libata: remove WARN() for DMA or PIO command without data
authorEric Biggers <ebiggers@google.com>
Sun, 4 Feb 2018 04:33:27 +0000 (20:33 -0800)
committerTejun Heo <tj@kernel.org>
Mon, 12 Feb 2018 17:20:34 +0000 (09:20 -0800)
syzkaller hit a WARN() in ata_qc_issue() when writing to /dev/sg0.  This
happened because it issued a READ_6 command with no data buffer.

Just remove the WARN(), as it doesn't appear indicate a kernel bug.  The
expected behavior is to fail the command, which the code does.

Here's a reproducer that works in QEMU when /dev/sg0 refers to a disk of
the default type ("82371SB PIIX3 IDE"):

    #include <fcntl.h>
    #include <unistd.h>

    int main()
    {
            char buf[42] = { [36] = 0x8 /* READ_6 */ };

            write(open("/dev/sg0", O_RDWR), buf, sizeof(buf));
    }

Fixes: f92a26365a72 ("libata: change ATA_QCFLAG_DMAMAP semantics")
Reported-by: syzbot+f7b556d1766502a69d85071d2ff08bd87be53d0f@syzkaller.appspotmail.com
Cc: <stable@vger.kernel.org> # v2.6.25+
Signed-off-by: Eric Biggers <ebiggers@google.com>
Signed-off-by: Tejun Heo <tj@kernel.org>
drivers/ata/libata-core.c

index 3c09122bf03825b6489a16b8f116bd05485aa1b4..61b09968d0326b267a93ca54500a1c7c3d6792f7 100644 (file)
@@ -5401,8 +5401,7 @@ void ata_qc_issue(struct ata_queued_cmd *qc)
         * We guarantee to LLDs that they will have at least one
         * non-zero sg if the command is a data command.
         */
-       if (WARN_ON_ONCE(ata_is_data(prot) &&
-                        (!qc->sg || !qc->n_elem || !qc->nbytes)))
+       if (ata_is_data(prot) && (!qc->sg || !qc->n_elem || !qc->nbytes))
                goto sys_err;
 
        if (ata_is_dma(prot) || (ata_is_pio(prot) &&