Currently when a user requests for the Quote generation, the Quote
generation handler (tdx_report_new()) only checks whether the VMM
successfully processes the Quote generation request (status !=
GET_QUOTE_IN_FLIGHT) and returns the output to the user without
validating the status of the output data. Since VMM can return error
even after processing the Quote request, returning success just after
successful processing will create confusion to the user. Although for
the failed request, output buffer length will be zero and can also be
used by the user to identify the failure case, it will be more clear to
return error for all failed cases. So validate the Quote output status
and return error code for all failed cases.
Fixes: f4738f56d1dc ("virt: tdx-guest: Add Quote generation support using TSM_REPORTS")
Reported-by: Xiaoyao Li <[email protected]>
Closes: https://lore.kernel.org/linux-coco/[email protected]/T/#u
Signed-off-by: Kuppuswamy Sathyanarayanan <[email protected]>
---
drivers/virt/coco/tdx-guest/tdx-guest.c | 6 ++++++
1 file changed, 6 insertions(+)
diff --git a/drivers/virt/coco/tdx-guest/tdx-guest.c b/drivers/virt/coco/tdx-guest/tdx-guest.c
index 1253bf76b570..61368318fa39 100644
--- a/drivers/virt/coco/tdx-guest/tdx-guest.c
+++ b/drivers/virt/coco/tdx-guest/tdx-guest.c
@@ -228,6 +228,12 @@ static int tdx_report_new(struct tsm_report *report, void *data)
goto done;
}
+ if (quote_buf->status != GET_QUOTE_SUCCESS) {
+ pr_err("GetQuote request failed, ret %llx\n", quote_buf->status);
+ ret = -EIO;
+ goto done;
+ }
+
buf = kvmemdup(quote_buf->data, quote_buf->out_len, GFP_KERNEL);
if (!buf) {
ret = -ENOMEM;
--
2.25.1
On Tue, Jan 09, 2024 at 05:46:04AM +0000, Kuppuswamy Sathyanarayanan wrote:
> Currently when a user requests for the Quote generation, the Quote
> generation handler (tdx_report_new()) only checks whether the VMM
> successfully processes the Quote generation request (status !=
> GET_QUOTE_IN_FLIGHT) and returns the output to the user without
> validating the status of the output data. Since VMM can return error
> even after processing the Quote request, returning success just after
> successful processing will create confusion to the user. Although for
> the failed request, output buffer length will be zero and can also be
> used by the user to identify the failure case, it will be more clear to
> return error for all failed cases. So validate the Quote output status
> and return error code for all failed cases.
Could you split commit message into several paragraphs? It would be easier
to get along.
It can be helpful to follow structure like:
<Background>
<Problem>
<Solution>
--
Kiryl Shutsemau / Kirill A. Shutemov
On 1/9/2024 5:17 AM, Kirill A . Shutemov wrote:
> On Tue, Jan 09, 2024 at 05:46:04AM +0000, Kuppuswamy Sathyanarayanan wrote:
>> Currently when a user requests for the Quote generation, the Quote
>> generation handler (tdx_report_new()) only checks whether the VMM
>> successfully processes the Quote generation request (status !=
>> GET_QUOTE_IN_FLIGHT) and returns the output to the user without
>> validating the status of the output data. Since VMM can return error
>> even after processing the Quote request, returning success just after
>> successful processing will create confusion to the user. Although for
>> the failed request, output buffer length will be zero and can also be
>> used by the user to identify the failure case, it will be more clear to
>> return error for all failed cases. So validate the Quote output status
>> and return error code for all failed cases.
>
> Could you split commit message into several paragraphs? It would be easier
> to get along.
>
> It can be helpful to follow structure like:
>
> <Background>
>
> <Problem>
>
> <Solution>
>
How about the following version?
During the TDX guest attestation process, TSM ConfigFS ABI is used by
the user attestation agent to get the signed VM measurement data (a.k.a
Quote), which can be used by a remote verifier to validate the
trustworthiness of the guest. When a user requests for the Quote data
via the ConfigFS ABI, the TDX Quote generation handler
(tdx_report_new()) forwards the request to VMM (or QE) via a hypercall,
and then shares the output with the user.
Currently, when handling the Quote generation request, tdx_report_new()
handler only checks whether the VMM successfully processed the request
and if it is true it returns success and shares the output to the user
without actually validating the output data. Since the VMM can return
error even after processing the Quote request, always returning success
for the processed requests is incorrect and will create confusion to
the user. Although for the failed request, output buffer length will
be zero and can also be used by the user to identify the failure case,
it will be more clear to return error for all failed cases.
So when handling the Quote generation request, validate the Quote data
output status and return error code for all failed cases.
--
Sathyanarayanan Kuppuswamy
Linux Kernel Developer
On Tue, Jan 09, 2024 at 07:56:56PM -0800, Kuppuswamy Sathyanarayanan wrote:
>
>
> On 1/9/2024 5:17 AM, Kirill A . Shutemov wrote:
> > On Tue, Jan 09, 2024 at 05:46:04AM +0000, Kuppuswamy Sathyanarayanan wrote:
> >> Currently when a user requests for the Quote generation, the Quote
> >> generation handler (tdx_report_new()) only checks whether the VMM
> >> successfully processes the Quote generation request (status !=
> >> GET_QUOTE_IN_FLIGHT) and returns the output to the user without
> >> validating the status of the output data. Since VMM can return error
> >> even after processing the Quote request, returning success just after
> >> successful processing will create confusion to the user. Although for
> >> the failed request, output buffer length will be zero and can also be
> >> used by the user to identify the failure case, it will be more clear to
> >> return error for all failed cases. So validate the Quote output status
> >> and return error code for all failed cases.
> >
> > Could you split commit message into several paragraphs? It would be easier
> > to get along.
> >
> > It can be helpful to follow structure like:
> >
> > <Background>
> >
> > <Problem>
> >
> > <Solution>
> >
>
> How about the following version?
>
> During the TDX guest attestation process, TSM ConfigFS ABI is used by
> the user attestation agent to get the signed VM measurement data (a.k.a
> Quote), which can be used by a remote verifier to validate the
> trustworthiness of the guest. When a user requests for the Quote data
> via the ConfigFS ABI, the TDX Quote generation handler
> (tdx_report_new()) forwards the request to VMM (or QE) via a hypercall,
> and then shares the output with the user.
>
> Currently, when handling the Quote generation request, tdx_report_new()
> handler only checks whether the VMM successfully processed the request
> and if it is true it returns success and shares the output to the user
> without actually validating the output data. Since the VMM can return
> error even after processing the Quote request, always returning success
> for the processed requests is incorrect and will create confusion to
> the user. Although for the failed request, output buffer length will
> be zero and can also be used by the user to identify the failure case,
> it will be more clear to return error for all failed cases.
>
> So when handling the Quote generation request, validate the Quote data
> output status and return error code for all failed cases.
I would drop the start of the sentence upto ',' here, but otherwise looks
good to me.
--
Kiryl Shutsemau / Kirill A. Shutemov