block: use generic power management
cover block: use generic power management
Linux Kernel Mentee: Remove Legacy Power Management.
The purpose of this patch series is to upgrade power management in block drivers. This has been done by upgrading .suspend() and .resume() callbacks.
The upgrade makes sure that the involvement of PCI Core does not change the order of operations executed in a driver. Thus, does not change its behavior.
In general, drivers with legacy PM, .suspend() and .resume() make use of PCI helper functions like pci_enable/disable_device_mem(), pci_set_power_state(), pci_save/restore_state(), pci_enable/disable_device(), etc. to complete their job.
The conversion requires the removal of those function calls, change the callbacks' definition accordingly and make use of dev_pm_ops structure.
All patches are compile-tested only.
Test tools:
- Compiler: gcc (GCC) 10.1.0
- allmodconfig build: make -j$(nproc) W=1 all
Vaibhav Gupta (3):
mtip32xx: use generic power management
rsxx: use generic power management
skd: use generic power management
drivers/block/mtip32xx/mtip32xx.c | 54 +++++++------------------------ drivers/block/rsxx/core.c | 9 ++++-- drivers/block/skd_main.c | 36 ++++++--------------- 3 files changed, 29 insertions(+), 70 deletions(-)
Signature · 1 line
2.27.0
PATCH 1/3 mtip32xx: use generic power management
Drivers using legacy PM have to manage PCI states and device's PM states themselves. They also need to take care of configuration registers.
With improved and powerful support of generic PM, PCI Core takes care of above mentioned, device-independent, jobs.
This driver makes use of PCI helper functions like pci_save/restore_state(), pci_disable_device(), pcim_enable_device(), pci_set_power_state() and pci_set_master() to do required operations. In generic mode, they are no longer needed.
Change function parameter in both .suspend() and .resume() to "struct device*" type. Use dev_get_drvdata() to get drv data.
Compile-tested only.
--- drivers/block/mtip32xx/mtip32xx.c | 54 +++++++------------------------ 1 file changed, 12 insertions(+), 42 deletions(-)
drivers/block/mtip32xx/mtip32xx.c +12 -42
@@ -4246,14 +4246,13 @@ static void mtip_pci_remove(struct pci_dev *pdev)* 0 Success * <0 Error */ -static int mtip_pci_suspend (struct pci_dev *pdev,pm_message_t mesg) +static int __maybe_unused mtip_pci_suspend (struct device *dev) {int rv =0 ; -struct driver_data *dd =pci_get_drvdata (pdev); +struct driver_data *dd =dev_get_drvdata (dev);if (!dd) { -dev_err (&pdev->dev, -"Driver private datastructure is NULL\n" ); +dev_err (dev,"Driver private datastructure is NULL\n" );return -EFAULT ; }
@@ -4261,21 +4260,8 @@ static int mtip_pci_suspend(struct pci_dev *pdev, pm_message_t mesg)/* Disable ports & interrupts then send standby immediate */ rv =mtip_block_suspend (dd); -if (rv <0 ) { -dev_err (&pdev->dev, -"Failed to suspend controller\n" ); -return rv; - } - -/* -* Save the pci config space to pdev structure & -* disable the device -*/ -pci_save_state (pdev); -pci_disable_device (pdev); - -/* Move to Low power state*/ -pci_set_power_state (pdev, PCI_D3hot); +if (rv <0 ) +dev_err (dev,"Failed to suspend controller\n" );return rv; }
@@ -4287,42 +4273,25 @@ static int mtip_pci_suspend(struct pci_dev *pdev, pm_message_t mesg)* 0 Success * <0 Error */ -static int mtip_pci_resume (struct pci_dev *pdev) +static int __maybe_unused mtip_pci_resume (struct device *dev) {int rv =0 ;struct driver_data *dd; - dd =pci_get_drvdata (pdev); + dd =dev_get_drvdata (dev);if (!dd) { -dev_err (&pdev->dev, -"Driver private datastructure is NULL\n" ); +dev_err (dev,"Driver private datastructure is NULL\n" );return -EFAULT ; } -/* Move the device to active State */ -pci_set_power_state (pdev,PCI_D0 ); - -/* Restore PCI configuration space */ -pci_restore_state (pdev); - -/* Enable the PCI device*/ - rv =pcim_enable_device (pdev); -if (rv <0 ) { -dev_err (&pdev->dev, -"Failed to enable card during resume\n" ); - goto err; - } -pci_set_master (pdev); -/* * Calls hbaReset, initPort, & startPort function * then enables interrupts */ rv =mtip_block_resume (dd);if (rv <0 ) -dev_err (&pdev->dev,"Unable to resume\n" ); +dev_err (dev,"Unable to resume\n" ); -err:clear_bit (MTIP_DDF_RESUME_BIT , &dd->dd_flag);return rv;
@@ -4353,14 +4322,15 @@ static const struct pci_device_id mtip_pci_tbl[] = { { 0 } }; +static SIMPLE_DEV_PM_OPS(mtip_pci_pm_ops, mtip_pci_suspend, mtip_pci_resume); +/* Structure that describes the PCI driver functions. */ static struct pci_driver mtip_pci_driver = { .name =MTIP_DRV_NAME , .id_table = mtip_pci_tbl, .probe = mtip_pci_probe, .remove = mtip_pci_remove, - .suspend = mtip_pci_suspend, - .resume = mtip_pci_resume, + .driver.pm = &mtip_pci_pm_ops, .shutdown = mtip_pci_shutdown, };
Signature · 1 line
2.27.0
PATCH 2/3 rsxx: use generic power management
In reply to Vaibhav GuptaDrivers using legacy PM have to manage PCI states and device's PM states themselves. They also need to take care of configuration registers.
With improved and powerful support of generic PM, PCI Core takes care of above mentioned, device-independent, jobs.
This driver has not defined .resume(), hence define it as NULL.
Change function parameter in both .suspend() to "struct device*" type and mark the parameter as "__attribute__((unused)) " as the function body is empty.
Compile-tested only.
--- drivers/block/rsxx/core.c | 9 +++++++-- 1 file changed, 7 insertions(+), 2 deletions(-)
drivers/block/rsxx/core.c +7 -2
@@ -1030,12 +1030,15 @@ static void rsxx_pci_remove(struct pci_dev *dev)kfree (card); } -static int rsxx_pci_suspend (struct pci_dev *dev,pm_message_t state) +static int __maybe_unused rsxx_pci_suspend ( + __attribute__((unused))struct device *dev) {/* We don't support suspend at this time. */ return -ENOSYS ; } +#define rsxx_pci_resume NULL +static void rsxx_pci_shutdown (struct pci_dev *dev) {struct rsxx_cardinfo *card =pci_get_drvdata (dev);
@@ -1071,12 +1074,14 @@ static const struct pci_device_id rsxx_pci_ids[] = {MODULE_DEVICE_TABLE (pci, rsxx_pci_ids); +static SIMPLE_DEV_PM_OPS(rsxx_pci_pm_ops, rsxx_pci_suspend, rsxx_pci_resume); +static struct pci_driver rsxx_pci_driver = { .name =DRIVER_NAME , .id_table = rsxx_pci_ids, .probe = rsxx_pci_probe, .remove = rsxx_pci_remove, - .suspend = rsxx_pci_suspend, + .driver.pm = &rsxx_pci_pm_ops, .shutdown = rsxx_pci_shutdown, .err_handler = &rsxx_err_handler, };
Signature · 1 line
2.27.0
PATCH 3/3 skd: use generic power management
In reply to Vaibhav GuptaDrivers using legacy PM have to manage PCI states and device's PM states themselves. They also need to take care of configuration registers.
With improved and powerful support of generic PM, PCI Core takes care of above mentioned, device-independent, jobs.
This driver makes use of PCI helper functions like pci_save/restore_state(), pci_enable/disable_device(), pci_request/release_regions(), pci_set_power_state() and pci_set_master() to do required operations. In generic mode, they are no longer needed.
Change function parameter in both .suspend() and .resume() to "struct device*" type. Use to_pci_dev() to get "struct pci_dev*" variable.
Compile-tested only.
--- drivers/block/skd_main.c | 36 ++++++++++-------------------------- 1 file changed, 10 insertions(+), 26 deletions(-)
drivers/block/skd_main.c +10 -26
@@ -3315,12 +3315,12 @@ static void skd_pci_remove(struct pci_dev *pdev)return ; } -static int skd_pci_suspend (struct pci_dev *pdev,pm_message_t state) +static int __maybe_unused skd_pci_suspend (struct device *dev) {int i; -struct skd_device *skdev; +struct pci_dev *pdev =to_pci_dev (dev); +struct skd_device *skdev =pci_get_drvdata (pdev); - skdev =pci_get_drvdata (pdev);if (!skdev) {dev_err (&pdev->dev,"no device data for PCI\n" );return -EIO ;
@@ -3337,35 +3337,23 @@ static int skd_pci_suspend(struct pci_dev *pdev, pm_message_t state)if (skdev->pcie_error_reporting_is_enabled)pci_disable_pcie_error_reporting (pdev); -pci_release_regions (pdev); -pci_save_state (pdev); -pci_disable_device (pdev); -pci_set_power_state (pdev,pci_choose_state (pdev, state));return 0 ; } -static int skd_pci_resume (struct pci_dev *pdev) +static int __maybe_unused skd_pci_resume (struct device *dev) {int i;int rc =0 ; -struct skd_device *skdev; - skdev =pci_get_drvdata (pdev); +struct pci_dev *pdev =to_pci_dev (dev); +struct skd_device *skdev =pci_get_drvdata (pdev);if (!skdev) {dev_err (&pdev->dev,"no device data for PCI\n" );return -1 ; } -pci_set_power_state (pdev,PCI_D0 ); -pci_enable_wake (pdev,PCI_D0 ,0 ); -pci_restore_state (pdev); +device_wakeup_disable (dev); - rc =pci_enable_device (pdev); -if (rc) -return rc; - rc =pci_request_regions (pdev,DRV_NAME ); -if (rc) - goto err_out; rc =dma_set_mask_and_coherent (&pdev->dev,DMA_BIT_MASK (64 ));if (rc) rc =dma_set_mask_and_coherent (&pdev->dev,DMA_BIT_MASK (32 ));
@@ -3374,7 +3362,6 @@ static int skd_pci_resume(struct pci_dev *pdev) goto err_out_regions; } -pci_set_master (pdev); rc =pci_enable_pcie_error_reporting (pdev);if (rc) { dev_err(&pdev->dev,
@@ -3427,10 +3414,6 @@ static int skd_pci_resume(struct pci_dev *pdev)pci_disable_pcie_error_reporting (pdev); err_out_regions: -pci_release_regions (pdev); - -err_out: -pci_disable_device (pdev);return rc; }
@@ -3450,13 +3433,14 @@ static void skd_pci_shutdown(struct pci_dev *pdev)skd_stop_device (skdev); } +static SIMPLE_DEV_PM_OPS(skd_pci_pm_ops, skd_pci_suspend, skd_pci_resume); +static struct pci_driver skd_driver = { .name =DRV_NAME , .id_table = skd_pci_tbl, .probe = skd_pci_probe, .remove = skd_pci_remove, - .suspend = skd_pci_suspend, - .resume = skd_pci_resume, + .driver.pm = &skd_pci_pm_ops, .shutdown = skd_pci_shutdown, };
Signature · 1 line
2.27.0
Re: [PATCH v1 3/3] skd: use generic power management
Quoted 12 lines · Vaibhav Gupta
Drivers using legacy PM have to manage PCI states and device's PM states themselves. They also need to take care of configuration registers.
With improved and powerful support of generic PM, PCI Core takes care of above mentioned, device-independent, jobs.
This driver makes use of PCI helper functions like pci_save/restore_state(), pci_enable/disable_device(), pci_request/release_regions(),
pci_set_power_state()andpci_set_master()to do required operations. In generic mode, they are no longer needed.
Quoted 4 lines
Change function parameter in both .suspend() and .resume() to "struct device*" type. Use
to_pci_dev()to get "struct pci_dev*" variable.Compile-tested only.
I do not think this belongs into the commit message. This was mentioned in the cover letter.
Quoted 11 lines
Signed-off-by Vaibhav Gupta <[email protected]>--- drivers/block/skd_main.c | 36 ++++++++++-------------------------- 1 file changed, 10 insertions(+), 26 deletions(-)drivers/block/skd_main.c +0
-0
drivers/block/skd_main.c static void skd_pci_remove(struct pci_dev *pdev) · line 3322
Quoted 10 lines
drivers/block/skd_main.c +3
-2@@ -3315,12 +3315,12 @@ static void skd_pci_remove(struct pci_dev *pdev)return ; } -static int skd_pci_suspend (struct pci_dev *pdev,pm_message_t state) +static int __maybe_unused skd_pci_suspend (struct device *dev) {int i; -struct skd_device *skdev; +struct pci_dev *pdev =to_pci_dev (dev); +struct skd_device *skdev =pci_get_drvdata (pdev);
You can just add the pdev declaration here. The other 2 changes are not needed.
drivers/block/skd_main.c static void skd_pci_remove(struct pci_dev *pdev) · line 3322
Quoted 2 lines
drivers/block/skd_main.c +0
-1- skdev =pci_get_drvdata (pdev);
You can keep this as is.
Quoted 12 lines
drivers/block/skd_main.c +0
-4if (!skdev) {dev_err (&pdev->dev,"no device data for PCI\n" );return -EIO ;@@ -3337,35 +3337,23 @@ static int skd_pci_suspend(struct pci_dev *pdev, pm_message_t state)if (skdev->pcie_error_reporting_is_enabled)pci_disable_pcie_error_reporting (pdev); -pci_release_regions (pdev); -pci_save_state (pdev); -pci_disable_device (pdev); -pci_set_power_state (pdev,pci_choose_state (pdev, state));return 0 ;
drivers/block/skd_main.c static int skd_pci_suspend(struct pci_dev *pdev, pm_message_t state) · line 3353
Quoted 10 lines
drivers/block/skd_main.c +1
-3} -static int skd_pci_resume (struct pci_dev *pdev) +static int __maybe_unused skd_pci_resume (struct device *dev) {int i;int rc =0 ; -struct skd_device *skdev; - skdev =pci_get_drvdata (pdev);
Same comment as above. Keep these as is and add only pdev declaration.
Quoted 58 lines
drivers/block/skd_main.c +6
-16+struct pci_dev *pdev =to_pci_dev (dev); +struct skd_device *skdev =pci_get_drvdata (pdev);if (!skdev) {dev_err (&pdev->dev,"no device data for PCI\n" );return -1 ; } -pci_set_power_state (pdev,PCI_D0 ); -pci_enable_wake (pdev,PCI_D0 ,0 ); -pci_restore_state (pdev); +device_wakeup_disable (dev); - rc =pci_enable_device (pdev); -if (rc) -return rc; - rc =pci_request_regions (pdev,DRV_NAME ); -if (rc) - goto err_out; rc =dma_set_mask_and_coherent (&pdev->dev,DMA_BIT_MASK (64 ));if (rc) rc =dma_set_mask_and_coherent (&pdev->dev,DMA_BIT_MASK (32 ));@@ -3374,7 +3362,6 @@ static int skd_pci_resume(struct pci_dev *pdev) goto err_out_regions; } -pci_set_master (pdev); rc =pci_enable_pcie_error_reporting (pdev);if (rc) { dev_err(&pdev->dev,@@ -3427,10 +3414,6 @@ static int skd_pci_resume(struct pci_dev *pdev)pci_disable_pcie_error_reporting (pdev); err_out_regions: -pci_release_regions (pdev); - -err_out: -pci_disable_device (pdev);return rc; }@@ -3450,13 +3433,14 @@ static void skd_pci_shutdown(struct pci_dev *pdev)skd_stop_device (skdev); } +static SIMPLE_DEV_PM_OPS(skd_pci_pm_ops, skd_pci_suspend, skd_pci_resume); +static struct pci_driver skd_driver = { .name =DRV_NAME , .id_table = skd_pci_tbl, .probe = skd_pci_probe, .remove = skd_pci_remove, - .suspend = skd_pci_suspend, - .resume = skd_pci_resume, + .driver.pm = &skd_pci_pm_ops, .shutdown = skd_pci_shutdown, };
Signature · 2 lines
Damien Le Moal Western Digital Research
Re: [PATCH v1 3/3] skd: use generic power management
Quoted 13 lines · Damien Le Moal
Quoted 12 lines · Vaibhav Gupta
Drivers using legacy PM have to manage PCI states and device's PM states themselves. They also need to take care of configuration registers.
With improved and powerful support of generic PM, PCI Core takes care of above mentioned, device-independent, jobs.
This driver makes use of PCI helper functions like pci_save/restore_state(), pci_enable/disable_device(), pci_request/release_regions(),
pci_set_power_state()andpci_set_master()to do required operations. In generic mode, they are no longer needed.
Quoted 8 lines
Quoted 4 lines
Change function parameter in both .suspend() and .resume() to "struct device*" type. Use
to_pci_dev()to get "struct pci_dev*" variable.Compile-tested only.
I do not think this belongs into the commit message. This was mentioned in the cover letter.
The messages in cover letter and commit message are bit similar. But the commit message has patch specific changes mentioned in it.
Quoted 14 lines
Quoted 14 lines
Signed-off-by Vaibhav Gupta <[email protected]>--- drivers/block/skd_main.c | 36 ++++++++++-------------------------- 1 file changed, 10 insertions(+), 26 deletions(-)drivers/block/skd_main.c +0
-0@@ -3315,12 +3315,12 @@ static void skd_pci_remove(struct pci_dev *pdev)return ; }
Quoted 10 lines
drivers/block/skd_main.c static void skd_pci_remove(struct pci_dev *pdev) · line 3322
Quoted 7 lines
drivers/block/skd_main.c +3
-2-static int skd_pci_suspend (struct pci_dev *pdev,pm_message_t state) +static int __maybe_unused skd_pci_suspend (struct device *dev) {int i; -struct skd_device *skdev; +struct pci_dev *pdev =to_pci_dev (dev); +struct skd_device *skdev =pci_get_drvdata (pdev);You can just add the pdev declaration here. The other 2 changes are not needed.
Okay.
Quoted 26 lines
drivers/block/skd_main.c static void skd_pci_remove(struct pci_dev *pdev) · line 3322
Quoted 2 lines
drivers/block/skd_main.c +0
-1- skdev =pci_get_drvdata (pdev);You can keep this as is.
Quoted 21 lines
drivers/block/skd_main.c +1
-6if (!skdev) {dev_err (&pdev->dev,"no device data for PCI\n" );return -EIO ;@@ -3337,35 +3337,23 @@ static int skd_pci_suspend(struct pci_dev *pdev, pm_message_t state)if (skdev->pcie_error_reporting_is_enabled)pci_disable_pcie_error_reporting (pdev); -pci_release_regions (pdev); -pci_save_state (pdev); -pci_disable_device (pdev); -pci_set_power_state (pdev,pci_choose_state (pdev, state));return 0 ; } -static int skd_pci_resume (struct pci_dev *pdev) +static int __maybe_unused skd_pci_resume (struct device *dev) {int i;int rc =0 ; -struct skd_device *skdev;
Quoted 4 lines
drivers/block/skd_main.c static int skd_pci_suspend(struct pci_dev *pdev, pm_message_t state) · line 3353
Quoted 1 line
drivers/block/skd_main.c +0
-1- skdev =pci_get_drvdata (pdev);Same comment as above. Keep these as is and add only pdev declaration.
Okay.
Sending v2 with suggested changes.
--Vaibhav Gupta
Quoted 63 lines
Quoted 58 lines
drivers/block/skd_main.c +6
-16+struct pci_dev *pdev =to_pci_dev (dev); +struct skd_device *skdev =pci_get_drvdata (pdev);if (!skdev) {dev_err (&pdev->dev,"no device data for PCI\n" );return -1 ; } -pci_set_power_state (pdev,PCI_D0 ); -pci_enable_wake (pdev,PCI_D0 ,0 ); -pci_restore_state (pdev); +device_wakeup_disable (dev); - rc =pci_enable_device (pdev); -if (rc) -return rc; - rc =pci_request_regions (pdev,DRV_NAME ); -if (rc) - goto err_out; rc =dma_set_mask_and_coherent (&pdev->dev,DMA_BIT_MASK (64 ));if (rc) rc =dma_set_mask_and_coherent (&pdev->dev,DMA_BIT_MASK (32 ));@@ -3374,7 +3362,6 @@ static int skd_pci_resume(struct pci_dev *pdev) goto err_out_regions; } -pci_set_master (pdev); rc =pci_enable_pcie_error_reporting (pdev);if (rc) { dev_err(&pdev->dev,@@ -3427,10 +3414,6 @@ static int skd_pci_resume(struct pci_dev *pdev)pci_disable_pcie_error_reporting (pdev); err_out_regions: -pci_release_regions (pdev); - -err_out: -pci_disable_device (pdev);return rc; }@@ -3450,13 +3433,14 @@ static void skd_pci_shutdown(struct pci_dev *pdev)skd_stop_device (skdev); } +static SIMPLE_DEV_PM_OPS(skd_pci_pm_ops, skd_pci_suspend, skd_pci_resume); +static struct pci_driver skd_driver = { .name =DRV_NAME , .id_table = skd_pci_tbl, .probe = skd_pci_probe, .remove = skd_pci_remove, - .suspend = skd_pci_suspend, - .resume = skd_pci_resume, + .driver.pm = &skd_pci_pm_ops, .shutdown = skd_pci_shutdown, };Signature · 2 lines
Damien Le Moal Western Digital Research
Re: [PATCH v1 3/3] skd: use generic power management
Quoted 17 lines · Vaibhav Gupta
Quoted 16 lines · Damien Le Moal
Quoted 15 lines · Vaibhav Gupta
Drivers using legacy PM have to manage PCI states and device's PM states themselves. They also need to take care of configuration registers.
