diff options
Diffstat (limited to 'fs/jffs2')
| -rw-r--r-- | fs/jffs2/README.Locking | 5 | ||||
| -rw-r--r-- | fs/jffs2/build.c | 75 | ||||
| -rw-r--r-- | fs/jffs2/dir.c | 11 | ||||
| -rw-r--r-- | fs/jffs2/file.c | 39 | ||||
| -rw-r--r-- | fs/jffs2/gc.c | 81 | ||||
| -rw-r--r-- | fs/jffs2/jffs2_fs_sb.h | 2 | ||||
| -rw-r--r-- | fs/jffs2/nodelist.h | 6 | ||||
| -rw-r--r-- | fs/jffs2/nodemgmt.c | 4 | ||||
| -rw-r--r-- | fs/jffs2/wbuf.c | 6 |
9 files changed, 146 insertions, 83 deletions
diff --git a/fs/jffs2/README.Locking b/fs/jffs2/README.Locking index 3ea36554107f..8918ac905a3b 100644 --- a/fs/jffs2/README.Locking +++ b/fs/jffs2/README.Locking @@ -2,10 +2,6 @@ JFFS2 LOCKING DOCUMENTATION --------------------------- -At least theoretically, JFFS2 does not require the Big Kernel Lock -(BKL), which was always helpfully obtained for it by Linux 2.4 VFS -code. It has its own locking, as described below. - This document attempts to describe the existing locking rules for JFFS2. It is not expected to remain perfectly up to date, but ought to be fairly close. @@ -69,6 +65,7 @@ Ordering constraints: any f->sem held. 2. Never attempt to lock two file mutexes in one thread. No ordering rules have been made for doing so. + 3. Never lock a page cache page with f->sem held. erase_completion_lock spinlock diff --git a/fs/jffs2/build.c b/fs/jffs2/build.c index 0ae91ad6df2d..b288c8ae1236 100644 --- a/fs/jffs2/build.c +++ b/fs/jffs2/build.c @@ -50,7 +50,8 @@ next_inode(int *i, struct jffs2_inode_cache *ic, struct jffs2_sb_info *c) static void jffs2_build_inode_pass1(struct jffs2_sb_info *c, - struct jffs2_inode_cache *ic) + struct jffs2_inode_cache *ic, + int *dir_hardlinks) { struct jffs2_full_dirent *fd; @@ -69,19 +70,21 @@ static void jffs2_build_inode_pass1(struct jffs2_sb_info *c, dbg_fsbuild("child \"%s\" (ino #%u) of dir ino #%u doesn't exist!\n", fd->name, fd->ino, ic->ino); jffs2_mark_node_obsolete(c, fd->raw); + /* Clear the ic/raw union so it doesn't cause problems later. */ + fd->ic = NULL; continue; } + /* From this point, fd->raw is no longer used so we can set fd->ic */ + fd->ic = child_ic; + child_ic->pino_nlink++; + /* If we appear (at this stage) to have hard-linked directories, + * set a flag to trigger a scan later */ if (fd->type == DT_DIR) { - if (child_ic->pino_nlink) { - JFFS2_ERROR("child dir \"%s\" (ino #%u) of dir ino #%u appears to be a hard link\n", - fd->name, fd->ino, ic->ino); - /* TODO: What do we do about it? */ - } else { - child_ic->pino_nlink = ic->ino; - } - } else - child_ic->pino_nlink++; + child_ic->flags |= INO_FLAGS_IS_DIR; + if (child_ic->pino_nlink > 1) + *dir_hardlinks = 1; + } dbg_fsbuild("increased nlink for child \"%s\" (ino #%u)\n", fd->name, fd->ino); /* Can't free scan_dents so far. We might need them in pass 2 */ @@ -95,8 +98,7 @@ static void jffs2_build_inode_pass1(struct jffs2_sb_info *c, */ static int jffs2_build_filesystem(struct jffs2_sb_info *c) { - int ret; - int i; + int ret, i, dir_hardlinks = 0; struct jffs2_inode_cache *ic; struct jffs2_full_dirent *fd; struct jffs2_full_dirent *dead_fds = NULL; @@ -120,7 +122,7 @@ static int jffs2_build_filesystem(struct jffs2_sb_info *c) /* Now scan the directory tree, increasing