On Mon, May 16, 2022 at 8:46 AM John Allen <[email protected]> wrote:
>
> For some sev ioctl interfaces, input may be passed that is less than or
> equal to SEV_FW_BLOB_MAX_SIZE, but larger than the data that PSP
> firmware returns. In this case, kmalloc will allocate memory that is the
> size of the input rather than the size of the data. Since PSP firmware
> doesn't fully overwrite the buffer, the sev ioctl interfaces with the
> issue may return uninitialized slab memory.
>
> Currently, all of the ioctl interfaces in the ccp driver are safe, but
> to prevent future problems, change all ioctl interfaces that allocate
> memory with kmalloc to use kzalloc.
>
> Fixes: e799035609e15 ("crypto: ccp: Implement SEV_PEK_CSR ioctl command")
> Fixes: 76a2b524a4b1d ("crypto: ccp: Implement SEV_PDH_CERT_EXPORT ioctl command")
> Fixes: d6112ea0cb344 ("crypto: ccp - introduce SEV_GET_ID2 command")
> Cc: [email protected]
> Reported-by: Andy Nguyen <[email protected]>
> Suggested-by: David Rientjes <[email protected]>
> Suggested-by: Peter Gonda <[email protected]>
> Signed-off-by: John Allen <[email protected]>
> ---
> v2:
> - Add fixes tags and CC [email protected]
> ---
static int sev_ioctl_do_platform_status(struct sev_issue_cmd *argp)
{
struct sev_user_data_status data;
int ret;
ret = __sev_do_cmd_locked(SEV_CMD_PLATFORM_STATUS, &data, &argp->error);
if (ret)
return ret;
if (copy_to_user((void __user *)argp->data, &data, sizeof(data)))
ret = -EFAULT;
return ret;
}
Would it be safer to memset @data here to all zeros too?
[AMD Official Use Only - General]
Hello Peter,
-----Original Message-----
From: Peter Gonda <[email protected]>
Sent: Monday, May 16, 2022 10:53 AM
To: Allen, John <[email protected]>
Cc: Herbert Xu <[email protected]>; Linux Crypto Mailing List <[email protected]>; Sean Christopherson <[email protected]>; Lendacky, Thomas <[email protected]>; Kalra, Ashish <[email protected]>; LKML <[email protected]>; Andy Nguyen <[email protected]>; David Rientjes <[email protected]>; [email protected]
Subject: Re: [PATCH v2] crypto: ccp - Use kzalloc for sev ioctl interfaces to prevent kernel memory leak
On Mon, May 16, 2022 at 8:46 AM John Allen <[email protected]> wrote:
>
> For some sev ioctl interfaces, input may be passed that is less than
> or equal to SEV_FW_BLOB_MAX_SIZE, but larger than the data that PSP
> firmware returns. In this case, kmalloc will allocate memory that is
> the size of the input rather than the size of the data. Since PSP
> firmware doesn't fully overwrite the buffer, the sev ioctl interfaces
> with the issue may return uninitialized slab memory.
>
> Currently, all of the ioctl interfaces in the ccp driver are safe, but
> to prevent future problems, change all ioctl interfaces that allocate
> memory with kmalloc to use kzalloc.
>
> Fixes: e799035609e15 ("crypto: ccp: Implement SEV_PEK_CSR ioctl
> command")
> Fixes: 76a2b524a4b1d ("crypto: ccp: Implement SEV_PDH_CERT_EXPORT
> ioctl command")
> Fixes: d6112ea0cb344 ("crypto: ccp - introduce SEV_GET_ID2 command")
> Cc: [email protected]
> Reported-by: Andy Nguyen <[email protected]>
> Suggested-by: David Rientjes <[email protected]>
> Suggested-by: Peter Gonda <[email protected]>
> Signed-off-by: John Allen <[email protected]>
> ---
> v2:
> - Add fixes tags and CC [email protected]
> ---
>static int sev_ioctl_do_platform_status(struct sev_issue_cmd *argp) { struct sev_user_data_status data; int ret;
>ret = __sev_do_cmd_locked(SEV_CMD_PLATFORM_STATUS, &data, &argp->error); if (ret) return ret;
>if (copy_to_user((void __user *)argp->data, &data, sizeof(data))) ret = -EFAULT;
>return ret;
>}
>Would it be safer to memset @data here to all zeros too?
