| Message ID | 20230105034327.1439-6-mario.limonciello@amd.com (mailing list archive) |
|---|---|
| State | New, archived |
| Headers | show |
| Series | Recover from failure to probe GPU | expand |
On 1/5/2023 9:12 AM, Mario Limonciello wrote: > All microcode runs a basic validation after it's been loaded. Each > IP block as part of init will run both. > > Introduce a wrapper for request_firmware and amdgpu_ucode_validate. > This wrapper will also remap any error codes from request_firmware > to -ENODEV. This is so that early_init will fail if firmware couldn't > be loaded instead of the IP block being disabled. > > Signed-off-by: Mario Limonciello <mario.limonciello@amd.com> > --- > v5->v6: > * Fix argument to be ** not * > --- > drivers/gpu/drm/amd/amdgpu/amdgpu_ucode.c | 36 +++++++++++++++++++++++ > drivers/gpu/drm/amd/amdgpu/amdgpu_ucode.h | 3 ++ > 2 files changed, 39 insertions(+) > > diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_ucode.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_ucode.c > index eafcddce58d3..8ebfec12da87 100644 > --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_ucode.c > +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_ucode.c > @@ -1312,3 +1312,39 @@ void amdgpu_ucode_ip_version_decode(struct amdgpu_device *adev, int block_type, > > snprintf(ucode_prefix, len, "%s_%d_%d_%d", ip_name, maj, min, rev); > } > + > +/* > + * amdgpu_ucode_request - Fetch and validate amdgpu microcode > + * > + * @adev: amdgpu device > + * @fw: pointer to load firmware to > + * @fw_name: firmware to load > + * > + * This is a helper that will use request_firmware and amdgpu_ucode_validate > + * to load and run basic validation on firmware. If the load fails, remap > + * the error code to -ENODEV, so that early_init functions will fail to load. > + */ > +int amdgpu_ucode_request(struct amdgpu_device *adev, const struct firmware **fw, > + const char *fw_name) > +{ > + int err = request_firmware(fw, fw_name, adev->dev); > + > + if (err) > + return -ENODEV; > + err = amdgpu_ucode_validate(*fw); > + if (err) > + dev_dbg(adev->dev, "\"%s\" failed to validate\n", fw_name); > + Missed this earlier. If validate fails, shouldn't this undo the request operation by calling release? Thanks, Lijo > + return err; > +} > + > +/* > + * amdgpu_ucode_release - Release firmware microcode > + * > + * @fw: pointer to firmware to release > + */ > +void amdgpu_ucode_release(const struct firmware **fw) > +{ > + release_firmware(*fw); > + *fw = NULL; > +} > diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_ucode.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_ucode.h > index 552e06929229..848579d4988b 100644 > --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_ucode.h > +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_ucode.h > @@ -544,6 +544,9 @@ void amdgpu_ucode_print_sdma_hdr(const struct common_firmware_header *hdr); > void amdgpu_ucode_print_psp_hdr(const struct common_firmware_header *hdr); > void amdgpu_ucode_print_gpu_info_hdr(const struct common_firmware_header *hdr); > int amdgpu_ucode_validate(const struct firmware *fw); > +int amdgpu_ucode_request(struct amdgpu_device *adev, const struct firmware **fw, > + const char *fw_name); > +void amdgpu_ucode_release(const struct firmware **fw); > bool amdgpu_ucode_hdr_version(union amdgpu_firmware_header *hdr, > uint16_t hdr_major, uint16_t hdr_minor); >
