From 59bdbabccaa470ed94aae7d94a1229c7b0ff4681 Mon Sep 17 00:00:00 2001 From: Andrzej Kacprowski Date: Thu, 5 Mar 2026 15:22:26 +0100 Subject: accel/ivpu: Remove boot params address setting via MMIO register The NPU 60XX uses the default boot params location specified in the firmware image header, consistent with earlier generations. Remove the unnecessary MMIO register write, freeing the AON register for future use. Fixes: 44e4c88951fa ("accel/ivpu: Implement warm boot flow for NPU6 and unify boot handling") Signed-off-by: Andrzej Kacprowski Reviewed-by: Karol Wachowski Signed-off-by: Karol Wachowski Link: https://patch.msgid.link/20260305142226.194995-1-andrzej.kacprowski@linux.intel.com (cherry picked from commit 81e62e7bf8b9309bf0febdf00940818f98bc23d8) Signed-off-by: Thomas Zimmermann --- drivers/accel/ivpu/ivpu_hw_40xx_reg.h | 6 ------ drivers/accel/ivpu/ivpu_hw_ip.c | 1 - 2 files changed, 7 deletions(-) (limited to 'drivers/accel') diff --git a/drivers/accel/ivpu/ivpu_hw_40xx_reg.h b/drivers/accel/ivpu/ivpu_hw_40xx_reg.h index 421242acb184..fc0ee8d637f9 100644 --- a/drivers/accel/ivpu/ivpu_hw_40xx_reg.h +++ b/drivers/accel/ivpu/ivpu_hw_40xx_reg.h @@ -121,12 +121,6 @@ #define VPU_50XX_HOST_SS_AON_PWR_ISLAND_STATUS_DLY 0x0003006cu #define VPU_50XX_HOST_SS_AON_PWR_ISLAND_STATUS_DLY_STATUS_DLY_MASK GENMASK(7, 0) -#define VPU_40XX_HOST_SS_AON_RETENTION0 0x0003000cu -#define VPU_40XX_HOST_SS_AON_RETENTION1 0x00030010u -#define VPU_40XX_HOST_SS_AON_RETENTION2 0x00030014u -#define VPU_40XX_HOST_SS_AON_RETENTION3 0x00030018u -#define VPU_40XX_HOST_SS_AON_RETENTION4 0x0003001cu - #define VPU_40XX_HOST_SS_AON_IDLE_GEN 0x00030200u #define VPU_40XX_HOST_SS_AON_IDLE_GEN_EN_MASK BIT_MASK(0) #define VPU_40XX_HOST_SS_AON_IDLE_GEN_HW_PG_EN_MASK BIT_MASK(1) diff --git a/drivers/accel/ivpu/ivpu_hw_ip.c b/drivers/accel/ivpu/ivpu_hw_ip.c index 959984c54341..37f95a0551ed 100644 --- a/drivers/accel/ivpu/ivpu_hw_ip.c +++ b/drivers/accel/ivpu/ivpu_hw_ip.c @@ -931,7 +931,6 @@ static int soc_cpu_boot_40xx(struct ivpu_device *vdev) static int soc_cpu_boot_60xx(struct ivpu_device *vdev) { - REGV_WR64(VPU_40XX_HOST_SS_AON_RETENTION1, vdev->fw->mem_bp->vpu_addr); soc_cpu_set_entry_point_40xx(vdev, vdev->fw->cold_boot_entry_point); return 0; -- cgit v1.2.3 From 6b13cb8f48a42ddf6dd98865b673a82e37ff238b Mon Sep 17 00:00:00 2001 From: Lizhi Hou Date: Tue, 10 Mar 2026 11:00:58 -0700 Subject: accel/amdxdna: Fix runtime suspend deadlock when there is pending job The runtime suspend callback drains the running job workqueue before suspending the device. If a job is still executing and calls pm_runtime_resume_and_get(), it can deadlock with the runtime suspend path. Fix this by moving pm_runtime_resume_and_get() from the job execution routine to the job submission routine, ensuring the device is resumed before the job is queued and avoiding the deadlock during runtime suspend. Fixes: 063db451832b ("accel/amdxdna: Enhance runtime power management") Reviewed-by: Mario Limonciello (AMD) Signed-off-by: Lizhi Hou Link: https://patch.msgid.link/20260310180058.336348-1-lizhi.hou@amd.com --- drivers/accel/amdxdna/aie2_ctx.c | 14 ++------------ drivers/accel/amdxdna/amdxdna_ctx.c | 10 ++++++++++ 2 files changed, 12 insertions(+), 12 deletions(-) (limited to 'drivers/accel') diff --git a/drivers/accel/amdxdna/aie2_ctx.c b/drivers/accel/amdxdna/aie2_ctx.c index afee5e667f77..c0d348884f74 100644 --- a/drivers/accel/amdxdna/aie2_ctx.c +++ b/drivers/accel/amdxdna/aie2_ctx.c @@ -165,7 +165,6 @@ aie2_sched_notify(struct amdxdna_sched_job *job) trace_xdna_job(&job->base, job->hwctx->name, "signaled fence", job->seq); - amdxdna_pm_suspend_put(job->hwctx->client->xdna); job->hwctx->priv->completed++; dma_fence_signal(fence); @@ -290,19 +289,11 @@ aie2_sched_job_run(struct