]> git.ipfire.org Git - thirdparty/kernel/linux.git/commitdiff
hfs: fix general protection fault in hfs_find_init()
authorViacheslav Dubeyko <slava@dubeyko.com>
Thu, 10 Jul 2025 21:36:57 +0000 (14:36 -0700)
committerViacheslav Dubeyko <slava@dubeyko.com>
Fri, 25 Jul 2025 22:40:16 +0000 (15:40 -0700)
The hfs_find_init() method can trigger the crash
if tree pointer is NULL:

[   45.746290][ T9787] Oops: general protection fault, probably for non-canonical address 0xdffffc0000000008: 0000 [#1] SMP KAI
[   45.747287][ T9787] KASAN: null-ptr-deref in range [0x0000000000000040-0x0000000000000047]
[   45.748716][ T9787] CPU: 2 UID: 0 PID: 9787 Comm: repro Not tainted 6.16.0-rc3 #10 PREEMPT(full)
[   45.750250][ T9787] Hardware name: QEMU Ubuntu 24.04 PC (i440FX + PIIX, 1996), BIOS 1.16.3-debian-1.16.3-2 04/01/2014
[   45.751983][ T9787] RIP: 0010:hfs_find_init+0x86/0x230
[   45.752834][ T9787] Code: c1 ea 03 80 3c 02 00 0f 85 9a 01 00 00 4c 8d 6b 40 48 c7 45 18 00 00 00 00 48 b8 00 00 00 00 00 fc
[   45.755574][ T9787] RSP: 0018:ffffc90015157668 EFLAGS: 00010202
[   45.756432][ T9787] RAX: dffffc0000000000 RBX: 0000000000000000 RCX: ffffffff819a4d09
[   45.757457][ T9787] RDX: 0000000000000008 RSI: ffffffff819acd3a RDI: ffffc900151576e8
[   45.758282][ T9787] RBP: ffffc900151576d0 R08: 0000000000000005 R09: 0000000000000000
[   45.758943][ T9787] R10: 0000000080000000 R11: 0000000000000001 R12: 0000000000000004
[   45.759619][ T9787] R13: 0000000000000040 R14: ffff88802c50814a R15: 0000000000000000
[   45.760293][ T9787] FS:  00007ffb72734540(0000) GS:ffff8880cec64000(0000) knlGS:0000000000000000
[   45.761050][ T9787] CS:  0010 DS: 0000 ES: 0000 CR0: 0000000080050033
[   45.761606][ T9787] CR2: 00007f9bd8225000 CR3: 000000010979a000 CR4: 00000000000006f0
[   45.762286][ T9787] Call Trace:
[   45.762570][ T9787]  <TASK>
[   45.762824][ T9787]  hfs_ext_read_extent+0x190/0x9d0
[   45.763269][ T9787]  ? submit_bio_noacct_nocheck+0x2dd/0xce0
[   45.763766][ T9787]  ? __pfx_hfs_ext_read_extent+0x10/0x10
[   45.764250][ T9787]  hfs_get_block+0x55f/0x830
[   45.764646][ T9787]  block_read_full_folio+0x36d/0x850
[   45.765105][ T9787]  ? __pfx_hfs_get_block+0x10/0x10
[   45.765541][ T9787]  ? const_folio_flags+0x5b/0x100
[   45.765972][ T9787]  ? __pfx_hfs_read_folio+0x10/0x10
[   45.766415][ T9787]  filemap_read_folio+0xbe/0x290
[   45.766840][ T9787]  ? __pfx_filemap_read_folio+0x10/0x10
[   45.767325][ T9787]  ? __filemap_get_folio+0x32b/0xbf0
[   45.767780][ T9787]  do_read_cache_folio+0x263/0x5c0
[   45.768223][ T9787]  ? __pfx_hfs_read_folio+0x10/0x10
[   45.768666][ T9787]  read_cache_page+0x5b/0x160
[   45.769070][ T9787]  hfs_btree_open+0x491/0x1740
[   45.769481][ T9787]  hfs_mdb_get+0x15e2/0x1fb0
[   45.769877][ T9787]  ? __pfx_hfs_mdb_get+0x10/0x10
[   45.770316][ T9787]  ? find_held_lock+0x2b/0x80
[   45.770731][ T9787]  ? lockdep_init_map_type+0x5c/0x280
[   45.771200][ T9787]  ? lockdep_init_map_type+0x5c/0x280
[   45.771674][ T9787]  hfs_fill_super+0x38e/0x720
[   45.772092][ T9787]  ? __pfx_hfs_fill_super+0x10/0x10
