iv/linux
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

This tag contains the following changes for kernel 5.10-rc1:

Browse Source 
- release the kernel context object after we reset the device. This is
   needed to prevent a race where the firmware still has some in-flight
   transcations that require the kernel context (and its memory mappings) to
   be alive.
 
 - replace constant numbers with defines in QMAN initialization of GAUDI
 
 - correct an error message text and add a few debug messages to help debug
   issues that happen during context open and close.
 -----BEGIN PGP SIGNATURE-----
 
 iQFKBAABCgA0FiEE7TEboABC71LctBLFZR1NuKta54AFAl9t2h4WHG9kZWQuZ2Fi
 YmF5QGdtYWlsLmNvbQAKCRBlHU24q1rngKJTCAClbA79YeWImufwUr2kcs3OhPmI
 QeBCpiAMM/1chKTTLjh5RiAGWYTjEpwRn3DXunnrwdfeqvSoYPw7GxgP/oWKbY9M
 ISdGS5igw7VaFhG6PF2IpmrRt3R8qV+QOhsto54zt9ZkUwwHw2vzkFYrFYUWa1Rq
 g6NBYO9O2O07EK/wfVln9mDNMjsHSaY1hltOfY6vmU00psWGGGDYUg0QpOUJeZBw
 1+Aw5jQPmflfTdcOzEO2Dsts4mVXiLVQxgvaohTXVTbhIXvUZA+aKMj8Xd3snghT
 /SehLKHi0LWUvzpsSNAjk4nw+kAcKatP8aKjpFfKcGoIc805E6fN4tAj5uTb
 =0VFf
 -----END PGP SIGNATURE-----

Merge tag 'misc-habanalabs-next-2020-09-25' of git://people.freedesktop.org/~gabbayo/linux into char-misc-next

Oded writes:

This tag contains the following changes for kernel 5.10-rc1:

- release the kernel context object after we reset the device. This is
  needed to prevent a race where the firmware still has some in-flight
  transcations that require the kernel context (and its memory mappings) to
  be alive.

- replace constant numbers with defines in QMAN initialization of GAUDI

- correct an error message text and add a few debug messages to help debug
  issues that happen during context open and close.

* tag 'misc-habanalabs-next-2020-09-25' of git://people.freedesktop.org/~gabbayo/linux:
  habanalabs/gaudi: configure QMAN LDMA registers properly
  habanalabs: add notice of device not idle
  habanalabs: add debug messages for opening/closing context
  habanalabs: release kernel context after hw_fini
  habanalabs: correct an error message
This commit is contained in:
Greg Kroah-Hartman 2020-09-25 14:36:48 +02:00
commit 5a487cf7ef
5 changed files with 70 additions and 30 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

13
drivers/misc/habanalabs/common/context.c
View File

@ -12,6 +12,7 @@
static void hl_ctx_fini(struct hl_ctx *ctx)
{
struct hl_device *hdev = ctx->hdev;
u64 idle_mask = 0;
int i;
/*
@ -28,6 +29,8 @@ static void hl_ctx_fini(struct hl_ctx *ctx)
kfree(ctx->cs_pending);
if (ctx->asid != HL_KERNEL_ASID_ID) {
dev_dbg(hdev->dev, "closing user context %d\n", ctx->asid);
/* The engines are stopped as there is no executing CS, but the
* Coresight might be still working by accessing addresses
* related to the stopped engines. Hence stop it explicitly.
@ -40,7 +43,15 @@ static void hl_ctx_fini(struct hl_ctx *ctx)
hl_cb_va_pool_fini(ctx);
hl_vm_ctx_fini(ctx);
hl_asid_free(hdev, ctx->asid);
if ((!hdev->pldm) && (hdev->pdev) &&
(!hdev->asic_funcs->is_device_idle(hdev,
&idle_mask, NULL)))
dev_notice(hdev->dev,
"device not idle after user context is closed (0x%llx)\n",
idle_mask);
} else {
dev_dbg(hdev->dev, "closing kernel context\n");
hl_mmu_ctx_fini(ctx);
}
}
@ -168,6 +179,8 @@ int hl_ctx_init(struct hl_device *hdev, struct hl_ctx *ctx, bool is_kernel_ctx)
dev_err(hdev->dev, "ctx_init failed\n");
goto err_cb_va_pool_fini;
}
dev_dbg(hdev->dev, "create user context %d\n", ctx->asid);
}
return 0;