nlink according to every dirent found. */ for_each_inode(i, c, ic) { if (ic->scan_dents) { - jffs2_build_inode_pass1(c, ic); + jffs2_build_inode_pass1(c, ic, &dir_hardlinks); cond_resched(); } } @@ -156,6 +158,20 @@ static int jffs2_build_filesystem(struct jffs2_sb_info *c) } dbg_fsbuild("pass 2a complete\n"); + + if (dir_hardlinks) { + /* If we detected directory hardlinks earlier, *hopefully* + * they are gone now because some of the links were from + * dead directories which still had some old dirents lying + * around and not yet garbage-collected, but which have + * been discarded above. So clear the pino_nlink field + * in each directory, so that the final scan below can + * print appropriate warnings. */ + for_each_inode(i, c, ic) { + if (ic->flags & INO_FLAGS_IS_DIR) + ic->pino_nlink = 0; + } + } dbg_fsbuild("freeing temporary data structures\n"); /* Finally, we can scan again and free the dirent structs */ @@ -163,6 +179,33 @@ static int jffs2_build_filesystem(struct jffs2_sb_info *c) while(ic->scan_dents) { fd = ic->scan_dents; ic->scan_dents = fd->next; + /* We do use the pino_nlink field to count nlink of + * directories during fs build, so set it to the + * parent ino# now. Now that there's hopefully only + * one. */ + if (fd->type == DT_DIR) { + if (!fd->ic) { + /* We'll have complained about it and marked the coresponding + raw node obsolete already. Just skip it. */ + continue; + } + + /* We *have* to have set this in jffs2_build_inode_pass1() */ + BUG_ON(!(fd->ic->flags & INO_FLAGS_IS_DIR)); + + /* We clear ic->pino_nlink ∀ directories' ic *only* if dir_hardlinks + * is set. Otherwise, we know this should never trigger anyway, so + * we don't do the check. And ic->pino_nlink still contains the nlink + * value (which is 1). */ + if (dir_hardlinks && fd->ic->pino_nlink) { + JFFS2_ERROR("child dir \"%s\" (ino #%u) of dir ino #%u is also hard linked from dir ino #%u\n", + fd->name, fd->ino, ic->ino, fd->ic->pino_nlink); + /* Should we unlink it from its previous parent? */ + } + + /* For directories, ic->pino_nlink holds that parent inode # */ + fd->ic->pino_nlink = ic->ino; + } jffs2_free_full_dirent(fd); } ic->scan_dents = NULL; @@ -241,11 +284,7 @@ static void jffs2_build_remove_unlinked_inode(struct jffs2_sb_info *c, /* Reduce nlink of the child. If it's now zero, stick it on the dead_fds list to be cleaned up later. Else just free the fd */ - - if (fd->type == DT_DIR) - child_ic->pino_nlink = 0; - else - child_ic->pino_nlink--; + child_ic->pino_nlink--; if (!child_ic->pino_nlink) { dbg_fsbuild("inode #%u (\"%s\") now has no links; adding to dead_fds list.\n", diff --git a/fs/jffs2/dir.c b/fs/jffs2/dir.c index d211b8e18566..30c4c9ebb693 100644 --- a/fs/jffs2/dir.c +++ b/fs/jffs2/dir.c @@ -843,9 +843,14 @@ static int jffs2_rename (struct inode *old_dir_i, struct dentry *old_dentry, pr_notice("%s(): Link succeeded, unlink failed (err %d). You now have a hard link\n", __func__, ret); - /* Might as well let the VFS know */ - d_instantiate(new_dentry, d_inode(old_dentry)); - ihold(d_inode(old_dentry)); + /* + * We can't keep the target in dcache after that. + * For one thing, we can't afford dentry aliases for directories. + * For