[   45.772549][ T9787]  ? snprintf+0xbe/0x100
[   45.772931][ T9787]  ? __pfx_snprintf+0x10/0x10
[   45.773350][ T9787]  ? do_raw_spin_lock+0x129/0x2b0
[   45.773796][ T9787]  ? find_held_lock+0x2b/0x80
[   45.774215][ T9787]  ? set_blocksize+0x40a/0x510
[   45.774636][ T9787]  ? sb_set_blocksize+0x176/0x1d0
[   45.775087][ T9787]  ? setup_bdev_super+0x369/0x730
[   45.775533][ T9787]  get_tree_bdev_flags+0x384/0x620
[   45.775985][ T9787]  ? __pfx_hfs_fill_super+0x10/0x10
[   45.776453][ T9787]  ? __pfx_get_tree_bdev_flags+0x10/0x10
[   45.776950][ T9787]  ? bpf_lsm_capable+0x9/0x10
[   45.777365][ T9787]  ? security_capable+0x80/0x260
[   45.777803][ T9787]  vfs_get_tree+0x8e/0x340
[   45.778203][ T9787]  path_mount+0x13de/0x2010
[   45.778604][ T9787]  ? kmem_cache_free+0x2b0/0x4c0
[   45.779052][ T9787]  ? __pfx_path_mount+0x10/0x10
[   45.779480][ T9787]  ? getname_flags.part.0+0x1c5/0x550
[   45.779954][ T9787]  ? putname+0x154/0x1a0
[   45.780335][ T9787]  __x64_sys_mount+0x27b/0x300
[   45.780758][ T9787]  ? __pfx___x64_sys_mount+0x10/0x10
[   45.781232][ T9787]  do_syscall_64+0xc9/0x480
[   45.781631][ T9787]  entry_SYSCALL_64_after_hwframe+0x77/0x7f
[   45.782149][ T9787] RIP: 0033:0x7ffb7265b6ca
[   45.782539][ T9787] Code: 48 8b 0d c9 17 0d 00 f7 d8 64 89 01 48 83 c8 ff c3 66 2e 0f 1f 84 00 00 00 00 00 0f 1f 44 00 00 48
[   45.784212][ T9787] RSP: 002b:00007ffc0c10cfb8 EFLAGS: 00000206 ORIG_RAX: 00000000000000a5
[   45.784935][ T9787] RAX: ffffffffffffffda RBX: 0000000000000000 RCX: 00007ffb7265b6ca
[   45.785626][ T9787] RDX: 0000200000000240 RSI: 0000200000000280 RDI: 00007ffc0c10d100
[   45.786316][ T9787] RBP: 00007ffc0c10d190 R08: 00007ffc0c10d000 R09: 0000000000000000
[   45.787011][ T9787] R10: 0000000000000048 R11: 0000000000000206 R12: 0000560246733250
[   45.787697][ T9787] R13: 0000000000000000 R14: 0000000000000000 R15: 0000000000000000
[   45.788393][ T9787]  </TASK>
[   45.788665][ T9787] Modules linked in:
[   45.789058][ T9787] ---[ end trace 0000000000000000 ]---
[   45.789554][ T9787] RIP: 0010:hfs_find_init+0x86/0x230
[   45.790028][ T9787] Code: c1 ea 03 80 3c 02 00 0f 85 9a 01 00 00 4c 8d 6b 40 48 c7 45 18 00 00 00 00 48 b8 00 00 00 00 00 fc
[   45.792364][ T9787] RSP: 0018:ffffc90015157668 EFLAGS: 00010202
[   45.793155][ T9787] RAX: dffffc0000000000 RBX: 0000000000000000 RCX: ffffffff819a4d09
[   45.794123][ T9787] RDX: 0000000000000008 RSI: ffffffff819acd3a RDI: ffffc900151576e8
[   45.795105][ T9787] RBP: ffffc900151576d0 R08: 0000000000000005 R09: 0000000000000000
[   45.796135][ T9787] R10: 0000000080000000 R11: 0000000000000001 R12: 0000000000000004
[   45.797114][ T9787] R13: 0000000000000040 R14: ffff88802c50814a R15: 0000000000000000
[   45.798024][ T9787] FS:  00007ffb72734540(0000) GS:ffff8880cec64000(0000) knlGS:0000000000000000
[   45.799019][ T9787] CS:  0010 DS: 0000 ES: 0000 CR0: 0000000080050033
[   45.799822][ T9787] CR2: 00007f9bd8225000 CR3: 000000010979a000 CR4: 00000000000006f0
[   45.800747][ T9787] Kernel panic - not syncing: Fatal exception

