[AMD Official Use Only]
With the typo fixed, the patch is:
Reviewed-by: Tao Zhou <tao.zh...@amd.com>
> -----Original Message-----
> From: Yang, Stanley <stanley.y...@amd.com>
> Sent: Thursday, January 20, 2022 11:25 AM
> To: Zhou1, Tao <tao.zh...@amd.com>; amd-gfx@lists.freedesktop.org
> Cc: Zhang, Hawking <hawking.zh...@amd.com>; Clements, John
> <john.cleme...@amd.com>; Ziya, Mohammad zafar
> <mohammadzafar.z...@amd.com>
> Subject: 回复: [PATCH Review 1/1] drm/amdgpu: fix convert bad page retiremt
>
> [AMD Official Use Only]
>
>
>
> > -----邮件原件-----
> > 发件人: Zhou1, Tao <tao.zh...@amd.com>
> > 发送时间: Thursday, January 20, 2022 11:09 AM
> > 收件人: Yang, Stanley <stanley.y...@amd.com>; amd-
> > g...@lists.freedesktop.org
> > 抄送: Zhang, Hawking <hawking.zh...@amd.com>; Clements, John
> > <john.cleme...@amd.com>; Ziya, Mohammad zafar
> > <mohammadzafar.z...@amd.com>; Yang, Stanley <stanley.y...@amd.com>
> > 主题: RE: [PATCH Review 1/1] drm/amdgpu: fix convert bad page retiremt
> >
> > [AMD Official Use Only]
> >
> >
> >
> > > -----Original Message-----
> > > From: Stanley.Yang <stanley.y...@amd.com>
> > > Sent: Thursday, January 20, 2022 12:29 AM
> > > To: amd-gfx@lists.freedesktop.org
> > > Cc: Zhang, Hawking <hawking.zh...@amd.com>; Zhou1, Tao
> > > <tao.zh...@amd.com>; Clements, John <john.cleme...@amd.com>;
> > Ziya,
> > > Mohammad zafar <mohammadzafar.z...@amd.com>; Yang, Stanley
> > > <stanley.y...@amd.com>
> > > Subject: [PATCH Review 1/1] drm/amdgpu: fix convert bad page
> > > retiremt
> > >
> > > Pmfw read ecc info registers and store values in eccinfo_table in
> > > the following order
> > >
> > > umc0 ch_inst 0, 1, 2 ... 7
> > > umc1 ch_inst 0, 1, 2 ... 7
> > > ...
> > > umc3 ch_inst 0, 1, 2 ... 7
> > >
> > > Driver should convert eccinfo_table_idx into channel_index according
> > > to channel_idx_tbe.
> > [Tao]: typo, channel_idx_tbe -> channel_idx_tbl
> >
> > The patch is OK for me, do we also need to apply the update to umc_v8_7.c?
> [Yang, Stanley] : This need to confirm the ecc info registers address
> definition in
> pmfw, it should be updated if definition mechanism of those registers
> addresses
> are same.
> >
> > >
> > > Change-Id: Icafe93e458912b729d2e30d655fd68be0e12124d
> > > Signed-off-by: Stanley.Yang <stanley.y...@amd.com>
> > > ---
> > > drivers/gpu/drm/amd/amdgpu/umc_v6_7.c | 26 ++++++++++++++--------
> > ----
> > > 1 file changed, 14 insertions(+), 12 deletions(-)
> > >
> > > diff --git a/drivers/gpu/drm/amd/amdgpu/umc_v6_7.c
> > > b/drivers/gpu/drm/amd/amdgpu/umc_v6_7.c
> > > index 526de1ca9b8d..f5a1ba7db75a 100644
> > > --- a/drivers/gpu/drm/amd/amdgpu/umc_v6_7.c
> > > +++ b/drivers/gpu/drm/amd/amdgpu/umc_v6_7.c
> > > @@ -58,29 +58,33 @@ static inline uint32_t
> > > get_umc_v6_7_channel_index(struct amdgpu_device *adev, }
> > >
> > > static void umc_v6_7_ecc_info_query_correctable_error_count(struct
> > > amdgpu_device *adev,
> > > - uint32_t channel_index,
> > > + uint32_t umc_inst, uint32_t
> > > ch_inst,
> > > unsigned long *error_count)
> {
> > > uint64_t mc_umc_status;
> > > + uint32_t eccinfo_table_idx;
> > > struct amdgpu_ras *ras = amdgpu_ras_get_context(adev);
> > >
> > > + eccinfo_table_idx = umc_inst * adev->umc.channel_inst_num +
> > ch_inst;
> > > /* check for SRAM correctable error
> > > MCUMC_STATUS is a 64 bit register */
> > > - mc_umc_status = ras-
> > >umc_ecc.ecc[channel_index].mca_umc_status;