13
drivers/misc/habanalabs/common/device.c
View File

@ -967,14 +967,13 @@ again:
flush_workqueue(hdev->eq_wq);
}
/* Release kernel context */
if ((hard_reset) && (hl_ctx_put(hdev->kernel_ctx) == 1))
hdev->kernel_ctx = NULL;
/* Reset the H/W. It will be in idle state after this returns */
hdev->asic_funcs->hw_fini(hdev, hard_reset);
if (hard_reset) {
/* Release kernel context */
if (hl_ctx_put(hdev->kernel_ctx) == 1)
hdev->kernel_ctx = NULL;
hl_vm_fini(hdev);
hl_mmu_fini(hdev);
hl_eq_reset(hdev, &hdev->event_queue);
@ -1465,13 +1464,13 @@ void hl_device_fini(struct hl_device *hdev)
hl_cb_pool_fini(hdev);
/* Reset the H/W. It will be in idle state after this returns */
hdev->asic_funcs->hw_fini(hdev, true);
/* Release kernel context */
if ((hdev->kernel_ctx) && (hl_ctx_put(hdev->kernel_ctx) != 1))
dev_err(hdev->dev, "kernel ctx is still alive\n");
/* Reset the H/W. It will be in idle state after this returns */
hdev->asic_funcs->hw_fini(hdev, true);
hl_vm_fini(hdev);
hl_mmu_fini(hdev);

4
drivers/misc/habanalabs/common/memory.c
View File

@ -77,8 +77,8 @@ static int alloc_device_memory(struct hl_ctx *ctx, struct hl_mem_in *args,
paddr = (u64) gen_pool_alloc(vm->dram_pg_pool, total_size);
if (!paddr) {
dev_err(hdev->dev,
"failed to allocate %llu huge contiguous pages\n",
num_pgs);
"failed to allocate %llu contiguous pages with total size of %llu\n",
num_pgs, total_size);
return -ENOMEM;
}
}

