The wrong define was used to check if the hqd is still active
Signed-off-by: Oded Gabbay oded.gabbay@gmail.com --- drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v7.c | 3 ++- 1 file changed, 2 insertions(+), 1 deletion(-)
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v7.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v7.c index 2daad33..6a45330 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v7.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v7.c @@ -450,7 +450,8 @@ static int kgd_hqd_destroy(struct kgd_dev *kgd, uint32_t reset_type,
while (true) { temp = RREG32(mmCP_HQD_ACTIVE); - if (temp & CP_HQD_ACTIVE__ACTIVE__SHIFT) + if (temp & (CP_HQD_ACTIVE__ACTIVE_MASK << + CP_HQD_ACTIVE__ACTIVE__SHIFT)) break; if (timeout == 0) { pr_err("kfd: cp queue preemption time out (%dms)\n",
A logical AND operation was used during mask and shift, instead of a bitwise AND operation. This patch fixes this bug by changing the operation to bitwise AND.
Signed-off-by: Oded Gabbay oded.gabbay@gmail.com --- drivers/gpu/drm/amd/amdkfd/kfd_device_queue_manager_cik.c | 2 +- drivers/gpu/drm/amd/amdkfd/kfd_device_queue_manager_vi.c | 2 +- 2 files changed, 2 insertions(+), 2 deletions(-)
diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_device_queue_manager_cik.c b/drivers/gpu/drm/amd/amdkfd/kfd_device_queue_manager_cik.c index 23ce774..c6f435a 100644 --- a/drivers/gpu/drm/amd/amdkfd/kfd_device_queue_manager_cik.c +++ b/drivers/gpu/drm/amd/amdkfd/kfd_device_queue_manager_cik.c @@ -143,7 +143,7 @@ static void init_sdma_vm(struct device_queue_manager *dqm, struct queue *q, get_sh_mem_bases_32(qpd_to_pdd(qpd)); else value |= ((get_sh_mem_bases_nybble_64(qpd_to_pdd(qpd))) << - SDMA0_RLC0_VIRTUAL_ADDR__SHARED_BASE__SHIFT) && + SDMA0_RLC0_VIRTUAL_ADDR__SHARED_BASE__SHIFT) & SDMA0_RLC0_VIRTUAL_ADDR__SHARED_BASE_MASK;
q->properties.sdma_vm_addr = value; diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_device_queue_manager_vi.c b/drivers/gpu/drm/amd/amdkfd/kfd_device_queue_manager_vi.c index 44c38e8..7e9cae9 100644 --- a/drivers/gpu/drm/amd/amdkfd/kfd_device_queue_manager_vi.c +++ b/drivers/gpu/drm/amd/amdkfd/kfd_device_queue_manager_vi.c @@ -155,7 +155,7 @@ static void init_sdma_vm(struct device_queue_manager *dqm, struct queue *q, get_sh_mem_bases_32(qpd_to_pdd(qpd)); else value |= ((get_sh_mem_bases_nybble_64(qpd_to_pdd(qpd))) << - SDMA0_RLC0_VIRTUAL_ADDR__SHARED_BASE__SHIFT) && + SDMA0_RLC0_VIRTUAL_ADDR__SHARED_BASE__SHIFT) & SDMA0_RLC0_VIRTUAL_ADDR__SHARED_BASE_MASK;
q->properties.sdma_vm_addr = value;
On Wed, Jul 29, 2015 at 3:44 AM, Oded Gabbay oded.gabbay@gmail.com wrote:
A logical AND operation was used during mask and shift, instead of a bitwise AND operation. This patch fixes this bug by changing the operation to bitwise AND.
