qemu bug where load linux kernel

[guangming liu]

i found a qemu bug ,when the qemu start and parse the kernel file .

This vulnerability can be exploited.

thanks

/****

(gdb) set args -nodefaults -device pc-testdev -device isa-debug-exit,iobase=0xf4,iosize=0x4 -vnc none -serial stdio -device pci-testdev -machine accel=kvm -m 2048 -smp 2 -cpu host -machine kernel_irqchip=split -kernel poc1
(gdb) r
Starting program: /usr/bin/qemu-system-x86_64 -nodefaults -device pc-testdev -device isa-debug-exit,iobase=0xf4,iosize=0x4 -vnc none -serial stdio -device pci-testdev -machine accel=kvm -m 2048 -smp 2 -cpu host -machine kernel_irqchip=split -kernel ./poc/poc1
[Thread debugging using libthread_db enabled]
Using host libthread_db library "/lib/x86_64-linux-gnu/libthread_db.so.1".
[New Thread 0x7fffe9a03700 (LWP 30066)]
[New Thread 0x7fffe9202700 (LWP 30068)]
[New Thread 0x7fffe8a01700 (LWP 30069)]

Thread 1 "qemu-system-x86" received signal SIGSEGV, Segmentation fault.
__memmove_avx_unaligned_erms () at ../sysdeps/x86_64/multiarch/memmove-vec-unaligned-erms.S:249
249 ../sysdeps/x86_64/multiarch/memmove-vec-unaligned-erms.S: No such file or directory.
(gdb) bt
#0 0x00007ffff2390b1f in __memmove_avx_unaligned_erms () at ../sysdeps/x86_64/multiarch/memmove-vec-unaligned-erms.S:249
#1 0x00005555559ebdcf in rom_copy ()
#2 0x00005555558dd1b3 in load_multiboot ()
#3 0x00005555558de1c3 in ()
#4 0x00005555558e19d1 in pc_memory_init ()
#5 0x00005555558e4ee3 in ()
#6 0x00005555559e8500 in machine_run_board_init ()
#7 0x0000555555834959 in main ()
(gdb) c
Continuing.
Couldn't get registers: No such process.
Couldn't get registers: No such process.
(gdb) [Thread 0x7fffe8a01700 (LWP 30069) exited]
[Thread 0x7fffe9202700 (LWP 30068) exited]
[Thread 0x7fffe9a03700 (LWP 30066) exited]

Program terminated with signal SIGSEGV, Segmentation fault.
The program no longer exists.

***/

[guangming liu (guangming1100)]

bug reason and how to fix it
/*
 * Copies memory from registered ROMs to dest. Any memory that is contained in
 * a ROM between addr and addr + size is copied. Note that this can involve
 * multiple ROMs, which need not start at addr and need not end at addr + size.
 */
int rom_copy(uint8_t *dest, hwaddr addr, size_t size)
{
    hwaddr end = addr + size;
    uint8_t *s, *d = dest;
    size_t l = 0;
    Rom *rom;

    QTAILQ_FOREACH(rom, &roms, next) {
        if (rom->fw_file) {
            continue;
        }
        if (rom->mr) {
            continue;
        }
        if (rom->addr + rom->romsize < addr) {
            continue;
        }
        if (rom->addr > end) {
            break;
        }

        d = dest + (rom->addr - addr);
        s = rom->data;
        l = rom->datasize;

        if ((d + l) > (dest + size)) {
            l = dest - d;
        }

        if (l > 0) {
            memcpy(d, s, l); //*****crash here how to fix check the size l.******//
        }

        if (rom->romsize > rom->datasize) {
            /* If datasize is less than romsize, it means that we didn't
             * allocate all the ROM because the trailing data are only zeros.
             */

            d += l;
            l = rom->romsize - rom->datasize;

            if ((d + l) > (dest + size)) {
                /* Rom size doesn't fit in the destination area. Adjust to avoid
                 * overflow.
                 */
                l = dest - d;
            }

            if (l > 0) {
                memset(d, 0x0, l);
            }
        }
    }

    return (d + l) - dest;
}

[Thomas Huth (th-huth)]

I can't reproduce the issue with your "poc" binary here. Which version of QEMU were you exactly using? Can you reproduce it with the latest version from the master branch?

Also there is already a size check some lines earlier:

        if ((d + l) > (dest + size)) {
            l = dest - d;
        }

Isn't that sufficient?

Also please explain how this vulnerability could be exploited? The code patch is not triggered by the guest, is it?

[guangming liu (guangming1100)]

hi ,

        if ((d + l) > (dest + size)) {
            l = dest - d;
        }
the l will be a very big Unsigned number.

the check was bypassed,try the new poc . i reproduce it with the latest
version on ubuntu . (apt install qemu , i got the latest version)

[guangming liu (guangming1100)]

hi Thomas,please try the new poc.
thanks

[Thomas Huth (th-huth)]

Thanks a lot! With the new poc, I was able to reproduce the crash.
I've forwarded the information to the QEMU security team (next time, it would be great if you could do that directly, see https://wiki.qemu.org/SecurityProcess for details), and after some discussion about the severity of the bug, I've now posted a patch to the mailing:

 https://lists.gnu.org/archive/html/qemu-devel/2019-09/msg05960.html

[Thomas Huth (th-huth)]

Fix has been merged:
https://git.qemu.org/?p=qemu.git;a=commitdiff;h=e423455c4f23a1a8