62
drivers/misc/habanalabs/gaudi/gaudi.c
View File

@ -1865,9 +1865,11 @@ static void gaudi_init_pci_dma_qman(struct hl_device *hdev, int dma_id,
WREG32(mmDMA0_QM_PQ_PI_0 + q_off, 0);
WREG32(mmDMA0_QM_PQ_CI_0 + q_off, 0);
WREG32(mmDMA0_QM_CP_LDMA_TSIZE_OFFSET_0 + q_off, 0x74);
WREG32(mmDMA0_QM_CP_LDMA_SRC_BASE_LO_OFFSET_0 + q_off, 0x14);
WREG32(mmDMA0_QM_CP_LDMA_DST_BASE_LO_OFFSET_0 + q_off, 0x1C);
WREG32(mmDMA0_QM_CP_LDMA_TSIZE_OFFSET_0 + q_off, QMAN_LDMA_SIZE_OFFSET);
WREG32(mmDMA0_QM_CP_LDMA_SRC_BASE_LO_OFFSET_0 + q_off,
QMAN_LDMA_SRC_OFFSET);
WREG32(mmDMA0_QM_CP_LDMA_DST_BASE_LO_OFFSET_0 + q_off,
QMAN_LDMA_DST_OFFSET);
WREG32(mmDMA0_QM_CP_MSG_BASE0_ADDR_LO_0 + q_off, mtr_base_en_lo);
WREG32(mmDMA0_QM_CP_MSG_BASE0_ADDR_HI_0 + q_off, mtr_base_en_hi);
@ -2025,13 +2027,19 @@ static void gaudi_init_hbm_dma_qman(struct hl_device *hdev, int dma_id,
WREG32(mmDMA0_QM_PQ_PI_0 + q_off, 0);
WREG32(mmDMA0_QM_PQ_CI_0 + q_off, 0);
WREG32(mmDMA0_QM_CP_LDMA_TSIZE_OFFSET_0 + q_off, 0x81BC);
WREG32(mmDMA0_QM_CP_LDMA_SRC_BASE_LO_OFFSET_0 + q_off, 0x81B4);
WREG32(mmDMA0_QM_CP_LDMA_DST_BASE_LO_OFFSET_0 + q_off, 0x1C);
WREG32(mmDMA0_QM_CP_LDMA_TSIZE_OFFSET_0 + q_off,
QMAN_CPDMA_SIZE_OFFSET);
WREG32(mmDMA0_QM_CP_LDMA_SRC_BASE_LO_OFFSET_0 + q_off,
QMAN_CPDMA_SRC_OFFSET);
WREG32(mmDMA0_QM_CP_LDMA_DST_BASE_LO_OFFSET_0 + q_off,
QMAN_CPDMA_DST_OFFSET);
} else {
WREG32(mmDMA0_QM_CP_LDMA_TSIZE_OFFSET_0 + q_off, 0x74);
WREG32(mmDMA0_QM_CP_LDMA_SRC_BASE_LO_OFFSET_0 + q_off, 0x14);
WREG32(mmDMA0_QM_CP_LDMA_DST_BASE_LO_OFFSET_0 + q_off, 0x1C);
WREG32(mmDMA0_QM_CP_LDMA_TSIZE_OFFSET_0 + q_off,
QMAN_LDMA_SIZE_OFFSET);
WREG32(mmDMA0_QM_CP_LDMA_SRC_BASE_LO_OFFSET_0 + q_off,
QMAN_LDMA_SRC_OFFSET);
WREG32(mmDMA0_QM_CP_LDMA_DST_BASE_LO_OFFSET_0 + q_off,
QMAN_LDMA_SIZE_OFFSET);
/* Configure RAZWI IRQ */
dma_qm_err_cfg = HBM_DMA_QMAN_GLBL_ERR_CFG_MSG_EN_MASK;
@ -2135,13 +2143,19 @@ static void gaudi_init_mme_qman(struct hl_device *hdev, u32 mme_offset,
WREG32(mmMME0_QM_PQ_PI_0 + q_off, 0);
WREG32(mmMME0_QM_PQ_CI_0 + q_off, 0);
WREG32(mmMME0_QM_CP_LDMA_TSIZE_OFFSET_0 + q_off, 0x81BC);
WREG32(mmMME0_QM_CP_LDMA_SRC_BASE_LO_OFFSET_0 + q_off, 0x81B4);
WREG32(mmMME0_QM_CP_LDMA_DST_BASE_LO_OFFSET_0 + q_off, 0x1C);
WREG32(mmMME0_QM_CP_LDMA_TSIZE_OFFSET_0 + q_off,
QMAN_CPDMA_SIZE_OFFSET);
WREG32(mmMME0_QM_CP_LDMA_SRC_BASE_LO_OFFSET_0 + q_off,
QMAN_CPDMA_SRC_OFFSET);
WREG32(mmMME0_QM_CP_LDMA_DST_BASE_LO_OFFSET_0 + q_off,
QMAN_CPDMA_DST_OFFSET);
} else {
WREG32(mmMME0_QM_CP_LDMA_TSIZE_OFFSET_0 + q_off, 0x74);
WREG32(mmMME0_QM_CP_LDMA_SRC_BASE_LO_OFFSET_0 + q_off, 0x14);
WREG32(mmMME0_QM_CP_LDMA_DST_BASE_LO_OFFSET_0 + q_off, 0x1C);
WREG32(mmMME0_QM_CP_LDMA_TSIZE_OFFSET_0 + q_off,
QMAN_LDMA_SIZE_OFFSET);
WREG32(mmMME0_QM_CP_LDMA_SRC_BASE_LO_OFFSET_0 + q_off,
QMAN_LDMA_SRC_OFFSET);
WREG32(mmMME0_QM_CP_LDMA_DST_BASE_LO_OFFSET_0 + q_off,
QMAN_LDMA_DST_OFFSET);
/* Configure RAZWI IRQ */
mme_id = mme_offset /
@ -2249,13 +2263,19 @@ static void gaudi_init_tpc_qman(struct hl_device *hdev, u32 tpc_offset,
WREG32(mmTPC0_QM_PQ_PI_0 + q_off, 0);
WREG32(mmTPC0_QM_PQ_CI_0 + q_off, 0);
WREG32(mmTPC0_QM_CP_LDMA_TSIZE_OFFSET_0 + q_off, 0x81BC);
WREG32(mmTPC0_QM_CP_LDMA_SRC_BASE_LO_OFFSET_0 + q_off, 0x81B4);
WREG32(mmTPC0_QM_CP_LDMA_DST_BASE_LO_OFFSET_0 + q_off, 0x1C);
WREG32(mmTPC0_QM_CP_LDMA_TSIZE_OFFSET_0 + q_off,
QMAN_CPDMA_SIZE_OFFSET);
WREG32(mmTPC0_QM_CP_LDMA_SRC_BASE_LO_OFFSET_0 + q_off,
QMAN_CPDMA_SRC_OFFSET);
WREG32(mmTPC0_QM_CP_LDMA_DST_BASE_LO_OFFSET_0 + q_off,
QMAN_CPDMA_DST_OFFSET);
} else {
WREG32(mmTPC0_QM_CP_LDMA_TSIZE_OFFSET_0 + q_off, 0x74);
WREG32(mmTPC0_QM_CP_LDMA_SRC_BASE_LO_OFFSET_0 + q_off, 0x14);
WREG32(mmTPC0_QM_CP_LDMA_DST_BASE_LO_OFFSET_0 + q_off, 0x1C);
WREG32(mmTPC0_QM_CP_LDMA_TSIZE_OFFSET_0 + q_off,
QMAN_LDMA_SIZE_OFFSET);
WREG32(mmTPC0_QM_CP_LDMA_SRC_BASE_LO_OFFSET_0 + q_off,
QMAN_LDMA_SRC_OFFSET);
WREG32(mmTPC0_QM_CP_LDMA_DST_BASE_LO_OFFSET_0 + q_off,
QMAN_LDMA_DST_OFFSET);
/* Configure RAZWI IRQ */
tpc_id = tpc_offset /

