<html> <head> <meta http-equiv="Content-Type" content="text/html; charset=us-ascii"> <style type="text/css" style="display:none;"> P {margin-top:0;margin-bottom:0;} </style> </head> <body dir="ltr"> [AMD Official Use Only] <div> <div style="font-family: Calibri, Arial, Helvetica, sans-serif; font-size: 12pt; color: rgb(0, 0, 0);"> Poison mode is a global setting currently, will we set it per IP block in the future?</div> <div style="font-family: Calibri, Arial, Helvetica, sans-serif; font-size: 12pt; color: rgb(0, 0, 0);"> For example, set poison mode for GFX but fatal error mode for SDMA?</div> <div style="font-family: Calibri, Arial, Helvetica, sans-serif; font-size: 12pt; color: rgb(0, 0, 0);"> </div> <div style="font-family: Calibri, Arial, Helvetica, sans-serif; font-size: 12pt; color: rgb(0, 0, 0);"> dgpu_mode is disabled when <span style="color: rgb(32, 31, 30); font-family: "Segoe UI", "Segoe UI Web (West European)", "Segoe UI", -apple-system, BlinkMacSystemFont, Roboto, "Helvetica Neue", sans-serif; font-size: 14.6667px; background-color: rgb(255, 255, 255); display: inline !important;">connected_to_cpu is 1, is irrelevant to IP block. </div> <div style="font-family: Calibri, Arial, Helvetica, sans-serif; font-size: 12pt; color: rgb(0, 0, 0);"> </div> <div style="font-family: Calibri, Arial, Helvetica, sans-serif; font-size: 12pt; color: rgb(0, 0, 0);"> Regards,</div> <div style="font-family: Calibri, Arial, Helvetica, sans-serif; font-size: 12pt; color: rgb(0, 0, 0);"> Tao</div> <div id="appendonsend"></div> <hr style="display:inline-block;width:98%" tabindex="-1"> <div id="divRplyFwdMsg" dir="ltr">From: Zhang, Hawking <Hawking.Zhang@amd.com> Sent: Saturday, September 18, 2021 4:59 PM To: Zhou1, Tao <Tao.Zhou1@amd.com>; amd-gfx@lists.freedesktop.org <amd-gfx@lists.freedesktop.org>; Clements, John <John.Clements@amd.com>; Yang, Stanley <Stanley.Yang@amd.com> Subject: RE: [PATCH 2/3] drm/amdgpu: set poison mode for RAS <div> </div> </div> <div class="BodyFragment"> <div class="PlainText">[AMD Official Use Only] + if (amdgpu_ras_is_poison_enabled(adev)) ras_cmd->ras_in_message.init_flags.poison_mode_en = 1; - else + if (!adev->gmc.xgmi.connected_to_cpu) ras_cmd->ras_in_message.init_flags.dgpu_mode = 1; I'd expect these flags are set in enable_feature command per IP block if needed. Instead of global setting at firmware/TA initialization phase, thoughts? Regards, Hawking -----Original Message----- From: Zhou1, Tao <Tao.Zhou1@amd.com> Sent: Saturday, September 18, 2021 16:08 To: amd-gfx@lists.freedesktop.org; Zhang, Hawking <Hawking.Zhang@amd.com>; Clements, John <John.Clements@amd.com>; Yang, Stanley <Stanley.Yang@amd.com> Cc: Zhou1, Tao <Tao.Zhou1@amd.com> Subject: [PATCH 2/3] drm/amdgpu: set poison mode for RAS Add RAS poison mode flag and tell PSP RAS TA about the info. Signed-off-by: Tao Zhou <tao.zhou1@amd.com> --- drivers/gpu/drm/amd/amdgpu/amdgpu_psp.c | 4 ++-- drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c | 28 +++++++++++++++++++++++++ drivers/gpu/drm/amd/amdgpu/amdgpu_ras.h | 5 +++++ 3 files changed, 35 insertions(+), 2 deletions(-) diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_psp.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_psp.c index 7d09b28889af..140b94da2f5a 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_psp.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_psp.c @@ -1442,9 +1442,9 @@ static int psp_ras_initialize(struct psp_context *psp) ras_cmd = (struct ta_ras_shared_memory *)psp->ras_context.context.mem_context.shared_buf; memset(ras_cmd, 0, sizeof(struct ta_ras_shared_memory)); - if (psp->adev->gmc.xgmi.connected_to_cpu) + if (amdgpu_ras_is_poison_enabled(adev)) ras_cmd->ras_in_message.init_flags.poison_mode_en = 1; - else + if (!adev->gmc.xgmi.connected_to_cpu) ras_cmd->ras_in_message.init_flags.dgpu_mode = 1; ret = psp_ras_load(psp); diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c index b5332db4d287..7b7e54fdd785 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c @@ -2180,6 +2180,7 @@ int amdgpu_ras_init(struct amdgpu_device *adev) { struct amdgpu_ras *con = amdgpu_ras_get_context(adev); int r; + bool df_poison, umc_poison; if (con) return 0; @@ -2249,6 +2250,23 @@ int amdgpu_ras_init(struct amdgpu_device *adev) goto release_con; } + /* Init poison mode, the default value is false */ + if (adev->df.funcs && + adev->df.funcs->query_ras_poison_mode && + adev->umc.ras_funcs && + adev->umc.ras_funcs->query_ras_poison_mode) { + df_poison = + adev->df.funcs->query_ras_poison_mode(adev); + umc_poison = + adev->umc.ras_funcs->query_ras_poison_mode(adev); + /* Only poison is set in both DF and UMC, we can enable it */ + if (df_poison && umc_poison) + con->poison_mode_en = true; + else if (df_poison != umc_poison) + dev_warn(adev->dev, "Poison setting is inconsistent in DF/UMC(%d:%d)!\n", + df_poison, umc_poison); + } + if (amdgpu_ras_fs_init(adev)) { r = -EINVAL; goto release_con; @@ -2292,6 +2310,16 @@ static int amdgpu_persistent_edc_harvesting(struct amdgpu_device *adev, return 0; } +bool amdgpu_ras_is_poison_enabled(struct amdgpu_device *adev) { + struct amdgpu_ras *con = amdgpu_ras_get_context(adev); + + if (!con) + return false; + + return con->poison_mode_en; +} + /* helper function to handle common stuff in ip late init phase */ int amdgpu_ras_late_init(struct amdgpu_device *adev, struct ras_common_if *ras_block, diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_ras.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_ras.h index 1670467c2054..044bd19b7cce 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_ras.h +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_ras.h @@ -345,6 +345,9 @@ struct amdgpu_ras { /* disable ras error count harvest in recovery */ bool disable_ras_err_cnt_harvest; + /* is poison mode */ + bool poison_mode_en; + /* RAS count errors delayed work */ struct delayed_work ras_counte_delay_work; atomic_t ras_ue_count; @@ -640,4 +643,6 @@ void amdgpu_release_ras_context(struct amdgpu_device *adev); int amdgpu_persistent_edc_harvesting_supported(struct amdgpu_device *adev); +bool amdgpu_ras_is_poison_enabled(struct amdgpu_device *adev); + #endif -- 2.17.1 </div> </div> </div> </body> </html>