CVE-2023-53143

In the Linux kernel, the following vulnerability has been resolved:<br /> <br /> ext4: fix another off-by-one fsmap error on 1k block filesystems<br /> <br /> Apparently syzbot figured out that issuing this FSMAP call:<br /> <br /> struct fsmap_head cmd = {<br /> .fmh_count = ...;<br /> .fmh_keys = {<br /> { .fmr_device = /* ext4 dev */, .fmr_physical = 0, },<br /> { .fmr_device = /* ext4 dev */, .fmr_physical = 0, },<br /> },<br /> ...<br /> };<br /> ret = ioctl(fd, FS_IOC_GETFSMAP, &amp;cmd);<br /> <br /> Produces this crash if the underlying filesystem is a 1k-block ext4<br /> filesystem:<br /> <br /> kernel BUG at fs/ext4/ext4.h:3331!<br /> invalid opcode: 0000 [#1] PREEMPT SMP<br /> CPU: 3 PID: 3227965 Comm: xfs_io Tainted: G W O 6.2.0-rc8-achx<br /> Hardware name: QEMU Standard PC (Q35 + ICH9, 2009), BIOS 1.15.0-1 04/01/2014<br /> RIP: 0010:ext4_mb_load_buddy_gfp+0x47c/0x570 [ext4]<br /> RSP: 0018:ffffc90007c03998 EFLAGS: 00010246<br /> RAX: ffff888004978000 RBX: ffffc90007c03a20 RCX: ffff888041618000<br /> RDX: 0000000000000000 RSI: 00000000000005a4 RDI: ffffffffa0c99b11<br /> RBP: ffff888012330000 R08: ffffffffa0c2b7d0 R09: 0000000000000400<br /> R10: ffffc90007c03950 R11: 0000000000000000 R12: 0000000000000001<br /> R13: 00000000ffffffff R14: 0000000000000c40 R15: ffff88802678c398<br /> FS: 00007fdf2020c880(0000) GS:ffff88807e100000(0000) knlGS:0000000000000000<br /> CS: 0010 DS: 0000 ES: 0000 CR0: 0000000080050033<br /> CR2: 00007ffd318a5fe8 CR3: 000000007f80f001 CR4: 00000000001706e0<br /> Call Trace:<br /> <br /> ext4_mballoc_query_range+0x4b/0x210 [ext4 dfa189daddffe8fecd3cdfd00564e0f265a8ab80]<br /> ext4_getfsmap_datadev+0x713/0x890 [ext4 dfa189daddffe8fecd3cdfd00564e0f265a8ab80]<br /> ext4_getfsmap+0x2b7/0x330 [ext4 dfa189daddffe8fecd3cdfd00564e0f265a8ab80]<br /> ext4_ioc_getfsmap+0x153/0x2b0 [ext4 dfa189daddffe8fecd3cdfd00564e0f265a8ab80]<br /> __ext4_ioctl+0x2a7/0x17e0 [ext4 dfa189daddffe8fecd3cdfd00564e0f265a8ab80]<br /> __x64_sys_ioctl+0x82/0xa0<br /> do_syscall_64+0x2b/0x80<br /> entry_SYSCALL_64_after_hwframe+0x46/0xb0<br /> RIP: 0033:0x7fdf20558aff<br /> RSP: 002b:00007ffd318a9e30 EFLAGS: 00000246 ORIG_RAX: 0000000000000010<br /> RAX: ffffffffffffffda RBX: 00000000000200c0 RCX: 00007fdf20558aff<br /> RDX: 00007fdf1feb2010 RSI: 00000000c0c0583b RDI: 0000000000000003<br /> RBP: 00005625c0634be0 R08: 00005625c0634c40 R09: 0000000000000001<br /> R10: 0000000000000000 R11: 0000000000000246 R12: 00007fdf1feb2010<br /> R13: 00005625be70d994 R14: 0000000000000800 R15: 0000000000000000<br /> <br /> For GETFSMAP calls, the caller selects a physical block device by<br /> writing its block number into fsmap_head.fmh_keys[01].fmr_device.<br /> To query mappings for a subrange of the device, the starting byte of the<br /> range is written to fsmap_head.fmh_keys[0].fmr_physical and the last<br /> byte of the range goes in fsmap_head.fmh_keys[1].fmr_physical.<br /> <br /> IOWs, to query what mappings overlap with bytes 3-14 of /dev/sda, you&amp;#39;d<br /> set the inputs as follows:<br /> <br /> fmh_keys[0] = { .fmr_device = major(8, 0), .fmr_physical = 3},<br /> fmh_keys[1] = { .fmr_device = major(8, 0), .fmr_physical = 14},<br /> <br /> Which would return you whatever is mapped in the 12 bytes starting at<br /> physical offset 3.<br /> <br /> The crash is due to insufficient range validation of keys[1] in<br /> ext4_getfsmap_datadev. On 1k-block filesystems, block 0 is not part of<br /> the filesystem, which means that s_first_data_block is nonzero.<br /> ext4_get_group_no_and_offset subtracts this quantity from the blocknr<br /> argument before cracking it into a group number and a block number<br /> within a group. IOWs, block group 0 spans blocks 1-8192 (1-based)<br /> instead of 0-8191 (0-based) like what happens with larger blocksizes.<br /> <br /> The net result of this encoding is that blocknr s_first_data_block);<br /> <br /> The division then operates on -1:<br /> <br /> offset = do_div(blocknr, EXT4_BLOCKS_PER_GROUP(sb)) &gt;&gt;<br /> EXT4_SB(sb)-&gt;s_cluster_bits;<br /> <br /> Leaving an impossibly large group number (2^32-1) in blocknr.<br /> ext4_getfsmap_check_keys checked that keys[0<br /> ---truncated---