From: Alex Deucher <alexdeucher-Re5JQEeQqe8AvxtiuMwx3w@public.gmane.org>
To: amd-gfx-PD4FTy7X32lNgt0PjOBp9y5qC8QIuHrW@public.gmane.org
Cc: Tao Zhou <tao.zhou1-5C7GfCeVMHo@public.gmane.org>,
Dennis Li <dennis.li-5C7GfCeVMHo@public.gmane.org>,
Hawking Zhang <Hawking.Zhang-5C7GfCeVMHo@public.gmane.org>
Subject: [PATCH 16/26] drm/amdgpu: query umc ras error address
Date: Wed, 31 Jul 2019 12:58:08 -0500 [thread overview]
Message-ID: <20190731175818.20159-17-alexander.deucher@amd.com> (raw)
In-Reply-To: <20190731175818.20159-1-alexander.deucher-5C7GfCeVMHo@public.gmane.org>
From: Tao Zhou <tao.zhou1@amd.com>
query umc ras error address, translate it to gpu 4k page view
and save it.
Signed-off-by: Tao Zhou <tao.zhou1@amd.com>
Reviewed-by: Hawking Zhang <Hawking.Zhang@amd.com>
Reviewed-by: Dennis Li <dennis.li@amd.com>
---
drivers/gpu/drm/amd/amdgpu/umc_v6_1.c | 80 +++++++++++++++++++++++++++
1 file changed, 80 insertions(+)
diff --git a/drivers/gpu/drm/amd/amdgpu/umc_v6_1.c b/drivers/gpu/drm/amd/amdgpu/umc_v6_1.c
index e05f3e68edb0..bff1a12f2cc9 100644
--- a/drivers/gpu/drm/amd/amdgpu/umc_v6_1.c
+++ b/drivers/gpu/drm/amd/amdgpu/umc_v6_1.c
@@ -31,6 +31,16 @@
#define smnMCA_UMC0_MCUMC_ADDRT0 0x50f10
+/*
+ * (addr / 256) * 8192, the higher 26 bits in ErrorAddr
+ * is the index of 8KB block
+ */
+#define ADDR_OF_8KB_BLOCK(addr) (((addr) & ~0xffULL) << 5)
+/* channel index is the index of 256B block */
+#define ADDR_OF_256B_BLOCK(channel_index) ((channel_index) << 8)
+/* offset in 256B block */
+#define OFFSET_IN_256B_BLOCK(addr) ((addr) & 0xffULL)
+
static uint32_t
umc_v6_1_channel_idx_tbl[UMC_V6_1_UMC_INSTANCE_NUM][UMC_V6_1_CHANNEL_INSTANCE_NUM] = {
{2, 18, 11, 27}, {4, 20, 13, 29},
@@ -158,6 +168,76 @@ static void umc_v6_1_query_ras_error_count(struct amdgpu_device *adev,
umc_v6_1_disable_umc_index_mode(adev);
}
+static void umc_v6_1_query_error_address(struct amdgpu_device *adev,
+ uint32_t umc_reg_offset, uint32_t channel_index,
+ struct ras_err_data *err_data)
+{
+ uint32_t lsb;
+ uint64_t mc_umc_status, err_addr;
+ uint32_t mc_umc_status_addr;
+
+ /* skip error address process if -ENOMEM */
+ if (!err_data->err_addr)
+ return;
+
+ mc_umc_status_addr =
+ SOC15_REG_OFFSET(UMC, 0, mmMCA_UMC_UMC0_MCUMC_STATUST0);
+ mc_umc_status = RREG64(mc_umc_status_addr + umc_reg_offset);
+
+ /* calculate error address if ue/ce error is detected */
+ 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, UECC) == 1 ||
+ REG_GET_FIELD(mc_umc_status, MCA_UMC_UMC0_MCUMC_STATUST0, CECC) == 1)) {
+ err_addr = RREG64_PCIE(smnMCA_UMC0_MCUMC_ADDRT0 + umc_reg_offset * 4);
+
+ /* the lowest lsb bits should be ignored */
+ lsb = REG_GET_FIELD(err_addr, MCA_UMC_UMC0_MCUMC_ADDRT0, LSB);
+ err_addr = REG_GET_FIELD(err_addr, MCA_UMC_UMC0_MCUMC_ADDRT0, ErrorAddr);
+ err_addr &= ~((0x1ULL << lsb) - 1);
+
+ /* translate umc channel address to soc pa, 3 parts are included */
+ err_data->err_addr[err_data->err_addr_cnt] =
+ ADDR_OF_8KB_BLOCK(err_addr)
+ | ADDR_OF_256B_BLOCK(channel_index)
+ | OFFSET_IN_256B_BLOCK(err_addr);
+
+ err_data->err_addr_cnt++;
+ }
+}
+
+static void umc_v6_1_query_ras_error_address(struct amdgpu_device *adev,
+ void *ras_error_status)
+{
+ struct ras_err_data *err_data = (struct ras_err_data *)ras_error_status;
+ uint32_t umc_inst, channel_inst, umc_reg_offset;
+ uint32_t channel_index, mc_umc_status_addr;
+
+ mc_umc_status_addr =
+ SOC15_REG_OFFSET(UMC, 0, mmMCA_UMC_UMC0_MCUMC_STATUST0);
+
+ for (umc_inst = 0; umc_inst < UMC_V6_1_UMC_INSTANCE_NUM; umc_inst++) {
+ /* enable the index mode to query eror count per channel */
+ umc_v6_1_enable_umc_index_mode(adev, umc_inst);
+ for (channel_inst = 0; channel_inst < UMC_V6_1_CHANNEL_INSTANCE_NUM; channel_inst++) {
+ /* calc the register offset according to channel instance */
+ umc_reg_offset = UMC_V6_1_PER_CHANNEL_OFFSET * channel_inst;
+ /* get channel index of interleaved memory */
+ channel_index = umc_v6_1_channel_idx_tbl[umc_inst][channel_inst];
+
+ umc_v6_1_query_error_address(adev, umc_reg_offset,
+ channel_index, err_data);
+
