec = status->mtime_server; + vnode->vfs_inode.i_ctime.tv_sec = status->mtime_client; vnode->vfs_inode.i_mtime = vnode->vfs_inode.i_ctime; vnode->vfs_inode.i_atime = vnode->vfs_inode.i_ctime; vnode->vfs_inode.i_version = data_version; @@ -734,8 +734,8 @@ int afs_fs_create(struct afs_server *server, memset(bp, 0, padsz); bp = (void *) bp + padsz; } - *bp++ = htonl(AFS_SET_MODE); - *bp++ = 0; /* mtime */ + *bp++ = htonl(AFS_SET_MODE | AFS_SET_MTIME); + *bp++ = htonl(vnode->vfs_inode.i_mtime.tv_sec); /* mtime */ *bp++ = 0; /* owner */ *bp++ = 0; /* group */ *bp++ = htonl(mode & S_IALLUGO); /* unix mode */ @@ -1003,8 +1003,8 @@ int afs_fs_symlink(struct afs_server *server, memset(bp, 0, c_padsz); bp = (void *) bp + c_padsz; } - *bp++ = htonl(AFS_SET_MODE); - *bp++ = 0; /* mtime */ + *bp++ = htonl(AFS_SET_MODE | AFS_SET_MTIME); + *bp++ = htonl(vnode->vfs_inode.i_mtime.tv_sec); /* mtime */ *bp++ = 0; /* owner */ *bp++ = 0; /* group */ *bp++ = htonl(S_IRWXUGO); /* unix mode */ @@ -1203,8 +1203,8 @@ static int afs_fs_store_data64(struct afs_server *server, *bp++ = htonl(vnode->fid.vnode); *bp++ = htonl(vnode->fid.unique); - *bp++ = 0; /* mask */ - *bp++ = 0; /* mtime */ + *bp++ = htonl(AFS_SET_MTIME); /* mask */ + *bp++ = htonl(vnode->vfs_inode.i_mtime.tv_sec); /* mtime */ *bp++ = 0; /* owner */ *bp++ = 0; /* group */ *bp++ = 0; /* unix mode */ @@ -1280,8 +1280,8 @@ int afs_fs_store_data(struct afs_server *server, struct afs_writeback *wb, *bp++ = htonl(vnode->fid.vnode); *bp++ = htonl(vnode->fid.unique); - *bp++ = 0; /* mask */ - *bp++ = 0; /* mtime */ + *bp++ = htonl(AFS_SET_MTIME); /* mask */ + *bp++ = htonl(vnode->vfs_inode.i_mtime.tv_sec); /* mtime */ *bp++ = 0; /* owner */ *bp++ = 0; /* group */ *bp++ = 0; /* unix mode */ diff --git a/fs/afs/inode.c b/fs/afs/inode.c index 4079c832ff27..aae55dd15108 100644 --- a/fs/afs/inode.c +++ b/fs/afs/inode.c @@ -85,7 +85,7 @@ static int afs_inode_map_status(struct afs_vnode *vnode, struct key *key) inode->i_uid = vnode->status.owner; inode->i_gid = vnode->status.group; inode->i_size = vnode->status.size; - inode->i_ctime.tv_sec = vnode->status.mtime_server; + inode->i_ctime.tv_sec = vnode->status.mtime_client; inode->i_ctime.tv_nsec = 0; inode->i_atime = inode->i_mtime = inode->i_ctime; inode->i_blocks = 0; -- cgit From 68ae849d7e674b83610bc7fdf74b21621a09b9ac Mon Sep 17 00:00:00 2001 From: David Howells Date: Thu, 16 Mar 2017 16:27:48 +0000 Subject: afs: Don't set PG_error on local EINTR or ENOMEM when filling a page Don't set PG_error on a page if we get local EINTR or ENOMEM when filling a page for writing. Signed-off-by: David Howells --- fs/afs/file.c | 12 ++++++++++-- 1 file changed, 10 insertions(+), 2 deletions(-) diff --git a/fs/afs/file.c b/fs/afs/file.c index b5829443ff69..0d5b8508869b 100644 --- a/fs/afs/file.c +++ b/fs/afs/file.c @@ -212,7 +212,13 @@ int afs_page_filler(void *data, struct page *page) fscache_uncache_page(vnode->cache, page); #endif BUG_ON(PageFsCache(page)); - goto error; + + if (ret == -EINTR || + ret == -ENOMEM || + ret == -ERESTARTSYS || + ret == -EAGAIN) + goto error; + goto io_error; } SetPageUptodate(page); @@ -231,10 +237,12 @@ int afs_page_filler(void *data, struct page *page) _leave(" = 0"); return 0; +io_error: + SetPageError(page); + goto error; enomem: ret = -ENOMEM; error: - SetPageError(page); unlock_page(page); _leave(" = %d", ret); return ret; -- cgit From 6d06b0d25209c80e99c1e89700f1e09694a3766b Mon Sep 17 00:00:00 2001 From: David Howells Date: Thu, 16 Mar 2017 16:27:48 +0000 Subject: afs: Fix page leak in afs_write_begin() afs_write_begin() leaks a ref and a lock on a page if afs_fill_page() fails. Fix the leak by unlocking and releasing the page in the error path. Signed-off-by: David Howells --- fs/afs/write.c | 7 +++++-- 1 file changed, 5 insertions(+), 2 deletions(-) diff --git a/fs/afs/write.c b/fs/afs/write.c index f1450ea09406..6e13e96c3db0 100644 --- a/fs/afs/write.c +++ b/fs/afs/write.c @@ -154,12 +154,12 @@ int afs_write_begin(struct file *file, struct