another, if there was a victim, we _can't_ set new inode + * for that sucker and we have to trigger mount eviction - the + * caller won't do it on its own since we are returning an error. + */ + d_invalidate(new_dentry); new_dir_i->i_mtime = new_dir_i->i_ctime = ITIME(now); return ret; } diff --git a/fs/jffs2/file.c b/fs/jffs2/file.c index c5ac5944bc1b..cad86bac3453 100644 --- a/fs/jffs2/file.c +++ b/fs/jffs2/file.c @@ -137,39 +137,33 @@ static int jffs2_write_begin(struct file *filp, struct address_space *mapping, struct page *pg; struct inode *inode = mapping->host; struct jffs2_inode_info *f = JFFS2_INODE_INFO(inode); - struct jffs2_sb_info *c = JFFS2_SB_INFO(inode->i_sb); - struct jffs2_raw_inode ri; - uint32_t alloc_len = 0; pgoff_t index = pos >> PAGE_CACHE_SHIFT; uint32_t pageofs = index << PAGE_CACHE_SHIFT; int ret = 0; - jffs2_dbg(1, "%s()\n", __func__); - - if (pageofs > inode->i_size) { - ret = jffs2_reserve_space(c, sizeof(ri), &alloc_len, - ALLOC_NORMAL, JFFS2_SUMMARY_INODE_SIZE); - if (ret) - return ret; - } - - mutex_lock(&f->sem); pg = grab_cache_page_write_begin(mapping, index, flags); - if (!pg) { - if (alloc_len) - jffs2_complete_reservation(c); - mutex_unlock(&f->sem); + if (!pg) return -ENOMEM; - } *pagep = pg; - if (alloc_len) { + jffs2_dbg(1, "%s()\n", __func__); + + if (pageofs > inode->i_size) { /* Make new hole frag from old EOF to new page */ + struct jffs2_sb_info *c = JFFS2_SB_INFO(inode->i_sb); + struct jffs2_raw_inode ri; struct jffs2_full_dnode *fn; + uint32_t alloc_len; jffs2_dbg(1, "Writing new hole frag 0x%x-0x%x between current EOF and new page\n", (unsigned int)inode->i_size, pageofs); + ret = jffs2_reserve_space(c, sizeof(ri), &alloc_len, + ALLOC_NORMAL, JFFS2_SUMMARY_INODE_SIZE); + if (ret) + goto out_page; + + mutex_lock(&f->sem); memset(&ri, 0, sizeof(ri)); ri.magic = cpu_to_je16(JFFS2_MAGIC_BITMASK); @@ -196,6 +190,7 @@ static int jffs2_write_begin(struct file *filp, struct address_space *mapping, if (IS_ERR(fn)) { ret = PTR_ERR(fn); jffs2_complete_reservation(c); + mutex_unlock(&f->sem); goto out_page; } ret = jffs2_add_full_dnode_to_inode(c, f, fn); @@ -210,10 +205,12 @@ static int jffs2_write_begin(struct file *filp, struct address_space *mapping, jffs2_mark_node_obsolete(c, fn->raw); jffs2_free_full_dnode(fn); jffs2_complete_reservation(c); + mutex_unlock(&f->sem); goto out_page; } jffs2_complete_reservation(c); inode->i_size = pageofs; + mutex_unlock(&f->sem); } /* @@ -222,18 +219,18 @@ static int jffs2_write_begin(struct file *filp, struct address_space *mapping, * case of a short-copy. */ if (!PageUptodate(pg)) { + mutex_lock(&f->sem); ret = jffs2_do_readpage_nolock(inode, pg); + mutex_unlock(&f->sem); if (ret) goto out_page; } - mutex_unlock(&f->sem); jffs2_dbg(1, "end write_begin(). pg->flags %lx\n", pg->flags); return ret; out_page: unlock_page(pg); page_cache_release(pg); - mutex_unlock(&f->sem); return ret; } diff --git a/fs/jffs2/gc.c b/fs/jffs2/gc.c index 5a2dec2b064c..7e553f286775 100644 --- a/fs/jffs2/gc.c +++ b/fs/jffs2/gc.c @@ -134,37 +134,59 @@ int jffs2_garbage_collect_pass(struct jffs2_sb_info *c) if (mutex_lock_interruptible(&c->alloc_sem)) return -EINTR; + for (;;) { + /* We can't start doing GC until we've finished checking + the node CRCs etc. */ + int bucket, want_ino; + spin_lock(&c->erase_completion_lock); if (!c->unchecked_size) break; - - /* We can't start doing GC yet. We haven't finished checking - the node CRCs etc. Do it now. */ - - /* checked_ino is protected by the alloc_sem */ - if (c->checked_ino > c->highest_ino && xattr) { - pr_crit("Checked all inodes but still 0x%x bytes of unchecked space?