drm_sched_job *sched_job) struct dma_fence *fence; int ret; - ret = amdxdna_pm_resume_get(hwctx->client->xdna); - if (ret) + if (!hwctx->priv->mbox_chann) return NULL; - if (!hwctx->priv->mbox_chann) { - amdxdna_pm_suspend_put(hwctx->client->xdna); - return NULL; - } - - if (!mmget_not_zero(job->mm)) { - amdxdna_pm_suspend_put(hwctx->client->xdna); + if (!mmget_not_zero(job->mm)) return ERR_PTR(-ESRCH); - } kref_get(&job->refcnt); fence = dma_fence_get(job->fence); @@ -333,7 +324,6 @@ aie2_sched_job_run(struct drm_sched_job *sched_job) out: if (ret) { - amdxdna_pm_suspend_put(hwctx->client->xdna); dma_fence_put(job->fence); aie2_job_put(job); mmput(job->mm); diff --git a/drivers/accel/amdxdna/amdxdna_ctx.c b/drivers/accel/amdxdna/amdxdna_ctx.c index 666dfd7b2a80..838430903a3e 100644 --- a/drivers/accel/amdxdna/amdxdna_ctx.c +++ b/drivers/accel/amdxdna/amdxdna_ctx.c @@ -17,6 +17,7 @@ #include "amdxdna_ctx.h" #include "amdxdna_gem.h" #include "amdxdna_pci_drv.h" +#include "amdxdna_pm.h" #define MAX_HWCTX_ID 255 #define MAX_ARG_COUNT 4095 @@ -445,6 +446,7 @@ put_shmem_bo: void amdxdna_sched_job_cleanup(struct amdxdna_sched_job *job) { trace_amdxdna_debug_point(job->hwctx->name, job->seq, "job release"); + amdxdna_pm_suspend_put(job->hwctx->client->xdna); amdxdna_arg_bos_put(job); amdxdna_gem_put_obj(job->cmd_bo); dma_fence_put(job->fence); @@ -482,6 +484,12 @@ int amdxdna_cmd_submit(struct amdxdna_client *client, goto cmd_put; } + ret = amdxdna_pm_resume_get(xdna); + if (ret) { + XDNA_ERR(xdna, "Resume failed, ret %d", ret); + goto put_bos; + } + idx = srcu_read_lock(&client->hwctx_srcu); hwctx = xa_load(&client->hwctx_xa, hwctx_hdl); if (!hwctx) { @@ -522,6 +530,8 @@ put_fence: dma_fence_put(job->fence); unlock_srcu: srcu_read_unlock(&client->hwctx_srcu, idx); + amdxdna_pm_suspend_put(xdna); +put_bos: amdxdna_arg_bos_put(job); cmd_put: amdxdna_gem_put_obj(job->cmd_bo); -- cgit v1.2.3 From e8ab57b56402697a9bef50b71aecc613f0d61846 Mon Sep 17 00:00:00 2001 From: Karol Wachowski Date: Mon, 23 Mar 2026 10:50:29 +0100 Subject: accel/ivpu: Add disable clock relinquish workaround for NVL-A0 Turn on disable clock relinquish workaround for Nova Lake A0. Without this workaround NPU may not power off correctly after inference, leading to unexpected system behavior. Fixes: 550f4dd2cedd ("accel/ivpu: Add support for Nova Lake's NPU") Cc: # v6.19+ Reviewed-by: Lizhi.hou Signed-off-by: Karol Wachowski Link: https://patch.msgid.link/20260323095029.64613-1-karol.wachowski@linux.intel.com --- drivers/accel/ivpu/ivpu_drv.h | 1 + drivers/accel/ivpu/ivpu_hw.c | 6 ++++-- 2 files changed, 5 insertions(+), 2 deletions(-) (limited to 'drivers/accel') diff --git a/drivers/accel/ivpu/ivpu_drv.h b/drivers/accel/ivpu/ivpu_drv.h index 5b34b6f50e69..f1b6155065ff 100644 --- a/drivers/accel/ivpu/ivpu_drv.h +++ b/drivers/accel/ivpu/ivpu_drv.h @@ -35,6 +35,7 @@ #define IVPU_HW_IP_60XX 60 #define IVPU_HW_IP_REV_LNL_B0 4 +#define IVPU_HW_IP_REV_NVL_A0 0 #define IVPU_HW_BTRS_MTL 1 #define IVPU_HW_BTRS_LNL 2 diff --git a/drivers/accel/ivpu/ivpu_hw.c b/drivers/accel/ivpu/ivpu_hw.c index d69cd0d93569..d4a9bcda4100 100644 --- a/drivers/accel/ivpu/ivpu_hw.c +++ b/drivers/accel/ivpu/ivpu_hw.c @@ -70,8 +70,10 @@ static void wa_init(struct ivpu_device *vdev) if (ivpu_hw_btrs_gen(vdev) == IVPU_HW_BTRS_MTL) vdev->wa.interrupt_clear_with_0 = ivpu_hw_btrs_irqs_clear_with_0_mtl(vdev); - if (ivpu_device_id(vdev) == PCI_DEVICE_ID_LNL && - ivpu_revision(vdev) < IVPU_HW_IP_REV_LNL_B0) + if ((ivpu_device_id(vdev) == PCI_DEVICE_ID_LNL && + ivpu_revision(vdev) < IVPU_HW_IP_REV_LNL_B0) || + (ivpu_device_id(vdev) == PCI_DEVICE_ID_NVL && + ivpu_revision(vdev) == IVPU_HW_IP_REV_NVL_A0)) vdev->wa.disable_clock_relinquish = true; if (ivpu_test_mode & IVPU_TEST_MODE_CLK_RELINQ_ENABLE) -- cgit v1.2.3