address_space *mapping, kfree(candidate); return -ENOMEM; } - *pagep = page; - /* page won't leak in error case: it eventually gets cleaned off LRU */ if (!PageUptodate(page) && len != PAGE_SIZE) { ret = afs_fill_page(vnode, key, pos & PAGE_MASK, PAGE_SIZE, page); if (ret < 0) { + unlock_page(page); + put_page(page); kfree(candidate); _leave(" = %d [prep]", ret); return ret; @@ -167,6 +167,9 @@ int afs_write_begin(struct file *file, struct address_space *mapping, SetPageUptodate(page); } + /* page won't leak in error case: it eventually gets cleaned off LRU */ + *pagep = page; + try_again: spin_lock(&vnode->writeback_lock); -- cgit From 7286a35e893176169b09715096a4aca557e2ccd2 Mon Sep 17 00:00:00 2001 From: David Howells Date: Thu, 16 Mar 2017 16:27:48 +0000 Subject: afs: Fix afs_kill_pages() Fix afs_kill_pages() in two ways: (1) If a writeback has been partially flushed, then if we try and kill the pages it contains, some of them may no longer be undergoing writeback and end_page_writeback() will assert. Fix this by checking to see whether the page in question is actually undergoing writeback before ending that writeback. (2) The loop that scans for pages to kill doesn't increase the first page index, and so the loop may not terminate, but it will try to process the same pages over and over again. Fix this by increasing the first page index to one after the last page we processed. Signed-off-by: David Howells --- fs/afs/write.c | 10 +++++++--- 1 file changed, 7 insertions(+), 3 deletions(-) diff --git a/fs/afs/write.c b/fs/afs/write.c index 6e13e96c3db0..134de0667898 100644 --- a/fs/afs/write.c +++ b/fs/afs/write.c @@ -321,10 +321,14 @@ static void afs_kill_pages(struct afs_vnode *vnode, bool error, ASSERTCMP(pv.nr, ==, count); for (loop = 0; loop < count; loop++) { - ClearPageUptodate(pv.pages[loop]); + struct page *page = pv.pages[loop]; + ClearPageUptodate(page); if (error) - SetPageError(pv.pages[loop]); - end_page_writeback(pv.pages[loop]); + SetPageError(page); + if (PageWriteback(page)) + end_page_writeback(page); + if (page->index >= first) + first = page->index + 1; } __pagevec_release(&pv); -- cgit From 445783d0ec173a52bef2e9b129de7d716a19b9fa Mon Sep 17 00:00:00 2001 From: David Howells Date: Thu, 16 Mar 2017 16:27:48 +0000 Subject: afs: Fix an off-by-one error in afs_send_pages() afs_send_pages() should only put the call into the AFS_CALL_AWAIT_REPLY state if it has sent all the pages - but the check it makes is incorrect and sometimes it will finish the loop early. Signed-off-by: David Howells --- fs/afs/rxrpc.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/fs/afs/rxrpc.c b/fs/afs/rxrpc.c index 22d26b369070..b12da6aa5412 100644 --- a/fs/afs/rxrpc.c +++ b/fs/afs/rxrpc.c @@ -315,7 +315,7 @@ static int afs_send_pages(struct afs_call *call, struct msghdr *msg) * packet as RxRPC might give us the reply before it * returns from sending the request. */ - if (first + nr >= last) + if (first + nr - 1 >= last) call->state = AFS_CALL_AWAIT_REPLY; ret = rxrpc_kernel_send_data(afs_socket, call->rxcall, msg, bytes); -- cgit From 954cd6dc02a65065aecb7150962c0870c5b0e322 Mon Sep 17 00:00:00 2001 From: David Howells Date: Thu, 16 Mar 2017 16:27:49 +0000 Subject: afs: Fix abort on signal while waiting for call completion Fix the way in which a call that's in progress and being waited for is aborted in the case that EINTR is detected. We should be sending RX_USER_ABORT rather than RX_CALL_DEAD as the abort code. Note that since the only two ways out of the loop are if the call completes or if a signal happens, the kill-the-call clause after the loop has finished can only happen in the case of EINTR. This means that we only have one abort case to deal with, not two, and the "KWC" case can never happen and so can be deleted. Note further that simply aborting the call isn't necessarily the best thing here since at this point: the request has been entirely sent and it's likely the server will do the operation anyway - whether we abort