\n", - c->unchecked_size); - jffs2_dbg_dump_block_lists_nolock(c); - spin_unlock(&c->erase_completion_lock); - mutex_unlock(&c->alloc_sem); - return -ENOSPC; - } - spin_unlock(&c->erase_completion_lock); if (!xattr) xattr = jffs2_verify_xattr(c); spin_lock(&c->inocache_lock); + /* Instead of doing the inodes in numeric order, doing a lookup + * in the hash for each possible number, just walk the hash + * buckets of *existing* inodes. This means that we process + * them out-of-order, but it can be a lot faster if there's + * a sparse inode# space. Which there often is. */ + want_ino = c->check_ino; + for (bucket = c->check_ino % c->inocache_hashsize ; bucket < c->inocache_hashsize; bucket++) { + for (ic = c->inocache_list[bucket]; ic; ic = ic->next) { + if (ic->ino < want_ino) + continue; + + if (ic->state != INO_STATE_CHECKEDABSENT && + ic->state != INO_STATE_PRESENT) + goto got_next; /* with inocache_lock held */ + + jffs2_dbg(1, "Skipping ino #%u already checked\n", + ic->ino); + } + want_ino = 0; + } - ic = jffs2_get_ino_cache(c, c->checked_ino++); + /* Point c->check_ino past the end of the last bucket. */ + c->check_ino = ((c->highest_ino + c->inocache_hashsize + 1) & + ~c->inocache_hashsize) - 1; - if (!ic) { - spin_unlock(&c->inocache_lock); - continue; - } + spin_unlock(&c->inocache_lock); + + pr_crit("Checked all inodes but still 0x%x bytes of unchecked space?\n", + c->unchecked_size); + jffs2_dbg_dump_block_lists_nolock(c); + mutex_unlock(&c->alloc_sem); + return -ENOSPC; + + got_next: + /* For next time round the loop, we want c->checked_ino to indicate + * the *next* one we want to check. And since we're walking the + * buckets rather than doing it sequentially, it's: */ + c->check_ino = ic->ino + c->inocache_hashsize; if (!ic->pino_nlink) { jffs2_dbg(1, "Skipping check of ino #%d with nlink/pino zero\n", @@ -176,8 +198,6 @@ int jffs2_garbage_collect_pass(struct jffs2_sb_info *c) switch(ic->state) { case INO_STATE_CHECKEDABSENT: case INO_STATE_PRESENT: - jffs2_dbg(1, "Skipping ino #%u already checked\n", - ic->ino); spin_unlock(&c->inocache_lock); continue; @@ -196,7 +216,7 @@ int jffs2_garbage_collect_pass(struct jffs2_sb_info *c) ic->ino); /* We need to come back again for the _same_ inode. We've made no progress in this case, but that should be OK */ - c->checked_ino--; + c->check_ino = ic->ino; mutex_unlock(&c->alloc_sem); sleep_on_spinunlock(&c->inocache_wq, &c->inocache_lock); @@ -1296,14 +1316,17 @@ static int jffs2_garbage_collect_dnode(struct jffs2_sb_info *c, struct jffs2_era BUG_ON(start > orig_start); } - /* First, use readpage() to read the appropriate page into the page cache */ - /* Q: What happens if we actually try to GC the _same_ page for which commit_write() - * triggered garbage collection in the first