hw/core/loader: Fix possible crash in rom_copy()
PATCH 1/1 hw/core/loader: Fix possible crash in rom_copy()
Both, "rom->addr" and "addr" are derived from the binary image that can be loaded with the "-kernel" paramer. The code in rom_copy() then calculates:
d = dest + (rom->addr - addr);
and uses "d" as destination in a memcpy() some lines later. Now with bad kernel images, it is possible that rom->addr is smaller than addr, thus "rom->addr - addr" gets negative and the memcpy() then tries to copy contents from the image to a bad memory location. In the best case, this just crashes QEMU, in the worst case, this could maybe be used to inject code from the kernel image into the QEMU binary, so we better fix it with an additional sanity check here.
--- hw/core/loader.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-)
hw/core/loader.c +1 -1
@@ -1281,7 +1281,7 @@ int rom_copy(uint8_t *dest, hwaddr addr, size_t size)if (rom->addr + rom->romsize < addr) {continue ; } -if (rom->addr > end) { +if (rom->addr > end || rom->addr < addr) {break ; }
Signature · 1 line
2.18.1
Quoted 14 lines · Thomas Huth
Both, "
rom->addr" and "addr" are derived from the binary image that can be loaded with the "-kernel" paramer. The code inrom_copy()then calculates:d = dest + (rom->addr - addr);and uses "d" as destination in a
memcpy()some lines later. Now with bad kernel images, it is possible thatrom->addris smaller than addr, thus "rom->addr- addr" gets negative and thememcpy()then tries to copy contents from the image to a bad memory location. In the best case, this just crashes QEMU, in the worst case, this could maybe be used to inject code from the kernel image into the QEMU binary, so we better fix it with an additional sanity check here.
Quoted 4 lines
Reported-by Guangming LiuSigned-off-by Thomas Huth <[email protected]>
Quoted 19 lines
--- hw/core/loader.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-)hw/core/loader.c +1
-1@@ -1281,7 +1281,7 @@ int rom_copy(uint8_t *dest, hwaddr addr, size_t size)if (rom->addr + rom->romsize < addr) {continue ; } -if (rom->addr > end) { +if (rom->addr > end || rom->addr < addr) {break ; }Signature · 1 line
2.18.1
Paolo Bonzini
AppliedQuoted 14 lines · Michael S. Tsirkin
Quoted 14 lines
Both, "
rom->addr" and "addr" are derived from the binary image that can be loaded with the "-kernel" paramer. The code inrom_copy()then calculates:d = dest + (rom->addr - addr);and uses "d" as destination in a
memcpy()some lines later. Now with bad kernel images, it is possible thatrom->addris smaller than addr, thus "rom->addr- addr" gets negative and thememcpy()then tries to copy contents from the image to a bad memory location. In the best case, this just crashes QEMU, in the worst case, this could maybe be used to inject code from the kernel image into the QEMU binary, so we better fix it with an additional sanity check here.
Quoted 6 lines
Quoted 4 lines
Reported-by Guangming LiuSigned-off-by Thomas Huth <[email protected]>Reviewed-by Michael S. Tsirkin <[email protected]>
Queued, thanks.
Paolo
Hi Thomas,
Quoted 14 lines · Thomas Huth
Both, "
rom->addr" and "addr" are derived from the binary image that can be loaded with the "-kernel" paramer. The code inrom_copy()then calculates:d = dest + (rom->addr - addr);and uses "d" as destination in a
memcpy()some lines later. Now with bad kernel images, it is possible thatrom->addris smaller than addr, thus "rom->addr- addr" gets negative and thememcpy()then tries to copy contents from the image to a bad memory location. In the best case, this just crashes QEMU, in the worst case, this could maybe be used to inject code from the kernel image into the QEMU binary, so we better fix it with an additional sanity check here.
Quoted 3 lines
"This page does not exist, or you may not have permission to see it."