8
drivers/misc/habanalabs/gaudi/gaudiP.h
View File

@ -84,6 +84,14 @@
#define DMA_CORE_OFFSET (mmDMA1_CORE_BASE - mmDMA0_CORE_BASE)
#define QMAN_LDMA_SRC_OFFSET (mmDMA0_CORE_SRC_BASE_LO - mmDMA0_CORE_CFG_0)
#define QMAN_LDMA_DST_OFFSET (mmDMA0_CORE_DST_BASE_LO - mmDMA0_CORE_CFG_0)
#define QMAN_LDMA_SIZE_OFFSET (mmDMA0_CORE_DST_TSIZE_0 - mmDMA0_CORE_CFG_0)
#define QMAN_CPDMA_SRC_OFFSET (mmDMA0_QM_CQ_PTR_LO_4 - mmDMA0_CORE_CFG_0)
#define QMAN_CPDMA_DST_OFFSET (mmDMA0_CORE_DST_BASE_LO - mmDMA0_CORE_CFG_0)
#define QMAN_CPDMA_SIZE_OFFSET (mmDMA0_QM_CQ_TSIZE_4 - mmDMA0_CORE_CFG_0)
#define SIF_RTR_CTRL_OFFSET (mmSIF_RTR_CTRL_1_BASE - mmSIF_RTR_CTRL_0_BASE)
#define NIF_RTR_CTRL_OFFSET (mmNIF_RTR_CTRL_1_BASE - mmNIF_RTR_CTRL_0_BASE)