With improved and powerful support of generic PM, PCI Core takes care of above mentioned, device-independent, jobs.
This driver makes use of PCI helper functions like pci_save/restore_state(), pci_enable/disable_device(), pci_request/release_regions(),
pci_set_power_state()andpci_set_master()to do required operations. In generic mode, they are no longer needed.Change function parameter in both .suspend() and .resume() to "struct device*" type. Use
to_pci_dev()to get "struct pci_dev*" variable.
Quoted 7 lines
Quoted 5 lines
Quoted 1 line
Compile-tested only.
I do not think this belongs into the commit message. This was mentioned in the cover letter.
The messages in cover letter and commit message are bit similar. But the commit message has patch specific changes mentioned in it.
My point was about the "compile-tested only" mention in the commit message. That should be mentioned in the cover letter only. The goal of the patch review process is also to get these patches tested by others if you do not have access to the hardware. That is fine, and I can test for the skd driver. But a patch mentioning that it is untested cannot be applied, for obvious reasons :)
So I only requested that you remove the "compiled tested only" note. The commit message could be simpler too, see followup comment.
Quoted 123 lines
Quoted 14 lines
Quoted 14 lines
Signed-off-by Vaibhav Gupta <[email protected]>--- drivers/block/skd_main.c | 36 ++++++++++-------------------------- 1 file changed, 10 insertions(+), 26 deletions(-)drivers/block/skd_main.c +0
-0@@ -3315,12 +3315,12 @@ static void skd_pci_remove(struct pci_dev *pdev)return ; }Quoted 10 lines
drivers/block/skd_main.c static void skd_pci_remove(struct pci_dev *pdev) · line 3322
Quoted 7 lines
drivers/block/skd_main.c +3
-2-static int skd_pci_suspend (struct pci_dev *pdev,pm_message_t state) +static int __maybe_unused skd_pci_suspend (struct device *dev) {int i; -struct skd_device *skdev; +struct pci_dev *pdev =to_pci_dev (dev); +struct skd_device *skdev =pci_get_drvdata (pdev);You can just add the pdev declaration here. The other 2 changes are not needed.
Okay.
Quoted 26 lines
drivers/block/skd_main.c static void skd_pci_remove(struct pci_dev *pdev) · line 3322
Quoted 2 lines
drivers/block/skd_main.c +0
-1- skdev =pci_get_drvdata (pdev);You can keep this as is.
Quoted 21 lines
drivers/block/skd_main.c +1
-6if (!skdev) {dev_err (&pdev->dev,"no device data for PCI\n" );return -EIO ;@@ -3337,35 +3337,23 @@ static int skd_pci_suspend(struct pci_dev *pdev, pm_message_t state)if (skdev->pcie_error_reporting_is_enabled)pci_disable_pcie_error_reporting (pdev); -pci_release_regions (pdev); -pci_save_state (pdev); -pci_disable_device (pdev); -pci_set_power_state (pdev,pci_choose_state (pdev, state));return 0 ; } -static int skd_pci_resume (struct pci_dev *pdev) +static int __maybe_unused skd_pci_resume (struct device *dev) {int i;int rc =0 ; -struct skd_device *skdev;Quoted 4 lines
drivers/block/skd_main.c static int skd_pci_suspend(struct pci_dev *pdev, pm_message_t state) · line 3353
Quoted 1 line
drivers/block/skd_main.c +0
-1- skdev =pci_get_drvdata (pdev);Same comment as above. Keep these as is and add only pdev declaration.
Okay.
Sending v2 with suggested changes.--Vaibhav Gupta
Quoted 63 lines
Quoted 58 lines
drivers/block/skd_main.c +6
-16+struct pci_dev *pdev =to_pci_dev (dev); +struct skd_device *skdev =pci_get_drvdata (pdev);if (!skdev) {dev_err (&pdev->dev,"no device data for PCI\n" );return -1 ; } -pci_set_power_state (pdev,PCI_D0 ); -pci_enable_wake (pdev,PCI_D0 ,0 ); -pci_restore_state (pdev); +device_wakeup_disable (dev); - rc =pci_enable_device (pdev); -if (rc) -return rc; - rc =pci_request_regions (pdev,DRV_NAME ); -if (rc) - goto err_out; rc =dma_set_mask_and_coherent (&pdev->dev,DMA_BIT_MASK (64 ));if (rc) rc =dma_set_mask_and_coherent (&pdev->dev,DMA_BIT_MASK (32 ));@@ -3374,7 +3362,6 @@ static int skd_pci_resume(struct pci_dev *pdev) goto err_out_regions; } -pci_set_master (pdev); rc =pci_enable_pcie_error_reporting (pdev);if (rc) { dev_err(&pdev->dev,@@ -3427,10 +3414,6 @@ static int skd_pci_resume(struct pci_dev *pdev)pci_disable_pcie_error_reporting (pdev); err_out_regions: -pci_release_regions (pdev); - -err_out: -pci_disable_device (pdev);return rc; }@@ -3450,13 +3433,14 @@ static void skd_pci_shutdown(struct pci_dev *pdev)skd_stop_device (skdev); } +static SIMPLE_DEV_PM_OPS(skd_pci_pm_ops, skd_pci_suspend, skd_pci_resume); +static struct pci_driver skd_driver = { .name =DRV_NAME , .id_table = skd_pci_tbl, .probe = skd_pci_probe, .remove = skd_pci_remove, - .suspend = skd_pci_suspend, - .resume = skd_pci_resume, + .driver.pm = &skd_pci_pm_ops, .shutdown = skd_pci_shutdown, };Signature · 2 lines
Damien Le Moal Western Digital Research
Signature · 2 lines
Damien Le Moal Western Digital Research
Re: [PATCH v1 3/3] skd: use generic power management
Quoted 14 lines · Damien Le Moal
Quoted 5 lines · Vaibhav Gupta
Quoted 4 lines · Damien Le Moal
Quoted 3 lines · Vaibhav Gupta
Change function parameter in both .suspend() and .resume() to "struct device*" type. Use
to_pci_dev()to get "struct pci_dev*" variable.Quoted 7 lines
Quoted 5 lines
Quoted 1 line
Compile-tested only.
I do not think this belongs into the commit message. This was mentioned in the cover letter.
The messages in cover letter and commit message are bit similar. But the commit message has patch specific changes mentioned in it.
Quoted 9 lines
My point was about the "compile-tested only" mention in the commit message. That should be mentioned in the cover letter only. The goal of the patch review process is also to get these patches tested by others if you do not have access to the hardware. That is fine, and I can test for the skd driver. But a patch mentioning that it is untested cannot be applied, for obvious reasons :)
So I only requested that you remove the "compiled tested only" note. The commit message could be simpler too, see followup comment.
Oh okay. Thanks for clarification and pointing this out :) . From now on, I will put "Compile-tested only." message only in cover letters.
Thanks.
--Vaibhav Gupta
Quoted 5 lines
Signature · 3 lines
-- Damien Le Moal Western Digital Research
v2 cover block: use generic power management
In reply to Damien Le MoalLinux Kernel Mentee: Remove Legacy Power Management.
The purpose of this patch series is to upgrade power management in block drivers. This has been done by upgrading .suspend() and .resume() callbacks.
The upgrade makes sure that the involvement of PCI Core does not change the order of operations executed in a driver. Thus, does not change its behavior.
In general, drivers with legacy PM, .suspend() and .resume() make use of PCI helper functions like pci_enable/disable_device_mem(), pci_set_power_state(), pci_save/restore_state(), pci_enable/disable_device(), etc. to complete their job.
The conversion requires the removal of those function calls, change the callbacks' definition accordingly and make use of dev_pm_ops structure.
All patches are compile-tested only.
Test tools:
- Compiler: gcc (GCC) 10.1.0
- allmodconfig build: make -j$(nproc) W=1 all
Vaibhav Gupta (3):
mtip32xx: use generic power management
rsxx: use generic power management
skd: use generic power management
drivers/block/mtip32xx/mtip32xx.c | 54 +++++++------------------------ drivers/block/rsxx/core.c | 9 ++++-- drivers/block/skd_main.c | 30 +++++------------ 3 files changed, 27 insertions(+), 66 deletions(-)
Signature · 1 line
2.27.0
PATCH v2 1/3 mtip32xx: use generic power management
Drivers using legacy PM have to manage PCI states and device's PM states themselves. They also need to take care of configuration registers.
With improved and powerful support of generic PM, PCI Core takes care of above mentioned, device-independent, jobs.
This driver makes use of PCI helper functions like pci_save/restore_state(), pci_disable_device(), pcim_enable_device(), pci_set_power_state() and pci_set_master() to do required operations. In generic mode, they are no longer needed.
Change function parameter in both .suspend() and .resume() to "struct device*" type. Use dev_get_drvdata() to get drv data.
Compile-tested only.