+ /* clear umc status */
+ WREG64(mc_umc_status_addr + umc_reg_offset, 0x0ULL);
+ /* clear error address register */
+ WREG64_PCIE(smnMCA_UMC0_MCUMC_ADDRT0 + umc_reg_offset * 4, 0x0ULL);
+ }
+ }
+
+ umc_v6_1_disable_umc_index_mode(adev);
+}
+
const struct amdgpu_umc_funcs umc_v6_1_funcs = {
.query_ras_error_count = umc_v6_1_query_ras_error_count,
+ .query_ras_error_address = umc_v6_1_query_ras_error_address,
};
--
2.20.1
_______________________________________________
amd-gfx mailing list
amd-gfx@lists.freedesktop.org
https://lists.freedesktop.org/mailman/listinfo/amd-gfx
next prev parent reply other threads:[~2019-07-31 17:58 UTC|newest]
Thread overview: 28+ messages / expand[flat|nested] mbox.gz Atom feed top
2019-07-31 17:57 [PATCH 00/26] Further RAS enablement for vega20 Alex Deucher
[not found] ` <20190731175818.20159-1-alexander.deucher-5C7GfCeVMHo@public.gmane.org>
2019-07-31 17:57 ` [PATCH 01/26] drm/amdgpu: move some ras data structure to amdgpu_ras.h Alex Deucher
2019-07-31 17:57 ` [PATCH 02/26] drm/amdgpu: init RSMU and UMC ip base address for vega20 Alex Deucher
2019-07-31 17:57 ` [PATCH 03/26] drm/amdgpu: add amdgpu_umc_functions structure Alex Deucher
2019-07-31 17:57 ` [PATCH 04/26] drm/amdgpu: add rsmu v_0_0_2 ip headers Alex Deucher
2019-07-31 17:57 ` [PATCH 05/26] drm/amdgpu: add umc v6_1_1 IP headers Alex Deucher
2019-07-31 17:57 ` [PATCH 06/26] drm/amdgpu: add umc v6_1 query error count support Alex Deucher
2019-07-31 17:57 ` [PATCH 07/26] drm/amdgpu: init umc v6_1 functions for vega20 Alex Deucher
2019-07-31 17:58 ` [PATCH 08/26] drm/amdgpu: querry umc error count Alex Deucher
2019-07-31 17:58 ` [PATCH 09/26] drm/amdgpu: add ras error count after each query (v2) Alex Deucher
2019-07-31 17:58 ` [PATCH 10/26] drm/amdgpu: add RREG64/WREG64(_PCIE) operations Alex Deucher
2019-07-31 17:58 ` [PATCH 11/26] drm/amdgpu: use 64bit operation macros for umc Alex Deucher
2019-07-31 17:58 ` [PATCH 12/26] drm/amdgpu: switch to amdgpu_umc structure Alex Deucher
2019-07-31 17:58 ` [PATCH 13/26] drm/amdgpu: update algorithm of umc uncorrectable error counting Alex Deucher
2019-07-31 17:58 ` [PATCH 14/26] drm/amdgpu: add support for recording ras error address Alex Deucher
2019-07-31 17:58 ` [PATCH 15/26] drm/amdgpu: add structures for umc error address translation Alex Deucher
2019-07-31 17:58 ` Alex Deucher [this message]
2019-07-31 17:58 ` [PATCH 17/26] drm/amdgpu: allow ras interrupt callback to return error data Alex Deucher
2019-07-31 17:58 ` [PATCH 18/26] drm/amdgpu: update interrupt callback for all ras clients Alex Deucher
2019-07-31 17:58 ` [PATCH 19/26] drm/amdgpu: add check for ras error type Alex Deucher
2019-07-31 17:58 ` [PATCH 20/26] drm/amdgpu: remove ras_reserve_vram in ras injection Alex Deucher
2019-07-31 17:58 ` [PATCH 21/26] drm/amd/include: add bitfield define for EDC registers Alex Deucher
2019-07-31 17:58 ` [PATCH 22/26] drm/amd/include: add define of TCP_EDC_CNT_NEW Alex Deucher
2019-07-31 17:58 ` [PATCH 23/26] drm/amdgpu: add define for gfx ras subblock Alex Deucher
2019-07-31 17:58 ` [PATCH 24/26] drm/amdgpu: add RAS callback for gfx Alex Deucher
[not found] ` <20190731175818.20159-25-alexander.deucher-5C7GfCeVMHo@public.gmane.org>
2019-08-01 5:40 ` Kevin Wang
2019-07-31 17:58 ` [PATCH 25/26] drm/amdgpu: support gfx ras error injection and err_cnt query Alex Deucher
2019-07-31 17:58 ` [PATCH 26/26] drm/amdgpu: disable inject for failed subblocks of gfx Alex Deucher
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
Avoid top-posting and favor interleaved quoting:
https://en.wikipedia.org/wiki/Posting_style#Interleaved_style
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to=20190731175818.20159-17-alexander.deucher@amd.com \
--to=alexdeucher-re5jqeeqqe8avxtiumwx3w@public.gmane.org \
--cc=Hawking.Zhang-5C7GfCeVMHo@public.gmane.org \
--cc=amd-gfx-PD4FTy7X32lNgt0PjOBp9y5qC8QIuHrW@public.gmane.org \
--cc=dennis.li-5C7GfCeVMHo@public.gmane.org \
--cc=tao.zhou1-5C7GfCeVMHo@public.gmane.org \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox