Re: [PATCH v12 43/46] virt: Add SEV-SNP guest driver

[Peter Gonda <pgonda@google.com>]

On Thu, Aug 25, 2022 at 12:54 PM Tom Lendacky <thomas.lendacky@amd.com> wrote:
>
> On 8/24/22 14:28, Peter Gonda wrote:
> > On Wed, Aug 24, 2022 at 12:01 PM Dionna Amalie Glaze
> > <dionnaglaze@google.com> wrote:
> >>
> >> Apologies for the necropost, but I noticed strange behavior testing my
> >> own Golang-based wrapper around the /dev/sev-guest driver.
> >>
> >>> +
> >>> +static int handle_guest_request(struct snp_guest_dev *snp_dev, u64 exit_code, int msg_ver,
> >>> +                               u8 type, void *req_buf, size_t req_sz, void *resp_buf,
> >>> +                               u32 resp_sz, __u64 *fw_err)
> >>> +{
> >>> +       unsigned long err;
> >>> +       u64 seqno;
> >>> +       int rc;
> >>> +
> >>> +       /* Get message sequence and verify that its a non-zero */
> >>> +       seqno = snp_get_msg_seqno(snp_dev);
> >>> +       if (!seqno)
> >>> +               return -EIO;
> >>> +
> >>> +       memset(snp_dev->response, 0, sizeof(struct snp_guest_msg));
> >>> +
> >>> +       /* Encrypt the userspace provided payload */
> >>> +       rc = enc_payload(snp_dev, seqno, msg_ver, type, req_buf, req_sz);
> >>> +       if (rc)
> >>> +               return rc;
> >>> +
> >>> +       /* Call firmware to process the request */
> >>> +       rc = snp_issue_guest_request(exit_code, &snp_dev->input, &err);
> >>> +       if (fw_err)
> >>> +               *fw_err = err;
> >>> +
> >>> +       if (rc)
> >>> +               return rc;
> >>> +
> >>
> >> The fw_err is written back regardless of rc, so since err is
> >> uninitialized, you can end up with garbage written back. I've worked
> >> around this by only caring about fw_err when the result is -EIO, but
> >> thought that I should bring this up.
> >
> > I also noticed that we use a u64 in snp_guest_request_ioctl.fw_err and
> > u32 in sev_issue_cmd.error when these should be errors from the
> > sev_ret_code enum IIUC.
>
> The reason for the u64 is that the Extended Guest Request can return a
> firmware error or a hypervisor error. To distinguish between the two, a
> firmware error is contained in the lower 32-bits, while a hypervisor error
> is contained in the upper 32-bits (e.g. when not enough contiguous pages
> of memory have been supplied).

Ah, makes sense. I was trying to think of a way to codify the state
described above where we error so early in the IOCTL or call that the
PSP is never called, something like below. I think using UINT32_MAX
still works with how u64 of Extended Guest Request is spec'd. Is this
interesting to clean up the PSP driver and internal calls, and the new
sev-guest driver?

diff --git a/arch/x86/kernel/sev.c b/arch/x86/kernel/sev.c
index 63dc626627a0..d1e605567d5e 100644
--- a/arch/x86/kernel/sev.c
+++ b/arch/x86/kernel/sev.c
@@ -22,6 +22,7 @@
 #include <linux/efi.h>
 #include <linux/platform_device.h>
 #include <linux/io.h>
+#include <linux/psp-sev.h>

 #include <asm/cpu_entry_area.h>
 #include <asm/stacktrace.h>
@@ -2177,6 +2178,8 @@ int snp_issue_guest_request(u64 exit_code,
struct snp_req_data *input, unsigned
        if (!fw_err)
                return -EINVAL;

+       fw_err = SEV_RET_NO_FW_CALL;
+
        /*
         * __sev_get_ghcb() needs to run with IRQs disabled because it is using
         * a per-CPU GHCB.
@@ -2209,6 +2212,8 @@ int snp_issue_guest_request(u64 exit_code,
struct snp_req_data *input, unsigned
                *fw_err = ghcb->save.sw_exit_info_2;

                ret = -EIO;
+       } else {
+               *fw_err = 0;
        }

 e_put:
diff --git a/drivers/crypto/ccp/sev-dev.c b/drivers/crypto/ccp/sev-dev.c
index 9f588c9728f8..e71d6e39aa2b 100644
--- a/drivers/crypto/ccp/sev-dev.c
+++ b/drivers/crypto/ccp/sev-dev.c
@@ -439,7 +439,7 @@ static int __sev_platform_init_locked(int *error)
 {
        struct psp_device *psp = psp_master;
        struct sev_device *sev;
-       int rc, psp_ret = -1;
+       int rc, psp_ret = SEV_RET_NO_FW_CALL;
        int (*init_function)(int *error);

        if (!psp || !psp->sev_data)
diff --git a/include/uapi/linux/psp-sev.h b/include/uapi/linux/psp-sev.h
index 91b4c63d5cbf..b8f2c129d63d 100644
--- a/include/uapi/linux/psp-sev.h
+++ b/include/uapi/linux/psp-sev.h
@@ -36,6 +36,11 @@ enum {
  * SEV Firmware status code
  */
...skipping...

 #include <asm/cpu_entry_area.h>
 #include <asm/stacktrace.h>
@@ -2177,6 +2178,8 @@ int snp_issue_guest_request(u64 exit_code,
struct snp_req_data *input, unsigned
        if (!fw_err)
                return -EINVAL;

+       fw_err = SEV_RET_NO_FW_CALL;
+
        /*
         * __sev_get_ghcb() needs to run with IRQs disabled because it is using
         * a per-CPU GHCB.
@@ -2209,6 +2212,8 @@ int snp_issue_guest_request(u64 exit_code,
struct snp_req_data *input, unsigned
                *fw_err = ghcb->save.sw_exit_info_2;

                ret = -EIO;
+       } else {
+               *fw_err = 0;
        }

 e_put:
diff --git a/drivers/crypto/ccp/sev-dev.c b/drivers/crypto/ccp/sev-dev.c
index 9f588c9728f8..e71d6e39aa2b 100644
--- a/drivers/crypto/ccp/sev-dev.c
+++ b/drivers/crypto/ccp/sev-dev.c
@@ -439,7 +439,7 @@ static int __sev_platform_init_locked(int *error)
 {
        struct psp_device *psp = psp_master;
        struct sev_device *sev;
-       int rc, psp_ret = -1;
+       int rc, psp_ret = SEV_RET_NO_FW_CALL;
        int (*init_function)(int *error);

        if (!psp || !psp->sev_data)
diff --git a/include/uapi/linux/psp-sev.h b/include/uapi/linux/psp-sev.h
index 91b4c63d5cbf..b8f2c129d63d 100644
--- a/include/uapi/linux/psp-sev.h
+++ b/include/uapi/linux/psp-sev.h
@@ -36,6 +36,11 @@ enum {
  * SEV Firmware status code
  */
 typedef enum {
+       /*
+        * This error code is not in the SEV spec but is added to convey that
+        * there was an error that prevented the SEV Firmware from being called.
+        */
+       SEV_RET_NO_FW_CALL = -1,
        SEV_RET_SUCCESS = 0,
        SEV_RET_INVALID_PLATFORM_STATE,
        SEV_RET_INVALID_GUEST_STATE,




> >
> >>
> >> --
> >> -Dionna Glaze, PhD (she/her)