It will be, but this command/function is safe as firmware will fill in the whole buffer here with the PLATFORM STATUS data retuned to the user.
Thanks,
Ashish
On Mon, May 16, 2022 at 9:02 AM Kalra, Ashish <[email protected]> wrote:
>
> [AMD Official Use Only - General]
>
> Hello Peter,
>
> -----Original Message-----
> From: Peter Gonda <[email protected]>
> Sent: Monday, May 16, 2022 10:53 AM
> To: Allen, John <[email protected]>
> Cc: Herbert Xu <[email protected]>; Linux Crypto Mailing List <[email protected]>; Sean Christopherson <[email protected]>; Lendacky, Thomas <[email protected]>; Kalra, Ashish <[email protected]>; LKML <[email protected]>; Andy Nguyen <[email protected]>; David Rientjes <[email protected]>; [email protected]
> Subject: Re: [PATCH v2] crypto: ccp - Use kzalloc for sev ioctl interfaces to prevent kernel memory leak
>
> On Mon, May 16, 2022 at 8:46 AM John Allen <[email protected]> wrote:
> >
> > For some sev ioctl interfaces, input may be passed that is less than
> > or equal to SEV_FW_BLOB_MAX_SIZE, but larger than the data that PSP
> > firmware returns. In this case, kmalloc will allocate memory that is
> > the size of the input rather than the size of the data. Since PSP
> > firmware doesn't fully overwrite the buffer, the sev ioctl interfaces
> > with the issue may return uninitialized slab memory.
> >
> > Currently, all of the ioctl interfaces in the ccp driver are safe, but
> > to prevent future problems, change all ioctl interfaces that allocate
> > memory with kmalloc to use kzalloc.
> >
> > Fixes: e799035609e15 ("crypto: ccp: Implement SEV_PEK_CSR ioctl
> > command")
> > Fixes: 76a2b524a4b1d ("crypto: ccp: Implement SEV_PDH_CERT_EXPORT
> > ioctl command")
> > Fixes: d6112ea0cb344 ("crypto: ccp - introduce SEV_GET_ID2 command")
> > Cc: [email protected]
> > Reported-by: Andy Nguyen <[email protected]>
> > Suggested-by: David Rientjes <[email protected]>
> > Suggested-by: Peter Gonda <[email protected]>
> > Signed-off-by: John Allen <[email protected]>
> > ---
> > v2:
> > - Add fixes tags and CC [email protected]
> > ---
>
>
> >static int sev_ioctl_do_platform_status(struct sev_issue_cmd *argp) { struct sev_user_data_status data; int ret;
>
> >ret = __sev_do_cmd_locked(SEV_CMD_PLATFORM_STATUS, &data, &argp->error); if (ret) return ret;
>
> >if (copy_to_user((void __user *)argp->data, &data, sizeof(data))) ret = -EFAULT;
>
> >return ret;
> >}
>
> >Would it be safer to memset @data here to all zeros too?
>
> It will be, but this command/function is safe as firmware will fill in the whole buffer here with the PLATFORM STATUS data retuned to the user.
That does seem safe for now but I thought we decided it would be
prudent to not trust the PSPs implementation here and clear all the
buffers that eventually get sent to userspace?
>
> Thanks,
> Ashish
[AMD Official Use Only - General]
Hello Peter,
-----Original Message-----
From: Peter Gonda <[email protected]>
Sent: Monday, May 16, 2022 12:13 PM
To: Kalra, Ashish <[email protected]>
Cc: Allen, John <[email protected]>; Herbert Xu <[email protected]>; Linux Crypto Mailing List <[email protected]>; Sean Christopherson <[email protected]>; Lendacky, Thomas <[email protected]>; LKML <[email protected]>; Andy Nguyen <[email protected]>; David Rientjes <[email protected]>; [email protected]
Subject: Re: [PATCH v2] crypto: ccp - Use kzalloc for sev ioctl interfaces to prevent kernel memory leak
On Mon, May 16, 2022 at 9:02 AM Kalra, Ashish <[email protected]> wrote:
>
> [AMD Official Use Only - General]
>
> Hello Peter,
>
> -----Original Message-----
> From: Peter Gonda <[email protected]>
> Sent: Monday, May 16, 2022 10:53 AM
> To: Allen, John <[email protected]>
> Cc: Herbert Xu <[email protected]>; Linux Crypto Mailing
> List <[email protected]>; Sean Christopherson
> <[email protected]>; Lendacky, Thomas <[email protected]>;
> Kalra, Ashish <[email protected]>; LKML
> <[email protected]>; Andy Nguyen <[email protected]>;
> David Rientjes <[email protected]>; [email protected]
> Subject: Re: [PATCH v2] crypto: ccp - Use kzalloc for sev ioctl
> interfaces to prevent kernel memory leak
>
> On Mon, May 16, 2022 at 8:46 AM John Allen <[email protected]> wrote:
> >
> > For some sev ioctl interfaces, input may be passed that is less than
> > or equal to SEV_FW_BLOB_MAX_SIZE, but larger than the data that PSP
> > firmware returns. In this case, kmalloc will allocate memory that is
> > the size of the input rather than the size of the data. Since PSP
> > firmware doesn't fully overwrite the buffer, the sev ioctl
> > interfaces with the issue may return uninitialized slab memory.
