jffs2: Fix lock acquisition order bug in jffs2_write_begin
Thomas Betker [Wed, 17 Oct 2012 20:59:30 +0000 (22:59 +0200)]
commit 5ffd3412ae5536a4c57469cb8ea31887121dcb2e upstream.

jffs2_write_begin() first acquires the page lock, then f->sem. This
causes an AB-BA deadlock with jffs2_garbage_collect_live(), which first
acquires f->sem, then the page lock:

jffs2_garbage_collect_live
    mutex_lock(&f->sem)                         (A)
    jffs2_garbage_collect_dnode
        jffs2_gc_fetch_page
            read_cache_page_async
                do_read_cache_page
                    lock_page(page)             (B)

jffs2_write_begin
    grab_cache_page_write_begin
        find_lock_page
            lock_page(page)                     (B)
    mutex_lock(&f->sem)                         (A)

We fix this by restructuring jffs2_write_begin() to take f->sem before
the page lock. However, we make sure that f->sem is not held when
calling jffs2_reserve_space(), as this is not permitted by the locking
rules.

The deadlock above was observed multiple times on an SoC with a dual
ARMv7 (Cortex-A9), running the long-term 3.4.11 kernel; it occurred
when using scp to copy files from a host system to the ARM target
system. The fix was heavily tested on the same target system.

Signed-off-by: Thomas Betker <thomas.betker@rohde-schwarz.com>
Acked-by: Joakim Tjernlund <Joakim.Tjernlund@transmode.se>
Signed-off-by: Artem Bityutskiy <artem.bityutskiy@linux.intel.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

fs/jffs2/file.c

index db3889b..8608f87 100644 (file)
@@ -138,33 +138,39 @@ 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 (!pg) {
+               if (alloc_len)
+                       jffs2_complete_reservation(c);
+               mutex_unlock(&f->sem);
                return -ENOMEM;
+       }
        *pagep = pg;
 
-       jffs2_dbg(1, "%s()\n", __func__);
-
-       if (pageofs > inode->i_size) {
+       if (alloc_len) {
                /* 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);
@@ -191,7 +197,6 @@ 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);
@@ -206,12 +211,10 @@ 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);
        }
 
        /*
@@ -220,18 +223,18 @@ 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;
 }