--- drivers/block/mtip32xx/mtip32xx.c | 54 +++++++------------------------ 1 file changed, 12 insertions(+), 42 deletions(-)
drivers/block/mtip32xx/mtip32xx.c +12 -42
@@ -4246,14 +4246,13 @@ static void mtip_pci_remove(struct pci_dev *pdev)* 0 Success * <0 Error */ -static int mtip_pci_suspend (struct pci_dev *pdev,pm_message_t mesg) +static int __maybe_unused mtip_pci_suspend (struct device *dev) {int rv =0 ; -struct driver_data *dd =pci_get_drvdata (pdev); +struct driver_data *dd =dev_get_drvdata (dev);if (!dd) { -dev_err (&pdev->dev, -"Driver private datastructure is NULL\n" ); +dev_err (dev,"Driver private datastructure is NULL\n" );return -EFAULT ; }
@@ -4261,21 +4260,8 @@ static int mtip_pci_suspend(struct pci_dev *pdev, pm_message_t mesg)/* Disable ports & interrupts then send standby immediate */ rv =mtip_block_suspend (dd); -if (rv <0 ) { -dev_err (&pdev->dev, -"Failed to suspend controller\n" ); -return rv; - } - -/* -* Save the pci config space to pdev structure & -* disable the device -*/ -pci_save_state (pdev); -pci_disable_device (pdev); - -/* Move to Low power state*/ -pci_set_power_state (pdev, PCI_D3hot); +if (rv <0 ) +dev_err (dev,"Failed to suspend controller\n" );return rv; }
@@ -4287,42 +4273,25 @@ static int mtip_pci_suspend(struct pci_dev *pdev, pm_message_t mesg)* 0 Success * <0 Error */ -static int mtip_pci_resume (struct pci_dev *pdev) +static int __maybe_unused mtip_pci_resume (struct device *dev) {int rv =0 ;struct driver_data *dd; - dd =pci_get_drvdata (pdev); + dd =dev_get_drvdata (dev);if (!dd) { -dev_err (&pdev->dev, -"Driver private datastructure is NULL\n" ); +dev_err (dev,"Driver private datastructure is NULL\n" );return -EFAULT ; } -/* Move the device to active State */ -pci_set_power_state (pdev,PCI_D0 ); - -/* Restore PCI configuration space */ -pci_restore_state (pdev); - -/* Enable the PCI device*/ - rv =pcim_enable_device (pdev); -if (rv <0 ) { -dev_err (&pdev->dev, -"Failed to enable card during resume\n" ); - goto err; - } -pci_set_master (pdev); -/* * Calls hbaReset, initPort, & startPort function * then enables interrupts */ rv =mtip_block_resume (dd);if (rv <0 ) -dev_err (&pdev->dev,"Unable to resume\n" ); +dev_err (dev,"Unable to resume\n" ); -err:clear_bit (MTIP_DDF_RESUME_BIT , &dd->dd_flag);return rv;
@@ -4353,14 +4322,15 @@ static const struct pci_device_id mtip_pci_tbl[] = { { 0 } }; +static SIMPLE_DEV_PM_OPS(mtip_pci_pm_ops, mtip_pci_suspend, mtip_pci_resume); +/* Structure that describes the PCI driver functions. */ static struct pci_driver mtip_pci_driver = { .name =MTIP_DRV_NAME , .id_table = mtip_pci_tbl, .probe = mtip_pci_probe, .remove = mtip_pci_remove, - .suspend = mtip_pci_suspend, - .resume = mtip_pci_resume, + .driver.pm = &mtip_pci_pm_ops, .shutdown = mtip_pci_shutdown, };
Signature · 1 line
2.27.0
PATCH v2 2/3 rsxx: use generic power management
In reply to Vaibhav GuptaDrivers using legacy PM have to manage PCI states and device's PM states themselves. They also need to take care of configuration registers.
With improved and powerful support of generic PM, PCI Core takes care of above mentioned, device-independent, jobs.
This driver has not defined .resume(), hence define it as NULL.
Change function parameter in both .suspend() to "struct device*" type and mark the parameter as "__attribute__((unused)) " as the function body is empty.
Compile-tested only.
--- drivers/block/rsxx/core.c | 9 +++++++-- 1 file changed, 7 insertions(+), 2 deletions(-)
drivers/block/rsxx/core.c +7 -2
@@ -1030,12 +1030,15 @@ static void rsxx_pci_remove(struct pci_dev *dev)kfree (card); } -static int rsxx_pci_suspend (struct pci_dev *dev,pm_message_t state) +static int __maybe_unused rsxx_pci_suspend ( + __attribute__((unused))struct device *dev) {/* We don't support suspend at this time. */ return -ENOSYS ; } +#define rsxx_pci_resume NULL +static void rsxx_pci_shutdown (struct pci_dev *dev) {struct rsxx_cardinfo *card =pci_get_drvdata (dev);
@@ -1071,12 +1074,14 @@ static const struct pci_device_id rsxx_pci_ids[] = {MODULE_DEVICE_TABLE (pci, rsxx_pci_ids); +static SIMPLE_DEV_PM_OPS(rsxx_pci_pm_ops, rsxx_pci_suspend, rsxx_pci_resume); +static struct pci_driver rsxx_pci_driver = { .name =DRIVER_NAME , .id_table = rsxx_pci_ids, .probe = rsxx_pci_probe, .remove = rsxx_pci_remove, - .suspend = rsxx_pci_suspend, + .driver.pm = &rsxx_pci_pm_ops, .shutdown = rsxx_pci_shutdown, .err_handler = &rsxx_err_handler, };
Signature · 1 line
2.27.0
PATCH v2 3/3 skd: use generic power management
In reply to Vaibhav GuptaDrivers using legacy PM have to manage PCI states and device's PM states themselves. They also need to take care of configuration registers.
With improved and powerful support of generic PM, PCI Core takes care of above mentioned, device-independent, jobs.
This driver makes use of PCI helper functions like pci_save/restore_state(), pci_enable/disable_device(), pci_request/release_regions(), pci_set_power_state() and pci_set_master() to do required operations. In generic mode, they are no longer needed.
Change function parameter in both .suspend() and .resume() to "struct device*" type. Use to_pci_dev() to get "struct pci_dev*" variable.
Compile-tested only.
--- drivers/block/skd_main.c | 30 ++++++++---------------------- 1 file changed, 8 insertions(+), 22 deletions(-)
drivers/block/skd_main.c +8 -22
@@ -3315,10 +3315,11 @@ static void skd_pci_remove(struct pci_dev *pdev)return ; } -static int skd_pci_suspend (struct pci_dev *pdev,pm_message_t state) +static int __maybe_unused skd_pci_suspend (struct device *dev) {int i;struct skd_device *skdev; +struct pci_dev *pdev =to_pci_dev (dev); skdev =pci_get_drvdata (pdev);if (!skdev) {
@@ -3337,18 +3338,15 @@ static int skd_pci_suspend(struct pci_dev *pdev, pm_message_t state)if (skdev->pcie_error_reporting_is_enabled)pci_disable_pcie_error_reporting (pdev); -pci_release_regions (pdev); -pci_save_state (pdev); -pci_disable_device (pdev); -pci_set_power_state (pdev,pci_choose_state (pdev, state));return 0 ; } -static int skd_pci_resume (struct pci_dev *pdev) +static int __maybe_unused skd_pci_resume (struct device *dev) {int i;int rc =0 ;struct skd_device *skdev; +struct pci_dev *pdev =to_pci_dev (dev); skdev =pci_get_drvdata (pdev);if (!skdev) {
@@ -3356,16 +3354,8 @@ static int skd_pci_resume(struct pci_dev *pdev)return -1 ; } -pci_set_power_state (pdev,PCI_D0 ); -pci_enable_wake (pdev,PCI_D0 ,0 ); -pci_restore_state (pdev); +device_wakeup_disable (dev); - rc =pci_enable_device (pdev); -if (rc) -return rc; - rc =pci_request_regions (pdev,DRV_NAME ); -if (rc) - goto err_out; rc =dma_set_mask_and_coherent (&pdev->dev,DMA_BIT_MASK (64 ));if (rc) rc =dma_set_mask_and_coherent (&pdev->dev,DMA_BIT_MASK (32 ));
@@ -3374,7 +3364,6 @@ static int skd_pci_resume(struct pci_dev *pdev) goto err_out_regions; } -pci_set_master (pdev); rc =pci_enable_pcie_error_reporting (pdev);if (rc) { dev_err(&pdev->dev,
@@ -3427,10 +3416,6 @@ static int skd_pci_resume(struct pci_dev *pdev)pci_disable_pcie_error_reporting (pdev); err_out_regions: -pci_release_regions (pdev); - -err_out: -pci_disable_device (pdev);return rc; }
@@ -3450,13 +3435,14 @@ static void skd_pci_shutdown(struct pci_dev *pdev)skd_stop_device (skdev); } +static SIMPLE_DEV_PM_OPS(skd_pci_pm_ops, skd_pci_suspend, skd_pci_resume); +static struct pci_driver skd_driver = { .name =DRV_NAME , .id_table = skd_pci_tbl, .probe = skd_pci_probe, .remove = skd_pci_remove, - .suspend = skd_pci_suspend, - .resume = skd_pci_resume, + .driver.pm = &skd_pci_pm_ops, .shutdown = skd_pci_shutdown, };
Signature · 1 line
2.27.0
Re: [PATCH v2 3/3] skd: use generic power management
Quoted 6 lines · Vaibhav Gupta
Drivers using legacy PM have to manage PCI states and device's PM states themselves. They also need to take care of configuration registers.
With improved and powerful support of generic PM, PCI Core takes care of above mentioned, device-independent, jobs.
Quoted 8 lines
This driver makes use of PCI helper functions like pci_save/restore_state(), pci_enable/disable_device(), pci_request/release_regions(),
pci_set_power_state()andpci_set_master()to do required operations. In generic mode, they are no longer needed.Change function parameter in both .suspend() and .resume() to "struct device*" type. Use
to_pci_dev()to get "struct pci_dev*" variable.
This commit message is rather vague, and the last sentence actually does not describe correctly the change. What about something very simple, yet clear, like this:
skd: use generic power management
Switch from the legacy .suspend()/.resume() power management interface to the generic power management interface using the single .driver.pm() method. This avoids the need for the driver to directly call most of the PCI helper functions and device power state control functions as the generic power management interface takes care of the necessary operations.
Quoted 91 lines
Compile-tested only.