place? - * A: I _think_ it's OK. read_cache_page shouldn't deadlock, we'll write out the - * page OK. We'll actually write it out again in commit_write, which is a little - * suboptimal, but at least we're correct. - */ + /* The rules state that we must obtain the page lock *before* f->sem, so + * drop f->sem temporarily. Since we also hold c->alloc_sem, nothing's + * actually going to *change* so we're safe; we only allow reading. + * + * It is important to note that jffs2_write_begin() will ensure that its + * page is marked Uptodate before allocating space. That means that if we + * end up here trying to GC the *same* page that jffs2_write_begin() is + * trying to write out, read_cache_page() will not deadlock. */ + mutex_unlock(&f->sem); pg_ptr = jffs2_gc_fetch_page(c, f, start, &pg); + mutex_lock(&f->sem); if (IS_ERR(pg_ptr)) { pr_warn("read_cache_page() returned error: %ld\n", diff --git a/fs/jffs2/jffs2_fs_sb.h b/fs/jffs2/jffs2_fs_sb.h index 046fee8b6e9b..778275f48a87 100644 --- a/fs/jffs2/jffs2_fs_sb.h +++ b/fs/jffs2/jffs2_fs_sb.h @@ -49,7 +49,7 @@ struct jffs2_sb_info { struct mtd_info *mtd; uint32_t highest_ino; - uint32_t checked_ino; + uint32_t check_ino; /* *NEXT* inode to be checked */ unsigned int flags; diff --git a/fs/jffs2/nodelist.h b/fs/jffs2/nodelist.h index fa35ff79ab35..0637271f3770 100644 --- a/fs/jffs2/nodelist.h +++ b/fs/jffs2/nodelist.h @@ -194,6 +194,7 @@ struct jffs2_inode_cache { #define INO_STATE_CLEARING 6 /* In clear_inode() */ #define INO_FLAGS_XATTR_CHECKED 0x01 /* has no duplicate xattr_ref */ +#define INO_FLAGS_IS_DIR 0x02 /* is a directory */ #define RAWNODE_CLASS_INODE_CACHE 0 #define RAWNODE_CLASS_XATTR_DATUM 1 @@ -249,7 +250,10 @@ struct jffs2_readinode_info struct jffs2_full_dirent { - struct jffs2_raw_node_ref *raw; + union { + struct jffs2_raw_node_ref *raw; + struct jffs2_inode_cache *ic; /* Just during part of build */ + }; struct jffs2_full_dirent *next; uint32_t version; uint32_t ino; /* == zero for unlink */ diff --git a/fs/jffs2/nodemgmt.c b/fs/jffs2/nodemgmt.c index b6bd4affd9ad..cda0774c2c9c 100644 --- a/fs/jffs2/nodemgmt.c +++ b/fs/jffs2/nodemgmt.c @@ -846,8 +846,8 @@ int jffs2_thread_should_wake(struct jffs2_sb_info *c) return 1; if (c->unchecked_size) { - jffs2_dbg(1, "jffs2_thread_should_wake(): unchecked_size %d, checked_ino #%d\n", - c->unchecked_size, c->checked_ino); + jffs2_dbg(1, "jffs2_thread_should_wake(): unchecked_size %d, check_ino #%d\n", + c->unchecked_size, c->check_ino); return 1; } diff --git a/fs/jffs2/wbuf.c b/fs/jffs2/wbuf.c index 5a3da3f52908..b25d28a21212 100644 --- a/fs/jffs2/wbuf.c +++ b/fs/jffs2/wbuf.c @@ -1183,22 +1183,20 @@ void jffs2_dirty_trigger(struct jffs2_sb_info *c) int jffs2_nand_flash_setup(struct jffs2_sb_info *c) { - struct nand_ecclayout *oinfo = c->mtd->ecclayout; - if (!c->mtd->oobsize) return 0; /* Cleanmarker is out-of-band, so inline size zero */ c->cleanmarker_size = 0; - if (!oinfo || oinfo->oobavail == 0) { + if (c->mtd->oobavail == 0) { pr_err("inconsistent device description\n"); return -EINVAL; } jffs2_dbg(1, "using OOB on NAND\n"); - c->oobavail = oinfo->oobavail; + c->oobavail = c->mtd->oobavail; /* Initialise write buffer */ init_rwsem(&c->wbuf_sem); |