This seems security related. Shouldn't we open a CVE for this?
https://wiki.qemu.org/SecurityProcess#CVE_allocation
Let's say I have write access to a LAN TFTP server used by some PXE bootloader where I can store my crafted nasty kernel, then I get this score:
CVSS Base Score: 9.6 CVSS Temporal Score: 8.6
Which seems quite high.
hw/core/loader.c int rom_copy(uint8_t *dest, hwaddr addr, size_t size)
Quoted 10 lines
Signed-off-by Thomas Huth <[email protected]>--- hw/core/loader.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-)hw/core/loader.c +0
-0@@ -1281,7 +1281,7 @@ int rom_copy(uint8_t *dest, hwaddr addr, size_t size)
$ git show 235f86ef014 Date: Thu Nov 12 21:53:11 2009 +0100
This function is old and poorly documented.
hw/core/loader.c int rom_copy(uint8_t *dest, hwaddr addr, size_t size) · line 1284
Quoted 5 lines
hw/core/loader.c +1
-1if (rom->addr + rom->romsize < addr) {continue ; } -if (rom->addr > end) { +if (rom->addr > end || rom->addr < addr) {
Quoted 4 lines
hw/core/loader.c +0
-0break ; }
Quoted 17 lines · Philippe Mathieu-Daudé
Hi Thomas,
Quoted 14 lines · Thomas Huth
Both, "
rom->addr" and "addr" are derived from the binary image that can be loaded with the "-kernel" paramer. The code inrom_copy()then calculates:d = dest + (rom->addr - addr);and uses "d" as destination in a
memcpy()some lines later. Now with bad kernel images, it is possible thatrom->addris smaller than addr, thus "rom->addr- addr" gets negative and thememcpy()then tries to copy contents from the image to a bad memory location. In the best case, this just crashes QEMU, in the worst case, this could maybe be used to inject code from the kernel image into the QEMU binary, so we better fix it with an additional sanity check here.
Quoted 8 lines
Quoted 3 lines
"This page does not exist, or you may not have permission to see it."
This seems security related. Shouldn't we open a CVE for this?
https://wiki.qemu.org/SecurityProcess#CVE_allocation
I wrote to the security team before writing the patch, so I assume a CVE number is already on the way. I'll reply to this thread when it is available.
Thomas
On 25/09/2019 22.51, Philippe Mathieu-Daudé wrote:
[...]
Quoted 9 lines
Let's say I have write access to a LAN TFTP server used by some PXE bootloader where I can store my crafted nasty kernel, then I get this score:
CVSS Base Score: 9.6 CVSS Temporal Score: 8.6Which seems quite high.
I don't think you can trigger this bug this way. If you load your kernel via a PXE server, the ELF parsing will be done by the bootloader, won't it? I think you can only trigger this bug here if you load your kernel via the "-kernel" command line parameter of QEMU (or the generic-loader device), so it's not a real guest escape, as far as I can see.
Thomas
Quoted 10 lines · Thomas Huth
On 25/09/2019 22.51, Philippe Mathieu-Daudé wrote:
[...]Quoted 8 lines
Let's say I have write access to a LAN TFTP server used by some PXE bootloader where I can store my crafted nasty kernel, then I get this score:
CVSS Base Score: 9.6 CVSS Temporal Score: 8.6
Quoted 7 lines
Quoted 1 line
Which seems quite high.
I don't think you can trigger this bug this way. If you load your kernel via a PXE server, the ELF parsing will be done by the bootloader, won't it? I think you can only trigger this bug here if you load your kernel via the "-kernel" command line parameter of QEMU (or the generic-loader device), so it's not a real guest escape, as far as I can see.
Ah indeed you are correct. You have to use the -kernel option to set kernel_filename. This reduce the scores:
CVSS Base Score: 8.4
CVSS Temporal Score: 7.6
Exploitability Subscore: 1.7
You could load a kernel stored on a NFS server, but it is unlikely a production case :)