Signed-off-by Vaibhav Gupta <[email protected]>--- drivers/block/skd_main.c | 30 ++++++++---------------------- 1 file changed, 8 insertions(+), 22 deletions(-)drivers/block/skd_main.c +7
-20@@ -3315,10 +3315,11 @@ static void skd_pci_remove(struct pci_dev *pdev)return ; } -static int skd_pci_suspend (struct pci_dev *pdev,pm_message_t state) +static int __maybe_unused skd_pci_suspend (struct device *dev) {int i;struct skd_device *skdev; +struct pci_dev *pdev =to_pci_dev (dev); skdev =pci_get_drvdata (pdev);if (!skdev) {@@ -3337,18 +3338,15 @@ static int skd_pci_suspend(struct pci_dev *pdev, pm_message_t state)if (skdev->pcie_error_reporting_is_enabled)pci_disable_pcie_error_reporting (pdev); -pci_release_regions (pdev); -pci_save_state (pdev); -pci_disable_device (pdev); -pci_set_power_state (pdev,pci_choose_state (pdev, state));return 0 ; } -static int skd_pci_resume (struct pci_dev *pdev) +static int __maybe_unused skd_pci_resume (struct device *dev) {int i;int rc =0 ;struct skd_device *skdev; +struct pci_dev *pdev =to_pci_dev (dev); skdev =pci_get_drvdata (pdev);if (!skdev) {@@ -3356,16 +3354,8 @@ static int skd_pci_resume(struct pci_dev *pdev)return -1 ; } -pci_set_power_state (pdev,PCI_D0 ); -pci_enable_wake (pdev,PCI_D0 ,0 ); -pci_restore_state (pdev); +device_wakeup_disable (dev); - rc =pci_enable_device (pdev); -if (rc) -return rc; - rc =pci_request_regions (pdev,DRV_NAME ); -if (rc) - goto err_out; rc =dma_set_mask_and_coherent (&pdev->dev,DMA_BIT_MASK (64 ));if (rc) rc =dma_set_mask_and_coherent (&pdev->dev,DMA_BIT_MASK (32 ));@@ -3374,7 +3364,6 @@ static int skd_pci_resume(struct pci_dev *pdev) goto err_out_regions; } -pci_set_master (pdev); rc =pci_enable_pcie_error_reporting (pdev);if (rc) { dev_err(&pdev->dev,@@ -3427,10 +3416,6 @@ static int skd_pci_resume(struct pci_dev *pdev)pci_disable_pcie_error_reporting (pdev); err_out_regions: -pci_release_regions (pdev); - -err_out: -pci_disable_device (pdev);return rc; }@@ -3450,13 +3435,14 @@ static void skd_pci_shutdown(struct pci_dev *pdev)skd_stop_device (skdev); } +static SIMPLE_DEV_PM_OPS(skd_pci_pm_ops, skd_pci_suspend, skd_pci_resume); +static struct pci_driver skd_driver = {
drivers/block/skd_main.c static void skd_pci_shutdown(struct pci_dev *pdev) · line 3448
Quoted 10 lines
drivers/block/skd_main.c +1
-2.name = DRV_NAME, .id_table = skd_pci_tbl, .probe = skd_pci_probe, .remove = skd_pci_remove, - .suspend = skd_pci_suspend, - .resume = skd_pci_resume, + .driver.pm = &skd_pci_pm_ops, .shutdown = skd_pci_shutdown, };
Apart from the commit message, this looks OK to me. I will give this a spin today on the hardware to check.
Signature · 2 lines
Damien Le Moal Western Digital Research
Re: [PATCH v2 3/3] skd: use generic power management
Quoted 20 lines · Damien Le Moal
Quoted 6 lines · Vaibhav Gupta
Drivers using legacy PM have to manage PCI states and device's PM states themselves. They also need to take care of configuration registers.
With improved and powerful support of generic PM, PCI Core takes care of above mentioned, device-independent, jobs.
Quoted 8 lines
This driver makes use of PCI helper functions like pci_save/restore_state(), pci_enable/disable_device(), pci_request/release_regions(),
pci_set_power_state()andpci_set_master()to do required operations. In generic mode, they are no longer needed.Change function parameter in both .suspend() and .resume() to "struct device*" type. Use
to_pci_dev()to get "struct pci_dev*" variable.This commit message is rather vague, and the last sentence actually does not describe correctly the change. What about something very simple, yet clear, like this:
Quoted 8 lines
skd: use generic power management
Switch from the legacy .suspend()/.resume() power management interface to the generic power management interface using the single .driver.pm() method. This avoids the need for the driver to directly call most of the PCI helper functions and device power state control functions as the generic power management interface takes care of the necessary operations.
Okay. I will improve on it. Just inform me after testing that if any other changes are required. I guess [PATCH 1/3] and [PATCH 2/3] are okay, so I will only send v3 of [PATCH 3/3] after suggested changes.
Quoted 96 lines
Quoted 96 lines
Compile-tested only.
Signed-off-by Vaibhav Gupta <[email protected]>--- drivers/block/skd_main.c | 30 ++++++++---------------------- 1 file changed, 8 insertions(+), 22 deletions(-)drivers/block/skd_main.c +7
-22@@ -3315,10 +3315,11 @@ static void skd_pci_remove(struct pci_dev *pdev)return ; } -static int skd_pci_suspend (struct pci_dev *pdev,pm_message_t state) +static int __maybe_unused skd_pci_suspend (struct device *dev) {int i;struct skd_device *skdev; +struct pci_dev *pdev =to_pci_dev (dev); skdev =pci_get_drvdata (pdev);if (!skdev) {@@ -3337,18 +3338,15 @@ static int skd_pci_suspend(struct pci_dev *pdev, pm_message_t state)if (skdev->pcie_error_reporting_is_enabled)pci_disable_pcie_error_reporting (pdev); -pci_release_regions (pdev); -pci_save_state (pdev); -pci_disable_device (pdev); -pci_set_power_state (pdev,pci_choose_state (pdev, state));return 0 ; } -static int skd_pci_resume (struct pci_dev *pdev) +static int __maybe_unused skd_pci_resume (struct device *dev) {int i;int rc =0 ;struct skd_device *skdev; +struct pci_dev *pdev =to_pci_dev (dev); skdev =pci_get_drvdata (pdev);if (!skdev) {@@ -3356,16 +3354,8 @@ static int skd_pci_resume(struct pci_dev *pdev)return -1 ; } -pci_set_power_state (pdev,PCI_D0 ); -pci_enable_wake (pdev,PCI_D0 ,0 ); -pci_restore_state (pdev); +device_wakeup_disable (dev); - rc =pci_enable_device (pdev); -if (rc) -return rc; - rc =pci_request_regions (pdev,DRV_NAME ); -if (rc) - goto err_out; rc =dma_set_mask_and_coherent (&pdev->dev,DMA_BIT_MASK (64 ));if (rc) rc =dma_set_mask_and_coherent (&pdev->dev,DMA_BIT_MASK (32 ));@@ -3374,7 +3364,6 @@ static int skd_pci_resume(struct pci_dev *pdev) goto err_out_regions; } -pci_set_master (pdev); rc =pci_enable_pcie_error_reporting (pdev);if (rc) { dev_err(&pdev->dev,@@ -3427,10 +3416,6 @@ static int skd_pci_resume(struct pci_dev *pdev)pci_disable_pcie_error_reporting (pdev); err_out_regions: -pci_release_regions (pdev); - -err_out: -pci_disable_device (pdev);return rc; }@@ -3450,13 +3435,14 @@ static void skd_pci_shutdown(struct pci_dev *pdev)skd_stop_device (skdev); } +static SIMPLE_DEV_PM_OPS(skd_pci_pm_ops, skd_pci_suspend, skd_pci_resume); +static struct pci_driver skd_driver = { .name =DRV_NAME , .id_table = skd_pci_tbl, .probe = skd_pci_probe, .remove = skd_pci_remove, - .suspend = skd_pci_suspend, - .resume = skd_pci_resume,
Quoted 9 lines
drivers/block/skd_main.c static void skd_pci_shutdown(struct pci_dev *pdev) · line 3448
Quoted 5 lines
drivers/block/skd_main.c +1
-0+ .driver.pm = &skd_pci_pm_ops, .shutdown = skd_pci_shutdown, };Apart from the commit message, this looks OK to me. I will give this a spin today on the hardware to check.
Thanks!
--Vaibhav Gupta
Quoted 4 lines
Signature · 3 lines
-- Damien Le Moal Western Digital Research
Re: [PATCH v2 3/3] skd: use generic power management
Quoted 136 lines · Vaibhav Gupta
Quoted 20 lines · Damien Le Moal
Quoted 6 lines · Vaibhav Gupta
Drivers using legacy PM have to manage PCI states and device's PM states themselves. They also need to take care of configuration registers.
With improved and powerful support of generic PM, PCI Core takes care of above mentioned, device-independent, jobs.
Quoted 8 lines
This driver makes use of PCI helper functions like pci_save/restore_state(), pci_enable/disable_device(), pci_request/release_regions(),
pci_set_power_state()andpci_set_master()to do required operations. In generic mode, they are no longer needed.Change function parameter in both .suspend() and .resume() to "struct device*" type. Use
to_pci_dev()to get "struct pci_dev*" variable.This commit message is rather vague, and the last sentence actually does not describe correctly the change. What about something very simple, yet clear, like this:
Quoted 8 lines
skd: use generic power management
Switch from the legacy .suspend()/.resume() power management interface to the generic power management interface using the single .driver.pm() method. This avoids the need for the driver to directly call most of the PCI helper functions and device power state control functions as the generic power management interface takes care of the necessary operations.
Okay. I will improve on it. Just inform me after testing that if any other changes are required. I guess [PATCH 1/3] and [PATCH 2/3] are okay, so I will only send v3 of [PATCH 3/3] after suggested changes.
Quoted 95 lines
Quoted 95 lines
Compile-tested only.