The hfs_fill_super() calls hfs_mdb_get() method that tries
to construct Extents Tree and Catalog Tree:

HFS_SB(sb)->ext_tree = hfs_btree_open(sb, HFS_EXT_CNID, hfs_ext_keycmp);
if (!HFS_SB(sb)->ext_tree) {
pr_err("unable to open extent tree\n");
goto out;
}
HFS_SB(sb)->cat_tree = hfs_btree_open(sb, HFS_CAT_CNID, hfs_cat_keycmp);
if (!HFS_SB(sb)->cat_tree) {
pr_err("unable to open catalog tree\n");
goto out;
}

However, hfs_btree_open() calls read_mapping_page() that
calls hfs_get_block(). And this method calls hfs_ext_read_extent():

static int hfs_ext_read_extent(struct inode *inode, u16 block)
{
struct hfs_find_data fd;
int res;

if (block >= HFS_I(inode)->cached_start &&
    block < HFS_I(inode)->cached_start + HFS_I(inode)->cached_blocks)
return 0;

res = hfs_find_init(HFS_SB(inode->i_sb)->ext_tree, &fd);
if (!res) {
res = __hfs_ext_cache_extent(&fd, inode, block);
hfs_find_exit(&fd);
}
return res;
}

The problem here that hfs_find_init() is trying to use
HFS_SB(inode->i_sb)->ext_tree that is not initialized yet.
It will be initailized when hfs_btree_open() finishes
the execution.

The patch adds checking of tree pointer in hfs_find_init()
and it reworks the logic of hfs_btree_open() by reading
the b-tree's header directly from the volume. The read_mapping_page()
is exchanged on filemap_grab_folio() that grab the folio from
mapping. Then, sb_bread() extracts the b-tree's header
content and copy it into the folio.

Reported-by: Wenzhi Wang <wenzhi.wang@uwaterloo.ca>
Signed-off-by: Viacheslav Dubeyko <slava@dubeyko.com>
cc: John Paul Adrian Glaubitz <glaubitz@physik.fu-berlin.de>
cc: Yangtao Li <frank.li@vivo.com>
cc: linux-fsdevel@vger.kernel.org
Link: https://lore.kernel.org/r/20250710213657.108285-1-slava@dubeyko.com
Signed-off-by: Viacheslav Dubeyko <slava@dubeyko.com>
fs/hfs/bfind.c
fs/hfs/btree.c
fs/hfs/extent.c
fs/hfs/hfs_fs.h

index ef9498a6e88acd39e81cbfc2d5d0406477da0a1c..34e9804e0f3601f54b27abe387458e16f32a6b4c 100644 (file)
@@ -16,6 +16,9 @@ int hfs_find_init(struct hfs_btree *tree, struct hfs_find_data *fd)
 {
        void *ptr;
 
+       if (!tree || !fd)
+               return -EINVAL;
+
        fd->tree = tree;
        fd->bnode = NULL;
        ptr = kmalloc(tree->max_key_len * 2 + 4, GFP_KERNEL);
index 2fa4b1f8cc7fb098a9406d600ded87922c2ba9ea..e86e1e235658fab0bad1c9de8890f3ea68030970 100644 (file)
@@ -21,8 +21,12 @@ struct hfs_btree *hfs_btree_open(struct super_block *sb, u32 id, btree_keycmp ke
        struct hfs_btree *tree;
        struct hfs_btree_header_rec *head;
        struct address_space *mapping;
-       struct page *page;
+       struct folio *folio;
+       struct buffer_head *bh;
        unsigned int size;
+       u16 dblock;
+       sector_t start_block;
+       loff_t offset;
 
        tree = kzalloc(sizeof(*tree), GFP_KERNEL);
        if (!tree)
@@ -75,12 +79,40 @@ struct hfs_btree *hfs_btree_open(struct super_block *sb, u32 id, btree_keycmp ke
        unlock_new_inode(tree->inode);
 
        mapping = tree->inode->i_mapping;
-       page = read_mapping_page(mapping, 0, NULL);
-       if (IS_ERR(page))
+       folio = filemap_grab_folio(mapping, 0);
+       if (IS_ERR(folio))
                goto free_inode;
 