Signed-off-by: Oded Gabbay oded.gabbay@gmail.com
Reviewed-by: Alex Deucher alexander.deucher@amd.com
drivers/gpu/drm/amd/amdkfd/kfd_device_queue_manager_cik.c | 2 +- drivers/gpu/drm/amd/amdkfd/kfd_device_queue_manager_vi.c | 2 +- 2 files changed, 2 insertions(+), 2 deletions(-)
diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_device_queue_manager_cik.c b/drivers/gpu/drm/amd/amdkfd/kfd_device_queue_manager_cik.c index 23ce774..c6f435a 100644 --- a/drivers/gpu/drm/amd/amdkfd/kfd_device_queue_manager_cik.c +++ b/drivers/gpu/drm/amd/amdkfd/kfd_device_queue_manager_cik.c @@ -143,7 +143,7 @@ static void init_sdma_vm(struct device_queue_manager *dqm, struct queue *q, get_sh_mem_bases_32(qpd_to_pdd(qpd)); else value |= ((get_sh_mem_bases_nybble_64(qpd_to_pdd(qpd))) <<
SDMA0_RLC0_VIRTUAL_ADDR__SHARED_BASE__SHIFT) &&
SDMA0_RLC0_VIRTUAL_ADDR__SHARED_BASE__SHIFT) & SDMA0_RLC0_VIRTUAL_ADDR__SHARED_BASE_MASK; q->properties.sdma_vm_addr = value;diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_device_queue_manager_vi.c b/drivers/gpu/drm/amd/amdkfd/kfd_device_queue_manager_vi.c index 44c38e8..7e9cae9 100644 --- a/drivers/gpu/drm/amd/amdkfd/kfd_device_queue_manager_vi.c +++ b/drivers/gpu/drm/amd/amdkfd/kfd_device_queue_manager_vi.c @@ -155,7 +155,7 @@ static void init_sdma_vm(struct device_queue_manager *dqm, struct queue *q, get_sh_mem_bases_32(qpd_to_pdd(qpd)); else value |= ((get_sh_mem_bases_nybble_64(qpd_to_pdd(qpd))) <<
SDMA0_RLC0_VIRTUAL_ADDR__SHARED_BASE__SHIFT) &&
SDMA0_RLC0_VIRTUAL_ADDR__SHARED_BASE__SHIFT) & SDMA0_RLC0_VIRTUAL_ADDR__SHARED_BASE_MASK; q->properties.sdma_vm_addr = value;-- 2.4.3
On Wed, Jul 29, 2015 at 3:44 AM, Oded Gabbay oded.gabbay@gmail.com wrote:
The wrong define was used to check if the hqd is still active
Signed-off-by: Oded Gabbay oded.gabbay@gmail.com
drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v7.c | 3 ++- 1 file changed, 2 insertions(+), 1 deletion(-)
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v7.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v7.c index 2daad33..6a45330 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v7.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v7.c @@ -450,7 +450,8 @@ static int kgd_hqd_destroy(struct kgd_dev *kgd, uint32_t reset_type,
while (true) { temp = RREG32(mmCP_HQD_ACTIVE);
if (temp & CP_HQD_ACTIVE__ACTIVE__SHIFT)
if (temp & (CP_HQD_ACTIVE__ACTIVE_MASK <<CP_HQD_ACTIVE__ACTIVE__SHIFT))
It's a bit moot here since the shift is 0, but I thought the mask normally included the shift already...
#define CP_MQD_BASE_ADDR__BASE_ADDR_MASK 0xfffffffc #define CP_MQD_BASE_ADDR__BASE_ADDR__SHIFT 0x2
#define CP_HPD_EOP_CONTROL__PEND_SIG_SEM_MASK 0x80000000 #define CP_HPD_EOP_CONTROL__PEND_SIG_SEM__SHIFT 0x1f
etc.
break; if (timeout == 0) { pr_err("kfd: cp queue preemption time out (%dms)\n",-- 2.4.3
dri-devel mailing list dri-devel@lists.freedesktop.org http://lists.freedesktop.org/mailman/listinfo/dri-devel
On Wed, Jul 29, 2015 at 11:49 AM, Ilia Mirkin imirkin@alum.mit.edu wrote:
On Wed, Jul 29, 2015 at 3:44 AM, Oded Gabbay oded.gabbay@gmail.com wrote:
The wrong define was used to check if the hqd is still active
Signed-off-by: Oded Gabbay oded.gabbay@gmail.com
drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v7.c | 3 ++- 1 file changed, 2 insertions(+), 1 deletion(-)
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v7.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v7.c index 2daad33..6a45330 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v7.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v7.c @@ -450,7 +450,8 @@ static int kgd_hqd_destroy(struct kgd_dev *kgd, uint32_t reset_type,
while (true) { temp = RREG32(mmCP_HQD_ACTIVE);
if (temp & CP_HQD_ACTIVE__ACTIVE__SHIFT)
if (temp & (CP_HQD_ACTIVE__ACTIVE_MASK <<CP_HQD_ACTIVE__ACTIVE__SHIFT))It's a bit moot here since the shift is 0, but I thought the mask normally included the shift already...
#define CP_MQD_BASE_ADDR__BASE_ADDR_MASK 0xfffffffc #define CP_MQD_BASE_ADDR__BASE_ADDR__SHIFT 0x2
#define CP_HPD_EOP_CONTROL__PEND_SIG_SEM_MASK 0x80000000 #define CP_HPD_EOP_CONTROL__PEND_SIG_SEM__SHIFT 0x1f
etc.
Absolutely correct and although it is muted here, still best practice is to fix that. Thanks!
Oded
break; if (timeout == 0) { pr_err("kfd: cp queue preemption time out (%dms)\n",-- 2.4.3
dri-devel mailing list dri-devel@lists.freedesktop.org http://lists.freedesktop.org/mailman/listinfo/dri-devel
dri-devel@lists.freedesktop.org
-
Alex Deucher -
Ilia Mirkin -
Oded Gabbay