Signed-off-by Vaibhav Gupta <[email protected]>--- drivers/block/skd_main.c | 30 ++++++++---------------------- 1 file changed, 8 insertions(+), 22 deletions(-)drivers/block/skd_main.c +7
-22@@ -3315,10 +3315,11 @@ static void skd_pci_remove(struct pci_dev *pdev)return ; } -static int skd_pci_suspend (struct pci_dev *pdev,pm_message_t state) +static int __maybe_unused skd_pci_suspend (struct device *dev) {int i;struct skd_device *skdev; +struct pci_dev *pdev =to_pci_dev (dev); skdev =pci_get_drvdata (pdev);if (!skdev) {@@ -3337,18 +3338,15 @@ static int skd_pci_suspend(struct pci_dev *pdev, pm_message_t state)if (skdev->pcie_error_reporting_is_enabled)pci_disable_pcie_error_reporting (pdev); -pci_release_regions (pdev); -pci_save_state (pdev); -pci_disable_device (pdev); -pci_set_power_state (pdev,pci_choose_state (pdev, state));return 0 ; } -static int skd_pci_resume (struct pci_dev *pdev) +static int __maybe_unused skd_pci_resume (struct device *dev) {int i;int rc =0 ;struct skd_device *skdev; +struct pci_dev *pdev =to_pci_dev (dev); skdev =pci_get_drvdata (pdev);if (!skdev) {@@ -3356,16 +3354,8 @@ static int skd_pci_resume(struct pci_dev *pdev)return -1 ; } -pci_set_power_state (pdev,PCI_D0 ); -pci_enable_wake (pdev,PCI_D0 ,0 ); -pci_restore_state (pdev); +device_wakeup_disable (dev); - rc =pci_enable_device (pdev); -if (rc) -return rc; - rc =pci_request_regions (pdev,DRV_NAME ); -if (rc) - goto err_out; rc =dma_set_mask_and_coherent (&pdev->dev,DMA_BIT_MASK (64 ));if (rc) rc =dma_set_mask_and_coherent (&pdev->dev,DMA_BIT_MASK (32 ));@@ -3374,7 +3364,6 @@ static int skd_pci_resume(struct pci_dev *pdev) goto err_out_regions; } -pci_set_master (pdev); rc =pci_enable_pcie_error_reporting (pdev);if (rc) { dev_err(&pdev->dev,@@ -3427,10 +3416,6 @@ static int skd_pci_resume(struct pci_dev *pdev)pci_disable_pcie_error_reporting (pdev); err_out_regions: -pci_release_regions (pdev); - -err_out: -pci_disable_device (pdev);return rc; }@@ -3450,13 +3435,14 @@ static void skd_pci_shutdown(struct pci_dev *pdev)skd_stop_device (skdev); } +static SIMPLE_DEV_PM_OPS(skd_pci_pm_ops, skd_pci_suspend, skd_pci_resume); +static struct pci_driver skd_driver = { .name =DRV_NAME , .id_table = skd_pci_tbl, .probe = skd_pci_probe, .remove = skd_pci_remove, - .suspend = skd_pci_suspend, - .resume = skd_pci_resume,Quoted 9 lines
drivers/block/skd_main.c static void skd_pci_shutdown(struct pci_dev *pdev) · line 3448
Quoted 5 lines
drivers/block/skd_main.c +1
-0+ .driver.pm = &skd_pci_pm_ops, .shutdown = skd_pci_shutdown, };Apart from the commit message, this looks OK to me. I will give this a spin today on the hardware to check.
Quoted 1 line
Thanks!
Tested this and it seems OK.
Of note, after having removed the device (echo 1 > /sys/block/skd0/device/remove) and rescaning its PCI slot, I got a lockdep splat:
[ 4449.098757] ======================================================
[ 4449.104945] WARNING: possible circular locking dependency detected
[ 4449.111136] 5.8.0-rc6+ #846 Not tainted
[ 4449.114980] ------------------------------------------------------
[ 4449.121168] tcsh/1550 is trying to acquire lock:
[ 4449.125798] ffffffffa85c3748 (pci_rescan_remove_lock){+.+.}-{3:3},
at: dev_rescan_store+0x38/0x60
[ 4449.134686]
[ 4449.134686] but task is already holding lock:
[ 4449.140522] ffff8f6cd352ed28 (kn->active#294){++++}-{0:0}, at:
kernfs_fop_write+0xad/0x1c0
[ 4449.148795]
[ 4449.148795] which lock already depends on the new lock.
[ 4449.148795]
[ 4449.156979]
[ 4449.156979] the existing dependency chain (in reverse order) is:
[ 4449.164464]
[ 4449.164464] -> #1 (kn->active#294){++++}-{0:0}:
[ 4449.170486] __kernfs_remove+0x276/0x2f0
[ 4449.174939] kernfs_remove_by_name_ns+0x41/0x80
[ 4449.180001] remove_files+0x2b/0x60
[ 4449.184018] sysfs_remove_group+0x38/0x80
[ 4449.188560] sysfs_remove_groups+0x29/0x40
[ 4449.193190] device_remove_attrs+0x4b/0x70
[ 4449.197818] device_del+0x167/0x3f0
[ 4449.201843] pci_remove_bus_device+0x70/0x110
[ 4449.206728] pci_stop_and_remove_bus_device_locked+0x1e/0x30
[ 4449.212915] remove_store+0x55/0x60
[ 4449.216939] kernfs_fop_write+0xd9/0x1c0
[ 4449.221398] vfs_write+0xc7/0x1f0
[ 4449.225239] ksys_write+0x58/0xd0
[ 4449.229089] do_syscall_64+0x4f/0x2c0
[ 4449.233284] entry_SYSCALL_64_after_hwframe+0x44/0xa9
[ 4449.238863]
[ 4449.238863] -> #0 (pci_rescan_remove_lock){+.+.}-{3:3}:
[ 4449.245575] __lock_acquire+0x1194/0x1fd0
[ 4449.250113] lock_acquire+0x97/0x390
[ 4449.254221] __mutex_lock+0x58/0x820
[ 4449.258330] dev_rescan_store+0x38/0x60
[ 4449.262697] kernfs_fop_write+0xd9/0x1c0
[ 4449.267154] vfs_write+0xc7/0x1f0
[ 4449.270999] ksys_write+0x58/0xd0
[ 4449.274846] do_syscall_64+0x4f/0x2c0
[ 4449.279041] entry_SYSCALL_64_after_hwframe+0x44/0xa9
[ 4449.284620]
[ 4449.284620] other info that might help us debug this:
[ 4449.284620]
[ 4449.292628] Possible unsafe locking scenario:
[ 4449.292628]
[ 4449.298555] CPU0 CPU1
[ 4449.303090] ---- ----
[ 4449.307630] lock(kn->active#294);
[
4449.311131] lock(pci_rescan_remove_lock
);
[ 4449.317838] lock(kn->active#294);
[ 4449.323855] lock(pci_rescan_remove_lock);
[ 4449.328048]
[ 4449.328048] *** DEADLOCK ***
[ 4449.328048]
[ 4449.333977] 3 locks held by tcsh/1550:
[ 4449.337731] #0: ffff8f6f18b7d448 (sb_writers#4){.+.+}-{0:0}, at:
vfs_write+0x174/0x1f0
[ 4449.345747] #1: ffff8f6e74d09a88 (&of->mutex){+.+.}-{3:3}, at:
kernfs_fop_write+0xa5/0x1c0
[ 4449.354111] #2: ffff8f6cd352ed28 (kn->active#294){++++}-{0:0}, at:
kernfs_fop_write+0xad/0x1c0
[ 4449.362821]
[ 4449.362821] stack backtrace:
[ 4449.367192] CPU: 15 PID: 1550 Comm: tcsh Not tainted 5.8.0-rc6+ #846
[ 4449.373548] Hardware name: Supermicro Super Server/X11DPL-i, BIOS
3.1 05/21/2019
[ 4449.380953] Call Trace:
[ 4449.383419] dump_stack+0x78/0xa0
[ 4449.386744] check_noncircular+0x12d/0x150
[ 4449.390853] __lock_acquire+0x1194/0x1fd0
[ 4449.394875] lock_acquire+0x97/0x390
[ 4449.398462] ? dev_rescan_store+0x38/0x60
[ 4449.402486] __mutex_lock+0x58/0x820
[ 4449.406067] ? dev_rescan_store+0x38/0x60
[ 4449.410090] ? dev_rescan_store+0x38/0x60
[ 4449.414113] ? lock_acquire+0x97/0x390
[ 4449.417876] ? kernfs_fop_write+0xad/0x1c0
[ 4449.421983] dev_rescan_store+0x38/0x60
[ 4449.425831] kernfs_fop_write+0xd9/0x1c0
[ 4449.429768] vfs_write+0xc7/0x1f0
[ 4449.433092] ksys_write+0x58/0xd0
[ 4449.436421] do_syscall_64+0x4f/0x2c0
[ 4449.440094] ? prepare_exit_to_usermode+0xa/0x40
[ 4449.444725] ? rcu_read_lock_sched_held+0x3f/0x50
[ 4449.449436] ? asm_exc_page_fault+0x8/0x30
[ 4449.453546] entry_SYSCALL_64_after_hwframe+0x44/0xa9
[ 4449.458608] RIP: 0033:0x7fcce58f45e7
[ 4449.462191] Code: Bad RIP value.
[ 4449.465426] RSP: 002b:00007fff2592c898 EFLAGS: 00000246 ORIG_RAX:
0000000000000001
[ 4449.473000] RAX: ffffffffffffffda RBX: 0000000000000002 RCX:
00007fcce58f45e7
[ 4449.480141] RDX: 0000000000000002 RSI: 0000556d30b96100 RDI:
0000000000000001
[ 4449.487283] RBP: 0000556d30b96100 R08: 0000000000000000 R09:
00007fff2592c7f0
[ 4449.494425] R10: 0000000000000000 R11: 0000000000000246 R12:
0000000000000001
[ 4449.501567] R13: 0000556d31fc4fb8 R14: 0000556d31fc37c0 R15:
0000556d31fc4fb4
[ 4449.520721] pci 0000:d8:00.0: [1b39:0001] type 00 class 0x018000
[ 4449.526802] pci 0000:d8:00.0: reg 0x10: [mem 0xfbe00000-0xfbe0ffff]
[ 4449.533121] pci 0000:d8:00.0: reg 0x14: [mem 0xfbe10000-0xfbe10fff]
[ 4449.539474] pci 0000:d8:00.0: reg 0x30: [mem 0xfbd00000-0xfbdfffff
pref]
[ 4449.546327] pci 0000:d8:00.0: supports D1 D2
[ 4449.550936] pci 0000:d8:00.0: BAR 6: assigned [mem 0xfbd00000-
0xfbdfffff pref]
[ 4449.559804] pci 0000:d8:00.0: BAR 0: assigned [mem 0xfbe00000-
0xfbe0ffff]
[ 4449.568849] pci 0000:d8:00.0: BAR 1: assigned [mem 0xfbe10000-
0xfbe10fff]
[ 4449.577178] skd 0000:d8:00.0: PCIe (5.0GT/s 4X) 64bit
[ 4449.582272] skd 0000:d8:00.0: bad enable of PCIe error reporting
rc=-5
[ 4449.589764] skd 0000:d8:00.0: s1120 state INIT(1)=>SOFT_RESET(8)
[ 4449.595788] skd 0000:d8:00.0: Driver state STARTING(3)=>STARTING(3)
[ 4449.829900] skd 0000:d8:00.0: s1120 state SOFT_RESET(8)=>INIT(1)
[ 4449.835929] skd 0000:d8:00.0: Driver state STARTING(3)=>STARTING(3)
[ 4450.076148] skd 0000:d8:00.0: Time sync driver=0x5f17d872
device=0x162bdfe9
[ 4450.083139] skd 0000:d8:00.0: s1120 state INIT(1)=>ONLINE(3)
[ 4450.088822] skd 0000:d8:00.0: Queue depth limit=64 dev=255 lowat=43
[ 4450.095101] skd 0000:d8:00.0: Driver state STARTING(3)=>STARTING(3)
[ 4450.101539] skd 0000:d8:00.0: Driver state STARTING(3)=>ONLINE(4)
[ 4450.107653] skd 0000:d8:00.0: STEC s1120 ONLINE
[ 4450.120422] skd1: p1But that looks completely unrelated to your patch. Will have a look at this.