+       folio_zero_range(folio, 0, folio_size(folio));
+
+       dblock = hfs_ext_find_block(HFS_I(tree->inode)->first_extents, 0);
+       start_block = HFS_SB(sb)->fs_start + (dblock * HFS_SB(sb)->fs_div);
+
+       size = folio_size(folio);
+       offset = 0;
+       while (size > 0) {
+               size_t len;
+
+               bh = sb_bread(sb, start_block);
+               if (!bh) {
+                       pr_err("unable to read tree header\n");
+                       goto put_folio;
+               }
+
+               len = min_t(size_t, folio_size(folio), sb->s_blocksize);
+               memcpy_to_folio(folio, offset, bh->b_data, sb->s_blocksize);
+
+               brelse(bh);
+
+               start_block++;
+               offset += len;
+               size -= len;
+       }
+
+       folio_mark_uptodate(folio);
+
        /* Load the header */
-       head = (struct hfs_btree_header_rec *)(kmap_local_page(page) +
+       head = (struct hfs_btree_header_rec *)(kmap_local_folio(folio, 0) +
                                               sizeof(struct hfs_bnode_desc));
        tree->root = be32_to_cpu(head->root);
        tree->leaf_count = be32_to_cpu(head->leaf_count);
@@ -95,22 +127,22 @@ struct hfs_btree *hfs_btree_open(struct super_block *sb, u32 id, btree_keycmp ke
 
        size = tree->node_size;
        if (!is_power_of_2(size))
-               goto fail_page;
+               goto fail_folio;
        if (!tree->node_count)
-               goto fail_page;
+               goto fail_folio;
        switch (id) {
        case HFS_EXT_CNID:
                if (tree->max_key_len != HFS_MAX_EXT_KEYLEN) {
                        pr_err("invalid extent max_key_len %d\n",
                               tree->max_key_len);
-                       goto fail_page;
+                       goto fail_folio;
                }
                break;
        case HFS_CAT_CNID:
                if (tree->max_key_len != HFS_MAX_CAT_KEYLEN) {
                        pr_err("invalid catalog max_key_len %d\n",
                               tree->max_key_len);
-                       goto fail_page;
+                       goto fail_folio;
                }
                break;
        default:
@@ -121,12 +153,15 @@ struct hfs_btree *hfs_btree_open(struct super_block *sb, u32 id, btree_keycmp ke
        tree->pages_per_bnode = (tree->node_size + PAGE_SIZE - 1) >> PAGE_SHIFT;
 
        kunmap_local(head);
-       put_page(page);
+       folio_unlock(folio);
+       folio_put(folio);
        return tree;
 
-fail_page:
+fail_folio:
        kunmap_local(head);
-       put_page(page);
+put_folio:
+       folio_unlock(folio);
+       folio_put(folio);
 free_inode:
        tree->inode->i_mapping->a_ops = &hfs_aops;
        iput(tree->inode);
index 4a0ce131e233fe6a717f47957cedf736ec58f7d1..580c62981dbd3d994a3c7b1413659b0a79bffe2b 100644 (file)
@@ -71,7 +71,7 @@ int hfs_ext_keycmp(const btree_key *key1, const btree_key *key2)
  *
  * Find a block within an extent record
  */
-static u16 hfs_ext_find_block(struct hfs_extent *ext, u16 off)
+u16 hfs_ext_find_block(struct hfs_extent *ext, u16 off)
 {
        int i;
        u16 count;
index a0c7cb0f79fcc9ac6a188012c7485be936cb68de..732c5c4c7545d6574356727dbc13d9f33413c0ff 100644 (file)
@@ -190,6 +190,7 @@ extern const struct inode_operations hfs_dir_inode_operations;
 
 /* extent.c */
 extern int hfs_ext_keycmp(const btree_key *, const btree_key *);
+extern u16 hfs_ext_find_block(struct hfs_extent *ext, u16 off);
 extern int hfs_free_fork(struct super_block *, struct hfs_cat_file *, int);
 extern int hfs_ext_write_extent(struct inode *);
 extern int hfs_extend_file(struct inode *);