On 1/4/23 23:07, Lazar, Lijo wrote: > > > On 1/5/2023 9:12 AM, Mario Limonciello wrote: >> All microcode runs a basic validation after it's been loaded. Each >> IP block as part of init will run both. >> >> Introduce a wrapper for request_firmware and amdgpu_ucode_validate. >> This wrapper will also remap any error codes from request_firmware >> to -ENODEV. This is so that early_init will fail if firmware couldn't >> be loaded instead of the IP block being disabled. >> >> Signed-off-by: Mario Limonciello <mario.limonciello@amd.com> >> --- >> v5->v6: >> * Fix argument to be ** not * >> --- >> drivers/gpu/drm/amd/amdgpu/amdgpu_ucode.c | 36 +++++++++++++++++++++++ >> drivers/gpu/drm/amd/amdgpu/amdgpu_ucode.h | 3 ++ >> 2 files changed, 39 insertions(+) >> >> diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_ucode.c >> b/drivers/gpu/drm/amd/amdgpu/amdgpu_ucode.c >> index eafcddce58d3..8ebfec12da87 100644 >> --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_ucode.c >> +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_ucode.c >> @@ -1312,3 +1312,39 @@ void amdgpu_ucode_ip_version_decode(struct >> amdgpu_device *adev, int block_type, >> snprintf(ucode_prefix, len, "%s_%d_%d_%d", ip_name, maj, min, rev); >> } >> + >> +/* >> + * amdgpu_ucode_request - Fetch and validate amdgpu microcode >> + * >> + * @adev: amdgpu device >> + * @fw: pointer to load firmware to >> + * @fw_name: firmware to load >> + * >> + * This is a helper that will use request_firmware and >> amdgpu_ucode_validate >> + * to load and run basic validation on firmware. If the load fails, >> remap >> + * the error code to -ENODEV, so that early_init functions will fail >> to load. >> + */ >> +int amdgpu_ucode_request(struct amdgpu_device *adev, const struct >> firmware **fw, >> + const char *fw_name) >> +{ >> + int err = request_firmware(fw, fw_name, adev->dev); >> + >> + if (err) >> + return -ENODEV; >> + err = amdgpu_ucode_validate(*fw); >> + if (err) >> + dev_dbg(adev->dev, "\"%s\" failed to validate\n", fw_name); >> + > > Missed this earlier. If validate fails, shouldn't this undo the request > operation by calling release? Actually that was original design, but there is one place in the codebase that expects that ucode validation can fail, and so leave the evaluate of error code and cleanup outside of helper. > > Thanks, > Lijo > >> + return err; >> +} >> + >> +/* >> + * amdgpu_ucode_release - Release firmware microcode >> + * >> + * @fw: pointer to firmware to release >> + */ >> +void amdgpu_ucode_release(const struct firmware **fw) >> +{ >> + release_firmware(*fw); >> + *fw = NULL; >> +} >> diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_ucode.h >> b/drivers/gpu/drm/amd/amdgpu/amdgpu_ucode.h >> index 552e06929229..848579d4988b 100644 >> --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_ucode.h >> +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_ucode.h >> @@ -544,6 +544,9 @@ void amdgpu_ucode_print_sdma_hdr(const struct >> common_firmware_header *hdr); >> void amdgpu_ucode_print_psp_hdr(const struct common_firmware_header >> *hdr); >> void amdgpu_ucode_print_gpu_info_hdr(const struct >> common_firmware_header *hdr); >> int amdgpu_ucode_validate(const struct firmware *fw); >> +int amdgpu_ucode_request(struct amdgpu_device *adev, const struct >> firmware **fw, >> + const char *fw_name); >> +void amdgpu_ucode_release(const struct firmware **fw); >> bool amdgpu_ucode_hdr_version(union amdgpu_firmware_header *hdr, >> uint16_t hdr_major, uint16_t hdr_minor);