Signature · 2 lines
Damien Le Moal Western Digital Research
Re: [PATCH v2 3/3] skd: use generic power management
Quoted 279 lines · Damien Le Moal
Quoted 136 lines · Vaibhav Gupta
Quoted 20 lines · Damien Le Moal
Quoted 6 lines · Vaibhav Gupta
Drivers using legacy PM have to manage PCI states and device's PM states themselves. They also need to take care of configuration registers.
With improved and powerful support of generic PM, PCI Core takes care of above mentioned, device-independent, jobs.
Quoted 8 lines
This driver makes use of PCI helper functions like pci_save/restore_state(), pci_enable/disable_device(), pci_request/release_regions(),
pci_set_power_state()andpci_set_master()to do required operations. In generic mode, they are no longer needed.Change function parameter in both .suspend() and .resume() to "struct device*" type. Use
to_pci_dev()to get "struct pci_dev*" variable.This commit message is rather vague, and the last sentence actually does not describe correctly the change. What about something very simple, yet clear, like this:
Quoted 8 lines
skd: use generic power management
Switch from the legacy .suspend()/.resume() power management interface to the generic power management interface using the single .driver.pm() method. This avoids the need for the driver to directly call most of the PCI helper functions and device power state control functions as the generic power management interface takes care of the necessary operations.
Okay. I will improve on it. Just inform me after testing that if any other changes are required. I guess [PATCH 1/3] and [PATCH 2/3] are okay, so I will only send v3 of [PATCH 3/3] after suggested changes.
Quoted 95 lines
Quoted 95 lines
Compile-tested only.
Signed-off-by Vaibhav Gupta <[email protected]>--- drivers/block/skd_main.c | 30 ++++++++---------------------- 1 file changed, 8 insertions(+), 22 deletions(-)drivers/block/skd_main.c +7
-22@@ -3315,10 +3315,11 @@ static void skd_pci_remove(struct pci_dev *pdev)return ; } -static int skd_pci_suspend (struct pci_dev *pdev,pm_message_t state) +static int __maybe_unused skd_pci_suspend (struct device *dev) {int i;struct skd_device *skdev; +struct pci_dev *pdev =to_pci_dev (dev); skdev =pci_get_drvdata (pdev);if (!skdev) {@@ -3337,18 +3338,15 @@ static int skd_pci_suspend(struct pci_dev *pdev, pm_message_t state)if (skdev->pcie_error_reporting_is_enabled)pci_disable_pcie_error_reporting (pdev); -pci_release_regions (pdev); -pci_save_state (pdev); -pci_disable_device (pdev); -pci_set_power_state (pdev,pci_choose_state (pdev, state));return 0 ; } -static int skd_pci_resume (struct pci_dev *pdev) +static int __maybe_unused skd_pci_resume (struct device *dev) {int i;int rc =0 ;struct skd_device *skdev; +struct pci_dev *pdev =to_pci_dev (dev); skdev =pci_get_drvdata (pdev);if (!skdev) {@@ -3356,16 +3354,8 @@ static int skd_pci_resume(struct pci_dev *pdev)return -1 ; } -pci_set_power_state (pdev,PCI_D0 ); -pci_enable_wake (pdev,PCI_D0 ,0 ); -pci_restore_state (pdev); +device_wakeup_disable (dev); - rc =pci_enable_device (pdev); -if (rc) -return rc; - rc =pci_request_regions (pdev,DRV_NAME ); -if (rc) - goto err_out; rc =dma_set_mask_and_coherent (&pdev->dev,DMA_BIT_MASK (64 ));if (rc) rc =dma_set_mask_and_coherent (&pdev->dev,DMA_BIT_MASK (32 ));@@ -3374,7 +3364,6 @@ static int skd_pci_resume(struct pci_dev *pdev) goto err_out_regions; } -pci_set_master (pdev); rc =pci_enable_pcie_error_reporting (pdev);if (rc) { dev_err(&pdev->dev,@@ -3427,10 +3416,6 @@ static int skd_pci_resume(struct pci_dev *pdev)pci_disable_pcie_error_reporting (pdev); err_out_regions: -pci_release_regions (pdev); - -err_out: -pci_disable_device (pdev);return rc; }@@ -3450,13 +3435,14 @@ static void skd_pci_shutdown(struct pci_dev *pdev)skd_stop_device (skdev); } +static SIMPLE_DEV_PM_OPS(skd_pci_pm_ops, skd_pci_suspend, skd_pci_resume); +static struct pci_driver skd_driver = { .name =DRV_NAME , .id_table = skd_pci_tbl, .probe = skd_pci_probe, .remove = skd_pci_remove, - .suspend = skd_pci_suspend, - .resume = skd_pci_resume,Quoted 9 lines
drivers/block/skd_main.c static void skd_pci_shutdown(struct pci_dev *pdev) · line 3448
Quoted 5 lines
drivers/block/skd_main.c +1
-0+ .driver.pm = &skd_pci_pm_ops, .shutdown = skd_pci_shutdown, };Apart from the commit message, this looks OK to me. I will give this a spin today on the hardware to check.
Quoted 1 line
Thanks!
Tested this and it seems OK.
Of note, after having removed the device (echo 1 >
/sys/block/skd0/device/remove) and rescaning its PCI slot, I got a lockdep splat:[ 4449.098757] ====================================================== [ 4449.104945] WARNING: possible circular locking dependency detected [ 4449.111136] 5.8.0-rc6+ #846 Not tainted [ 4449.114980] ------------------------------------------------------ [ 4449.121168] tcsh/1550 is trying to acquire lock: [ 4449.125798] ffffffffa85c3748 (pci_rescan_remove_lock){+.+.}-{3:3}, at: dev_rescan_store+0x38/0x60 [ 4449.134686] [ 4449.134686] but task is already holding lock: [ 4449.140522] ffff8f6cd352ed28 (kn->active#294){++++}-{0:0}, at: kernfs_fop_write+0xad/0x1c0 [ 4449.148795] [ 4449.148795] which lock already depends on the new lock. [ 4449.148795] [ 4449.156979] [ 4449.156979] the existing dependency chain (in reverse order) is: [ 4449.164464] [ 4449.164464] -> #1 (kn->active#294){++++}-{0:0}: [ 4449.170486] __kernfs_remove+0x276/0x2f0 [ 4449.174939] kernfs_remove_by_name_ns+0x41/0x80 [ 4449.180001] remove_files+0x2b/0x60 [ 4449.184018] sysfs_remove_group+0x38/0x80 [ 4449.188560] sysfs_remove_groups+0x29/0x40 [ 4449.193190] device_remove_attrs+0x4b/0x70 [ 4449.197818] device_del+0x167/0x3f0 [ 4449.201843] pci_remove_bus_device+0x70/0x110 [ 4449.206728] pci_stop_and_remove_bus_device_locked+0x1e/0x30 [ 4449.212915] remove_store+0x55/0x60 [ 4449.216939] kernfs_fop_write+0xd9/0x1c0 [ 4449.221398] vfs_write+0xc7/0x1f0 [ 4449.225239] ksys_write+0x58/0xd0 [ 4449.229089] do_syscall_64+0x4f/0x2c0 [ 4449.233284] entry_SYSCALL_64_after_hwframe+0x44/0xa9 [ 4449.238863] [ 4449.238863] -> #0 (pci_rescan_remove_lock){+.+.}-{3:3}: [ 4449.245575] __lock_acquire+0x1194/0x1fd0 [ 4449.250113] lock_acquire+0x97/0x390 [ 4449.254221] __mutex_lock+0x58/0x820 [ 4449.258330] dev_rescan_store+0x38/0x60 [ 4449.262697] kernfs_fop_write+0xd9/0x1c0 [ 4449.267154] vfs_write+0xc7/0x1f0 [ 4449.270999] ksys_write+0x58/0xd0 [ 4449.274846] do_syscall_64+0x4f/0x2c0 [ 4449.279041] entry_SYSCALL_64_after_hwframe+0x44/0xa9 [ 4449.284620] [ 4449.284620] other info that might help us debug this: [ 4449.284620] [ 4449.292628] Possible unsafe locking scenario: [ 4449.292628] [ 4449.298555] CPU0 CPU1 [ 4449.303090] ---- ---- [ 4449.307630] lock(kn->active#294); [ 4449.311131] lock(pci_rescan_remove_lock ); [ 4449.317838] lock(kn->active#294); [ 4449.323855] lock(pci_rescan_remove_lock); [ 4449.328048] [ 4449.328048] *** DEADLOCK *** [ 4449.328048] [ 4449.333977] 3 locks held by tcsh/1550: [ 4449.337731] #0: ffff8f6f18b7d448 (sb_writers#4){.+.+}-{0:0}, at: vfs_write+0x174/0x1f0 [ 4449.345747] #1: ffff8f6e74d09a88 (&of->mutex){+.+.