it or not. In future, we should punt the handling of the remainder of the call off to a background thread. Reported-by: Marc Dionne Signed-off-by: David Howells --- fs/afs/rxrpc.c | 19 ++++++------------- 1 file changed, 6 insertions(+), 13 deletions(-) diff --git a/fs/afs/rxrpc.c b/fs/afs/rxrpc.c index b12da6aa5412..8f76b13d5549 100644 --- a/fs/afs/rxrpc.c +++ b/fs/afs/rxrpc.c @@ -517,7 +517,6 @@ call_complete: */ static int afs_wait_for_call_to_complete(struct afs_call *call) { - const char *abort_why; int ret; DECLARE_WAITQUEUE(myself, current); @@ -536,13 +535,8 @@ static int afs_wait_for_call_to_complete(struct afs_call *call) continue; } - abort_why = "KWC"; - ret = call->error; - if (call->state == AFS_CALL_COMPLETE) - break; - abort_why = "KWI"; - ret = -EINTR; - if (signal_pending(current)) + if (call->state == AFS_CALL_COMPLETE || + signal_pending(current)) break; schedule(); } @@ -550,15 +544,14 @@ static int afs_wait_for_call_to_complete(struct afs_call *call) remove_wait_queue(&call->waitq, &myself); __set_current_state(TASK_RUNNING); - /* kill the call */ + /* Kill off the call if it's still live. */ if (call->state < AFS_CALL_COMPLETE) { - _debug("call incomplete"); + _debug("call interrupted"); rxrpc_kernel_abort_call(afs_socket, call->rxcall, - RX_CALL_DEAD, -ret, abort_why); - } else if (call->error < 0) { - ret = call->error; + RX_USER_ABORT, -EINTR, "KWI"); } + ret = call->error; _debug("call complete"); afs_put_call(call); _leave(" = %d", ret); -- cgit From 65a151094edeb04e8f5f6f1502028e2383e81bb8 Mon Sep 17 00:00:00 2001 From: David Howells Date: Thu, 16 Mar 2017 16:27:49 +0000 Subject: afs: ->writepage() shouldn't call clear_page_dirty_for_io() The ->writepage() op shouldn't call clear_page_dirty_for_io() as that has already been called by the caller. Fix afs_writepage() by moving the call out of afs_write_back_from_locked_page() to afs_writepages_region() where it is needed. Signed-off-by: David Howells --- fs/afs/write.c | 6 +++--- 1 file changed, 3 insertions(+), 3 deletions(-) diff --git a/fs/afs/write.c b/fs/afs/write.c index 134de0667898..e5f150bccfb5 100644 --- a/fs/afs/write.c +++ b/fs/afs/write.c @@ -231,7 +231,7 @@ flush_conflicting_wb: if (wb->state == AFS_WBACK_PENDING) wb->state = AFS_WBACK_CONFLICTING; spin_unlock(&vnode->writeback_lock); - if (PageDirty(page)) { + if (clear_page_dirty_for_io(page)) { ret = afs_write_back_from_locked_page(wb, page); if (ret < 0) { afs_put_writeback(candidate); @@ -353,8 +353,6 @@ static int afs_write_back_from_locked_page(struct afs_writeback *wb, _enter(",%lx", primary_page->index); count = 1; - if (!clear_page_dirty_for_io(primary_page)) - BUG(); if (test_set_page_writeback(primary_page)) BUG(); @@ -542,6 +540,8 @@ static int afs_writepages_region(struct address_space *mapping, wb->state = AFS_WBACK_WRITING; spin_unlock(&wb->vnode->writeback_lock); + if (!clear_page_dirty_for_io(page)) + BUG(); ret = afs_write_back_from_locked_page(wb, page); unlock_page(page); put_page(page); -- cgit From c5051c7bc777dffa5661569dec5997f432b9a34a Mon Sep 17 00:00:00 2001 From: David Howells Date: Thu, 16 Mar 2017 16:27:49 +0000 Subject: afs: Don't wait for page writeback with the page lock held Drop the page lock before waiting for page writeback. Signed-off-by: David Howells --- fs/afs/write.c | 9 ++++----- 1 file changed, 4 insertions(+), 5 deletions(-) diff --git a/fs/afs/write.c b/fs/afs/write.c index e5f150bccfb5..2d2fccd5044b 100644 --- a/fs/afs/write.c +++ b/fs/afs/write.c @@ -518,17 +518,16 @@ static int afs_writepages_region(struct address_space *mapping, */ lock_page(page); - if (page->mapping != mapping) { + if (page->mapping != mapping || !PageDirty(page)) { unlock_page(page); put_page(page); continue; } - if (wbc->sync_mode != WB_SYNC_NONE) - wait_on_page_writeback(page); - - if (PageWriteback(page) || !PageDirty(page)) { + if (PageWriteback(page)) { unlock_page(page); + if (wbc->sync_mode != WB_SYNC_NONE) + wait_on_page_writeback(page); put_page(page); continue; } -- cgit