On 1/5/2023 10:53 AM, Mario Limonciello wrote: > On 1/4/23 23:07, Lazar, Lijo wrote: >> >> >> On 1/5/2023 9:12 AM, Mario Limonciello wrote: >>> All microcode runs a basic validation after it's been loaded. Each >>> IP block as part of init will run both. >>> >>> Introduce a wrapper for request_firmware and amdgpu_ucode_validate. >>> This wrapper will also remap any error codes from request_firmware >>> to -ENODEV. This is so that early_init will fail if firmware couldn't >>> be loaded instead of the IP block being disabled. >>> >>> Signed-off-by: Mario Limonciello <mario.limonciello@amd.com> >>> --- >>> v5->v6: >>> * Fix argument to be ** not * >>> --- >>> drivers/gpu/drm/amd/amdgpu/amdgpu_ucode.c | 36 +++++++++++++++++++++++ >>> drivers/gpu/drm/amd/amdgpu/amdgpu_ucode.h | 3 ++ >>> 2 files changed, 39 insertions(+) >>> >>> diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_ucode.c >>> b/drivers/gpu/drm/amd/amdgpu/amdgpu_ucode.c >>> index eafcddce58d3..8ebfec12da87 100644 >>> --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_ucode.c >>> +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_ucode.c >>> @@ -1312,3 +1312,39 @@ void amdgpu_ucode_ip_version_decode(struct >>> amdgpu_device *adev, int block_type, >>> snprintf(ucode_prefix, len, "%s_%d_%d_%d", ip_name, maj, min, >>> rev); >>> } >>> + >>> +/* >>> + * amdgpu_ucode_request - Fetch and validate amdgpu microcode >>> + * >>> + * @adev: amdgpu device >>> + * @fw: pointer to load firmware to >>> + * @fw_name: firmware to load >>> + * >>> + * This is a helper that will use request_firmware and >>> amdgpu_ucode_validate >>> + * to load and run basic validation on firmware. If the load fails, >>> remap >>> + * the error code to -ENODEV, so that early_init functions will fail >>> to load. >>> + */ >>> +int amdgpu_ucode_request(struct amdgpu_device *adev, const struct >>> firmware **fw, >>> + const char *fw_name) >>> +{ >>> + int err = request_firmware(fw, fw_name, adev->dev); >>> + >>> + if (err) >>> + return -ENODEV; >>> + err = amdgpu_ucode_validate(*fw); >>> + if (err) >>> + dev_dbg(adev->dev, "\"%s\" failed to validate\n", fw_name); >>> + >> >> Missed this earlier. If validate fails, shouldn't this undo the >> request operation by calling release? > > Actually that was original design, but there is one place in the > codebase that expects that ucode validation can fail, and so leave the > evaluate of error code and cleanup outside of helper. > I see. Does request_firmware assure that fw pointer be always NULL if it fails? Or should that be done here if request_ fails? In subsequent patches, I see clients calling release without checking what caused the failure. Thanks, Lijo >> >> Thanks, >> Lijo >> >>> + return err; >>> +} >>> + >>> +/* >>> + * amdgpu_ucode_release - Release firmware microcode >>> + * >>> + * @fw: pointer to firmware to release >>> + */ >>> +void amdgpu_ucode_release(const struct firmware **fw) >>> +{ >>> + release_firmware(*fw); >>> + *fw = NULL; >>> +} >>> diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_ucode.h >>> b/drivers/gpu/drm/amd/amdgpu/amdgpu_ucode.h >>> index 552e06929229..848579d4988b 100644 >>> --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_ucode.h >>> +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_ucode.h >>> @@ -544,6 +544,9 @@ void amdgpu_ucode_print_sdma_hdr(const struct >>> common_firmware_header *hdr); >>> void amdgpu_ucode_print_psp_hdr(const struct common_firmware_header >>> *hdr); >>> void amdgpu_ucode_print_gpu_info_hdr(const struct >>> common_firmware_header *hdr); >>> int amdgpu_ucode_validate(const struct firmware *fw); >>> +int amdgpu_ucode_request(struct amdgpu_device *adev, const struct >>> firmware **fw, >>> + const char *fw_name); >>> +void amdgpu_ucode_release(const struct firmware **fw); >>> bool amdgpu_ucode_hdr_version(union amdgpu_firmware_header *hdr, >>> uint16_t hdr_major, uint16_t hdr_minor); >