> >
> > Currently, all of the ioctl interfaces in the ccp driver are safe,
> > but to prevent future problems, change all ioctl interfaces that
> > allocate memory with kmalloc to use kzalloc.
> >
> > Fixes: e799035609e15 ("crypto: ccp: Implement SEV_PEK_CSR ioctl
> > command")
> > Fixes: 76a2b524a4b1d ("crypto: ccp: Implement SEV_PDH_CERT_EXPORT
> > ioctl command")
> > Fixes: d6112ea0cb344 ("crypto: ccp - introduce SEV_GET_ID2 command")
> > Cc: [email protected]
> > Reported-by: Andy Nguyen <[email protected]>
> > Suggested-by: David Rientjes <[email protected]>
> > Suggested-by: Peter Gonda <[email protected]>
> > Signed-off-by: John Allen <[email protected]>
> > ---
> > v2:
> > - Add fixes tags and CC [email protected]
> > ---
>
>
> >static int sev_ioctl_do_platform_status(struct sev_issue_cmd *argp) {
> >struct sev_user_data_status data; int ret;
>
> >ret = __sev_do_cmd_locked(SEV_CMD_PLATFORM_STATUS, &data,
> >&argp->error); if (ret) return ret;
>
> >if (copy_to_user((void __user *)argp->data, &data, sizeof(data))) ret
> >= -EFAULT;
>
> >return ret;
> >}
>
> >Would it be safer to memset @data here to all zeros too?
>
> It will be, but this command/function is safe as firmware will fill in the whole buffer here with the PLATFORM STATUS data retuned to the user.
> That does seem safe for now but I thought we decided it would be prudent to not trust the PSPs implementation here and clear all the buffers that eventually get sent to userspace?
Yes, but the issue is when the user programs a buffer size larger the one filled in by the firmware. In this case firmware is always going to fill up the whole buffer with
PLATFORM_STATUS data, so it will be always be safe. The issue is mainly with the kernel side doing a copy_to_user() based on user programmed length instead of the firmware
returned buffer length.
Thanks,
Ashish
On Mon, May 16, 2022, Kalra, Ashish wrote:
> > >Would it be safer to memset @data here to all zeros too?
> >
> > It will be, but this command/function is safe as firmware will fill in the
> > whole buffer here with the PLATFORM STATUS data retuned to the user.
>
> > That does seem safe for now but I thought we decided it would be prudent to
> > not trust the PSPs implementation here and clear all the buffers that
> > eventually get sent to userspace?
>
> Yes, but the issue is when the user programs a buffer size larger the one
> filled in by the firmware. In this case firmware is always going to fill up
> the whole buffer with PLATFORM_STATUS data, so it will be always be safe. The
> issue is mainly with the kernel side doing a copy_to_user() based on user
> programmed length instead of the firmware returned buffer length.
Peter's point is that it costs the kernel very little to be paranoid and not make
assumptions about whether or not the PSP will fill an entire struct as expected.
I agree it feels a bit silly since all fields are output, but on the other hand the
PSP spec just says:
The following data structure is written to memory at STATUS_PADDR
and the data structure has several reserved fields. I don't love assuming that the
PSP will always write zeros for the reserved fields and not do something like:
if (rmp_initialized)
data[3] |= IS_RMP_INIT;
else
data[3] &= ~IS_RMP_INIT;
Given that zeroing @data in the kernel is easy and this is not a hot patch, I
prefer the paranoid approach unless the PSP spec is much more explicit in saying
that it writes all bits and bytes on success.