> > > + mc_umc_status = ras-
> > > >umc_ecc.ecc[eccinfo_table_idx].mca_umc_status;
> > > if (REG_GET_FIELD(mc_umc_status,
> > > MCA_UMC_UMC0_MCUMC_STATUST0, Val) == 1 &&
> > > REG_GET_FIELD(mc_umc_status,
> > > MCA_UMC_UMC0_MCUMC_STATUST0, CECC) == 1)
> > > *error_count += 1;
> > > }
> > >
> > > static void
> > > umc_v6_7_ecc_info_querry_uncorrectable_error_count(struct
> > > amdgpu_device *adev,
> > > - uint32_t channel_index,
> > > + uint32_t umc_inst,
> > > uint32_t ch_inst,
> > > unsigned long
> > *error_count) {
> > > uint64_t mc_umc_status;
> > > + uint32_t eccinfo_table_idx;
> > > struct amdgpu_ras *ras = amdgpu_ras_get_context(adev);
> > >
> > > + eccinfo_table_idx = umc_inst * adev->umc.channel_inst_num +
> > ch_inst;
> > > /* check the MCUMC_STATUS */
> > > - mc_umc_status = ras-
> > >umc_ecc.ecc[channel_index].mca_umc_status;
> > > + mc_umc_status = ras-
> > > >umc_ecc.ecc[eccinfo_table_idx].mca_umc_status;
> > > if ((REG_GET_FIELD(mc_umc_status,
> > > MCA_UMC_UMC0_MCUMC_STATUST0, Val) == 1) &&
> > > (REG_GET_FIELD(mc_umc_status,
> > > MCA_UMC_UMC0_MCUMC_STATUST0, Deferred) == 1 ||
> > > REG_GET_FIELD(mc_umc_status,
> > > MCA_UMC_UMC0_MCUMC_STATUST0, UECC) == 1 || @@ -97,19 +101,15
> > @@ static
> > > void umc_v6_7_ecc_info_query_ras_error_count(struct amdgpu_device
> > > *adev,
> > >
> > > uint32_t umc_inst = 0;
> > > uint32_t ch_inst = 0;
> > > - uint32_t channel_index = 0;
> > >
> > > /*TODO: driver needs to toggle DF Cstate to ensure
> > > * safe access of UMC registers. Will add the protection */
> > > LOOP_UMC_INST_AND_CH(umc_inst, ch_inst) {
> > > - channel_index = get_umc_v6_7_channel_index(adev,
> > > - umc_inst,
> > > - ch_inst);
> > > umc_v6_7_ecc_info_query_correctable_error_count(adev,
> > > - channel_index,
> > > + umc_inst, ch_inst,
> > > &(err_data->ce_count));
> > >
> > umc_v6_7_ecc_info_querry_uncorrectable_error_count(adev,
> > > - channel_index,
> > > + umc_inst, ch_inst,
> > > &(err_data-
> > > >ue_count));
> > > }
> > > }
> > > @@ -122,12 +122,14 @@ static void
> > > umc_v6_7_ecc_info_query_error_address(struct amdgpu_device *adev,
> > > uint64_t mc_umc_status, err_addr, retired_page;
> > > struct eeprom_table_record *err_rec;
> > > uint32_t channel_index;
> > > + uint32_t eccinfo_table_idx;
> > > struct amdgpu_ras *ras = amdgpu_ras_get_context(adev);
> > >
> > > + eccinfo_table_idx = umc_inst * adev->umc.channel_inst_num +
> > ch_inst;
> > > channel_index =
> > > adev->umc.channel_idx_tbl[umc_inst * adev-
> > > >umc.channel_inst_num + ch_inst];
> > >
> > > - mc_umc_status = ras-
> > >umc_ecc.ecc[channel_index].mca_umc_status;
> > > + mc_umc_status = ras-
> > > >umc_ecc.ecc[eccinfo_table_idx].mca_umc_status;
> > >
> > > if (mc_umc_status == 0)
> > > return;
> > > @@ -142,7 +144,7 @@ static void
> > > umc_v6_7_ecc_info_query_error_address(struct amdgpu_device *adev,
> > > (REG_GET_FIELD(mc_umc_status,
> > > MCA_UMC_UMC0_MCUMC_STATUST0, UECC) == 1 ||
> > > REG_GET_FIELD(mc_umc_status,
> > > MCA_UMC_UMC0_MCUMC_STATUST0, CECC) == 1)) {
> > >
> > > - err_addr = ras-
> > >umc_ecc.ecc[channel_index].mca_umc_addr;
> > > + err_addr = ras-
> > > >umc_ecc.ecc[eccinfo_table_idx].mca_umc_addr;
> > > err_addr = REG_GET_FIELD(err_addr,
> MCA_UMC_UMC0_MCUMC_ADDRT0,
> > > ErrorAddr);
> > >
> > > /* translate umc channel address to soc pa, 3 parts are
> > included */
> > > --
> > > 2.17.1