On 1/4/23 23:29, Lazar, Lijo wrote: > > > On 1/5/2023 10:53 AM, Mario Limonciello wrote: >> On 1/4/23 23:07, Lazar, Lijo wrote: >>> >>> >>> On 1/5/2023 9:12 AM, Mario Limonciello wrote: >>>> All microcode runs a basic validation after it's been loaded. Each >>>> IP block as part of init will run both. >>>> >>>> Introduce a wrapper for request_firmware and amdgpu_ucode_validate. >>>> This wrapper will also remap any error codes from request_firmware >>>> to -ENODEV. This is so that early_init will fail if firmware couldn't >>>> be loaded instead of the IP block being disabled. >>>> >>>> Signed-off-by: Mario Limonciello <mario.limonciello@amd.com> >>>> --- >>>> v5->v6: >>>> * Fix argument to be ** not * >>>> --- >>>> drivers/gpu/drm/amd/amdgpu/amdgpu_ucode.c | 36 >>>> +++++++++++++++++++++++ >>>> drivers/gpu/drm/amd/amdgpu/amdgpu_ucode.h | 3 ++ >>>> 2 files changed, 39 insertions(+) >>>> >>>> diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_ucode.c >>>> b/drivers/gpu/drm/amd/amdgpu/amdgpu_ucode.c >>>> index eafcddce58d3..8ebfec12da87 100644 >>>> --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_ucode.c >>>> +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_ucode.c >>>> @@ -1312,3 +1312,39 @@ void amdgpu_ucode_ip_version_decode(struct >>>> amdgpu_device *adev, int block_type, >>>> snprintf(ucode_prefix, len, "%s_%d_%d_%d", ip_name, maj, min, >>>> rev); >>>> } >>>> + >>>> +/* >>>> + * amdgpu_ucode_request - Fetch and validate amdgpu microcode >>>> + * >>>> + * @adev: amdgpu device >>>> + * @fw: pointer to load firmware to >>>> + * @fw_name: firmware to load >>>> + * >>>> + * This is a helper that will use request_firmware and >>>> amdgpu_ucode_validate >>>> + * to load and run basic validation on firmware. If the load fails, >>>> remap >>>> + * the error code to -ENODEV, so that early_init functions will >>>> fail to load. >>>> + */ >>>> +int amdgpu_ucode_request(struct amdgpu_device *adev, const struct >>>> firmware **fw, >>>> + const char *fw_name) >>>> +{ >>>> + int err = request_firmware(fw, fw_name, adev->dev); >>>> + >>>> + if (err) >>>> + return -ENODEV; >>>> + err = amdgpu_ucode_validate(*fw); >>>> + if (err) >>>> + dev_dbg(adev->dev, "\"%s\" failed to validate\n", fw_name); >>>> + >>> >>> Missed this earlier. If validate fails, shouldn't this undo the >>> request operation by calling release? >> >> Actually that was original design, but there is one place in the >> codebase that expects that ucode validation can fail, and so leave the >> evaluate of error code and cleanup outside of helper. >> > > I see. Does request_firmware assure that fw pointer be always NULL if it > fails? Or should that be done here if request_ fails? In subsequent > patches, I see clients calling release without checking what caused the > failure. Yes, according to docs (https://docs.kernel.org/next/driver-api/firmware/request_firmware.html#request-firmware-api-expected-driver-use) > > Thanks, > Lijo > >>> >>> Thanks, >>> Lijo >>> >>>> + return err; >>>> +} >>>> + >>>> +/* >>>> + * amdgpu_ucode_release - Release firmware microcode >>>> + * >>>> + * @fw: pointer to firmware to release >>>> + */ >>>> +void amdgpu_ucode_release(const struct firmware **fw) >>>> +{ >>>> + release_firmware(*fw); >>>> + *fw = NULL; >>>> +} >>>> diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_ucode.h >>>> b/drivers/gpu/drm/amd/amdgpu/amdgpu_ucode.h >>>> index 552e06929229..848579d4988b 100644 >>>> --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_ucode.h >>>> +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_ucode.h >>>> @@ -544,6 +544,9 @@ void amdgpu_ucode_print_sdma_hdr(const struct >>>> common_firmware_header *hdr); >>>> void amdgpu_ucode_print_psp_hdr(const struct >>>> common_firmware_header *hdr); >>>> void amdgpu_ucode_print_gpu_info_hdr(const struct >>>> common_firmware_header *hdr); >>>> int amdgpu_ucode_validate(const struct firmware *fw); >>>> +int amdgpu_ucode_request(struct amdgpu_device *adev, const struct >>>> firmware **fw, >>>> + const char *fw_name); >>>> +void amdgpu_ucode_release(const struct firmware **fw); >>>> bool amdgpu_ucode_hdr_version(union amdgpu_firmware_header *hdr, >>>> uint16_t hdr_major, uint16_t hdr_minor); >>
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_ucode.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_ucode.c index eafcddce58d3..8ebfec12da87 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_ucode.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_ucode.c @@ -1312,3 +1312,39 @@ void amdgpu_ucode_ip_version_decode(struct amdgpu_device *adev, int block_type, snprintf(ucode_prefix, len, "%s_%d_%d_%d", ip_name, maj, min, rev); } + +/* + * amdgpu_ucode_request - Fetch and validate amdgpu microcode + * + * @adev: amdgpu device + * @fw: pointer to load firmware to + * @fw_name: firmware to load + * + * This is a helper that will use request_firmware and amdgpu_ucode_validate + * to load and run basic validation on firmware. If the load fails, remap + * the error code to -ENODEV, so that early_init functions will fail to load. + */ +int amdgpu_ucode_request(struct amdgpu_device *adev, const struct firmware **fw, + const char *fw_name) +{ + int err = request_firmware(fw, fw_name, adev->dev); + + if (err) + return -ENODEV; + err = amdgpu_ucode_validate(*fw); + if (err) + dev_dbg(adev->dev, "\"%s\" failed to validate\n", fw_name); + + return err; +} + +/* + * amdgpu_ucode_release - Release firmware microcode + * + * @fw: pointer to firmware to release + */ +void amdgpu_ucode_release(const struct firmware **fw) +{ + release_firmware(*fw); + *fw = NULL; +} diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_ucode.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_ucode.h index 552e06929229..848579d4988b 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_ucode.h +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_ucode.h @@ -544,6 +544,9 @@ void amdgpu_ucode_print_sdma_hdr(const struct common_firmware_header *hdr); void amdgpu_ucode_print_psp_hdr(const struct common_firmware_header *hdr); void amdgpu_ucode_print_gpu_info_hdr(const struct common_firmware_header *hdr); int amdgpu_ucode_validate(const struct firmware *fw); +int amdgpu_ucode_request(struct amdgpu_device *adev, const struct firmware **fw, + const char *fw_name); +void amdgpu_ucode_release(const struct firmware **fw); bool amdgpu_ucode_hdr_version(union amdgpu_firmware_header *hdr, uint16_t hdr_major, uint16_t hdr_minor);
All microcode runs a basic validation after it's been loaded. Each IP block as part of init will run both. Introduce a wrapper for request_firmware and amdgpu_ucode_validate. This wrapper will also remap any error codes from request_firmware to -ENODEV. This is so that early_init will fail if firmware couldn't be loaded instead of the IP block being disabled. Signed-off-by: Mario Limonciello <mario.limonciello@amd.com> --- v5->v6: * Fix argument to be ** not * --- drivers/gpu/drm/amd/amdgpu/amdgpu_ucode.c | 36 +++++++++++++++++++++++ drivers/gpu/drm/amd/amdgpu/amdgpu_ucode.h | 3 ++ 2 files changed, 39 insertions(+)