ASoC: Intel: avs: Fixes for avs_pci_probe()
Subsystems (3)
Cezary Rojewski
maintainercover ASoC: Intel: avs: Fixes for avs_pci_probe()
Two small changes that aim to improve code quality around error-path area for the avs_pci_probe() function.
1) The synchronization-change is here to make sure both happy and error paths are streamlines, no functional changes there.
2) The fix addresses null-ptr-deref and similar problems is pcim_request_all_regions() fails.
Cezary Rojewski (2):
ASoC: Intel: avs: Synchronize happy and error paths in PCI probe
ASoC: Intel: avs: Fix uninitialized pointer error in probe()
sound/soc/intel/avs/core.c | 11 ++++++----- 1 file changed, 6 insertions(+), 5 deletions(-)
Signature · 1 line
2.25.1
Cezary Rojewski
maintainerPATCH 1/2 ASoC: Intel: avs: Synchronize happy and error paths in PCI probe
Error path in avs_pci_probe() does not match its happy path.
--- sound/soc/intel/avs/core.c | 8 ++++---- 1 file changed, 4 insertions(+), 4 deletions(-)
sound/soc/intel/avs/core.c +4 -4
@@ -505,11 +505,11 @@ static int avs_pci_probe(struct pci_dev *pci, const struct pci_device_id *id)return 0 ; err_i915_init: -pci_free_irq (pci,0 , adev); -pci_free_irq (pci,0 , bus); -pci_free_irq_vectors (pci); -pci_clear_master (pci);pci_set_drvdata (pci,NULL ); +pci_clear_master (pci); +pci_free_irq (pci,0 , adev); +pci_free_irq (pci,0 , bus); +pci_free_irq_vectors (pci); err_acquire_irq:snd_hdac_bus_free_stream_pages (bus);snd_hdac_ext_stream_free_all (bus);
Signature · 1 line
2.25.1
Mark Brown
maintainerRe: [PATCH 1/2] ASoC: Intel: avs: Synchronize happy and error paths in PCI probe
sound/soc/intel/avs/core.c static int avs_pci_probe(struct pci_dev *pci, const struct pci_device_id *id) · line 512
Quoted 10 lines · Cezary Rojewski
err_i915_init:sound/soc/intel/avs/core.c +4
-4-pci_free_irq (pci,0 , adev); -pci_free_irq (pci,0 , bus); -pci_free_irq_vectors (pci); -pci_clear_master (pci);pci_set_drvdata (pci,NULL ); +pci_clear_master (pci); +pci_free_irq (pci,0 , adev); +pci_free_irq (pci,0 , bus); +pci_free_irq_vectors (pci);
The ordering before clearing the driver data actually looks sensible there, though looking at the interrupt handlers quickly they don't seem to pay attention to driver data. Really you could just delete the pci_set_drvdata(), it should be redundant (and IIRC the driver core does that anyway).
Cezary Rojewski
maintainerRe: [PATCH 1/2] ASoC: Intel: avs: Synchronize happy and error paths in PCI probe
Quoted 18 lines · Mark Brown
sound/soc/intel/avs/core.c static int avs_pci_probe(struct pci_dev *pci, const struct pci_device_id *id) · line 512
Quoted 10 lines · Cezary Rojewski
err_i915_init:sound/soc/intel/avs/core.c +4
-4-pci_free_irq (pci,0 , adev); -pci_free_irq (pci,0 , bus); -pci_free_irq_vectors (pci); -pci_clear_master (pci);pci_set_drvdata (pci,NULL ); +pci_clear_master (pci); +pci_free_irq (pci,0 , adev); +pci_free_irq (pci,0 , bus); +pci_free_irq_vectors (pci);The ordering before clearing the driver data actually looks sensible there, though looking at the interrupt handlers quickly they don't seem to pay attention to driver data. Really you could just delete the
pci_set_drvdata(), it should be redundant (and IIRC the driver core does that anyway).
While I do agree that pci_set_drvdata(NULL) could be skipped, I feel better with having a true reverse order in the error path:
happy path:
pci_alloc_irq_vectors(); pci_request_irq(bus); pci_request_irq(adev); pci_set_master();
error path:
pci_clear_master(pci); pci_free_irq(pci, 0, adev); pci_free_irq(pci, 0, bus); pci_free_irq_vectors(pci);
Is there anything wrong with the approach before I send v2?
Mark Brown
maintainerRe: [PATCH 1/2] ASoC: Intel: avs: Synchronize happy and error paths in PCI probe
Quoted 6 lines · Cezary Rojewski
happy path:
pci_alloc_irq_vectors(); pci_request_irq(bus); pci_request_irq(adev); pci_set_master();
Quoted 7 lines
error path:
pci_clear_master(pci); pci_free_irq(pci, 0, adev); pci_free_irq(pci, 0, bus); pci_free_irq_vectors(pci);Is there anything wrong with the approach before I send v2?
Unless there's some issue with having master disabled but interrupts disabled that should be fine.
Cezary Rojewski
maintainerRe: [PATCH 1/2] ASoC: Intel: avs: Synchronize happy and error paths in PCI probe
Quoted 14 lines · Mark Brown
Quoted 11 lines · Cezary Rojewski
happy path:
pci_alloc_irq_vectors(); pci_request_irq(bus); pci_request_irq(adev); pci_set_master();error path:
pci_clear_master(pci); pci_free_irq(pci, 0, adev); pci_free_irq(pci, 0, bus); pci_free_irq_vectors(pci);
Quoted 4 lines
Quoted 1 line
Is there anything wrong with the approach before I send v2?
Unless there's some issue with having master disabled but interrupts disabled that should be fine.
After your comments, I've investigated the drivers/ code and it seems it's more common to put pci_set_master() early in the probe() rather than at the end of it. That may trigger an update to avs/ and would render this very patch redundant.
Because of that, I've decided to drop this patch. Will re-iterate on the subject but don't want to block the followup patch.
Good review Mark!
Cezary Rojewski
maintainerPATCH 2/2 ASoC: Intel: avs: Fix uninitialized pointer error in probe()
In reply to Cezary RojewskiIf pcim_request_all_regions() fails, error path operates on uninitialized 'bus' pointer. Found out by Coverity static analyzer.
--- sound/soc/intel/avs/core.c | 3 ++- 1 file changed, 2 insertions(+), 1 deletion(-)
sound/soc/intel/avs/core.c +2 -1
@@ -445,6 +445,8 @@ static int avs_pci_probe(struct pci_dev *pci, const struct pci_device_id *id) adev =devm_kzalloc (dev,sizeof (*adev),GFP_KERNEL );if (!adev)return -ENOMEM ; + bus = &adev->base.core; + ret =avs_bus_init (adev, pci, id);if (ret <0 ) {dev_err (dev,"failed to init avs bus: %d\n" , ret);
@@ -455,7 +457,6 @@ static int avs_pci_probe(struct pci_dev *pci, const struct pci_device_id *id)if (ret <0 )return ret; - bus = &adev->base.core; bus->addr =pci_resource_start (pci,0 ); bus->remap_addr =pci_ioremap_bar (pci,0 );if (!bus->remap_addr) {
Signature · 1 line
2.25.1