}-{3:3}, at: kernfs_fop_write+0xa5/0x1c0 [ 4449.354111] #2: ffff8f6cd352ed28 (kn->active#294){++++}-{0:0}, at: kernfs_fop_write+0xad/0x1c0 [ 4449.362821] [ 4449.362821] stack backtrace: [ 4449.367192] CPU: 15 PID: 1550 Comm: tcsh Not tainted 5.8.0-rc6+ #846 [ 4449.373548] Hardware name: Supermicro Super Server/X11DPL-i, BIOS 3.1 05/21/2019 [ 4449.380953] Call Trace: [ 4449.383419] dump_stack+0x78/0xa0 [ 4449.386744] check_noncircular+0x12d/0x150 [ 4449.390853] __lock_acquire+0x1194/0x1fd0 [ 4449.394875] lock_acquire+0x97/0x390 [ 4449.398462] ? dev_rescan_store+0x38/0x60 [ 4449.402486] __mutex_lock+0x58/0x820 [ 4449.406067] ? dev_rescan_store+0x38/0x60 [ 4449.410090] ? dev_rescan_store+0x38/0x60 [ 4449.414113] ? lock_acquire+0x97/0x390 [ 4449.417876] ? kernfs_fop_write+0xad/0x1c0 [ 4449.421983] dev_rescan_store+0x38/0x60 [ 4449.425831] kernfs_fop_write+0xd9/0x1c0 [ 4449.429768] vfs_write+0xc7/0x1f0 [ 4449.433092] ksys_write+0x58/0xd0 [ 4449.436421] do_syscall_64+0x4f/0x2c0 [ 4449.440094] ? prepare_exit_to_usermode+0xa/0x40 [ 4449.444725] ? rcu_read_lock_sched_held+0x3f/0x50 [ 4449.449436] ? asm_exc_page_fault+0x8/0x30 [ 4449.453546] entry_SYSCALL_64_after_hwframe+0x44/0xa9 [ 4449.458608] RIP: 0033:0x7fcce58f45e7 [ 4449.462191] Code: Bad RIP value. [ 4449.465426] RSP: 002b:00007fff2592c898 EFLAGS: 00000246 ORIG_RAX: 0000000000000001 [ 4449.473000] RAX: ffffffffffffffda RBX: 0000000000000002 RCX: 00007fcce58f45e7 [ 4449.480141] RDX: 0000000000000002 RSI: 0000556d30b96100 RDI: 0000000000000001 [ 4449.487283] RBP: 0000556d30b96100 R08: 0000000000000000 R09: 00007fff2592c7f0 [ 4449.494425] R10: 0000000000000000 R11: 0000000000000246 R12: 0000000000000001 [ 4449.501567] R13: 0000556d31fc4fb8 R14: 0000556d31fc37c0 R15: 0000556d31fc4fb4 [ 4449.520721] pci 0000:d8:00.0: [1b39:0001] type 00 class 0x018000 [ 4449.526802] pci 0000:d8:00.0: reg 0x10: [mem 0xfbe00000-0xfbe0ffff] [ 4449.533121] pci 0000:d8:00.0: reg 0x14: [mem 0xfbe10000-0xfbe10fff] [ 4449.539474] pci 0000:d8:00.0: reg 0x30: [mem 0xfbd00000-0xfbdfffff pref] [ 4449.546327] pci 0000:d8:00.0: supports D1 D2 [ 4449.550936] pci 0000:d8:00.0: BAR 6: assigned [mem 0xfbd00000- 0xfbdfffff pref] [ 4449.559804] pci 0000:d8:00.0: BAR 0: assigned [mem 0xfbe00000- 0xfbe0ffff] [ 4449.568849] pci 0000:d8:00.0: BAR 1: assigned [mem 0xfbe10000- 0xfbe10fff] [ 4449.577178] skd 0000:d8:00.0: PCIe (5.0GT/s 4X) 64bit [ 4449.582272] skd 0000:d8:00.0: bad enable of PCIe error reporting rc=-5 [ 4449.589764] skd 0000:d8:00.0: s1120 state INIT(1)=>SOFT_RESET(8) [ 4449.595788] skd 0000:d8:00.0: Driver state STARTING(3)=>STARTING(3) [ 4449.829900] skd 0000:d8:00.0: s1120 state SOFT_RESET(8)=>INIT(1) [ 4449.835929] skd 0000:d8:00.0: Driver state STARTING(3)=>STARTING(3) [ 4450.076148] skd 0000:d8:00.0: Time sync driver=0x5f17d872 device=0x162bdfe9 [ 4450.083139] skd 0000:d8:00.0: s1120 state INIT(1)=>ONLINE(3) [ 4450.088822] skd 0000:d8:00.0: Queue depth limit=64 dev=255 lowat=43 [ 4450.095101] skd 0000:d8:00.0: Driver state STARTING(3)=>STARTING(3) [ 4450.101539] skd 0000:d8:00.0: Driver state STARTING(3)=>ONLINE(4) [ 4450.107653] skd 0000:d8:00.0: STEC s1120 ONLINE [ 4450.120422] skd1: p1
Quoted 3 lines
But that looks completely unrelated to your patch. Will have a look at this.
Thanks. I will send the patch-series as v3 with updated commit message as suggested.
--Vaibhav Gupta
Quoted 3 lines
Signature · 2 lines
Damien Le Moal Western Digital Research
v3 cover block: use generic power management
In reply to Damien Le MoalLinux Kernel Mentee: Remove Legacy Power Management.
The purpose of this patch series is to upgrade power management in block drivers. This has been done by upgrading .suspend() and .resume() callbacks.
The upgrade makes sure that the involvement of PCI Core does not change the order of operations executed in a driver. Thus, does not change its behavior.
In general, drivers with legacy PM, .suspend() and .resume() make use of PCI helper functions like pci_enable/disable_device_mem(), pci_set_power_state(), pci_save/restore_state(), pci_enable/disable_device(), etc. to complete their job.
The conversion requires the removal of those function calls, change the callbacks' definition accordingly and make use of dev_pm_ops structure.
All patches are compile-tested only.
Test tools:
- Compiler: gcc (GCC) 10.1.0
- allmodconfig build: make -j$(nproc) W=1 all
v3: v2 needed some changes in commit messages.
Vaibhav Gupta (3):
mtip32xx: use generic power management
rsxx: use generic power management
skd: use generic power management
drivers/block/mtip32xx/mtip32xx.c | 54 +++++++------------------------ drivers/block/rsxx/core.c | 9 ++++-- drivers/block/skd_main.c | 30 +++++------------ 3 files changed, 27 insertions(+), 66 deletions(-)
Signature · 1 line
2.27.0
PATCH v3 1/3 mtip32xx: use generic power management
Drivers using legacy power management .suspen()/.resume() callbacks have to manage PCI states and device's PM states themselves. They also need to take care of standard configuration registers.
Switch to generic power management framework using a single "struct dev_pm_ops" variable to take the unnecessary load from the driver. This also avoids the need for the driver to directly call most of the PCI helper functions and device power state control functions, as through the generic framework PCI Core takes care of the necessary operations, and drivers are required to do only device-specific jobs.
--- drivers/block/mtip32xx/mtip32xx.c | 54 +++++++------------------------ 1 file changed, 12 insertions(+), 42 deletions(-)
drivers/block/mtip32xx/mtip32xx.c +12 -42
@@ -4246,14 +4246,13 @@ static void mtip_pci_remove(struct pci_dev *pdev)* 0 Success * <0 Error */ -static int mtip_pci_suspend (struct pci_dev *pdev,pm_message_t mesg) +static int __maybe_unused mtip_pci_suspend (struct device *dev) {int rv =0 ; -struct driver_data *dd =pci_get_drvdata (pdev); +struct driver_data *dd =dev_get_drvdata (dev);if (!dd) { -dev_err (&pdev->dev, -"Driver private datastructure is NULL\n" ); +dev_err (dev,"Driver private datastructure is NULL\n" );return -EFAULT ; }
@@ -4261,21 +4260,8 @@ static int mtip_pci_suspend(struct pci_dev *pdev, pm_message_t mesg)/* Disable ports & interrupts then send standby immediate */ rv =mtip_block_suspend (dd); -if (rv <0 ) { -dev_err (&pdev->dev, -"Failed to suspend controller\n" ); -return rv; - } - -/* -* Save the pci config space to pdev structure & -* disable the device -*/ -pci_save_state (pdev); -pci_disable_device (pdev); - -/* Move to Low power state*/ -pci_set_power_state (pdev, PCI_D3hot); +if (rv <0 ) +dev_err (dev,"Failed to suspend controller\n" );return rv; }
@@ -4287,42 +4273,25 @@ static int mtip_pci_suspend(struct pci_dev *pdev, pm_message_t mesg)* 0 Success * <0 Error */ -static int mtip_pci_resume (struct pci_dev *pdev) +static int __maybe_unused mtip_pci_resume (struct device *dev) {int rv =0 ;struct driver_data *dd; - dd =pci_get_drvdata (pdev); + dd =dev_get_drvdata (dev);if (!dd) { -dev_err (&pdev->dev, -"Driver private datastructure is NULL\n" ); +dev_err (dev,"Driver private datastructure is NULL\n" );return -EFAULT ; } -/* Move the device to active State */ -pci_set_power_state (pdev,PCI_D0 ); - -/* Restore PCI configuration space */ -pci_restore_state (pdev); - -/* Enable the PCI device*/ - rv =pcim_enable_device (pdev); -if (rv <0 ) { -dev_err (&pdev->dev, -"Failed to enable card during resume\n" ); - goto err; - } -pci_set_master (pdev); -/* * Calls hbaReset, initPort, & startPort function * then enables interrupts */ rv =mtip_block_resume (dd);if (rv <0 ) -dev_err (&pdev->dev,"Unable to resume\n" ); +dev_err (dev,"Unable to resume\n" ); -err:clear_bit (MTIP_DDF_RESUME_BIT , &dd->dd_flag);return rv;
@@ -4353,14 +4322,15 @@ static const struct pci_device_id mtip_pci_tbl[] = { { 0 } }; +static SIMPLE_DEV_PM_OPS(mtip_pci_pm_ops, mtip_pci_suspend, mtip_pci_resume); +/* Structure that describes the PCI driver functions. */ static struct pci_driver mtip_pci_driver = { .name =MTIP_DRV_NAME , .id_table = mtip_pci_tbl, .probe = mtip_pci_probe, .remove = mtip_pci_remove, - .suspend = mtip_pci_suspend, - .resume = mtip_pci_resume, + .driver.pm = &mtip_pci_pm_ops, .shutdown = mtip_pci_shutdown, };
Signature · 1 line
2.27.0
PATCH v3 2/3 rsxx: use generic power management
In reply to Vaibhav GuptaDrivers using legacy power management .suspen()/.resume() callbacks have to manage PCI states and device's PM states themselves. They also need to take care of standard configuration registers.
Switch to generic power management framework using a single "struct dev_pm_ops" variable to take the unnecessary load from the driver. This also avoids the need for the driver to directly call most of the PCI helper functions and device power state control functions, as through the generic framework PCI Core takes care of the necessary operations, and drivers are required to do only device-specific jobs.
--- drivers/block/rsxx/core.c | 9 +++++++-- 1 file changed, 7 insertions(+), 2 deletions(-)
drivers/block/rsxx/core.c +7 -2
@@ -1030,12 +1030,15 @@ static void rsxx_pci_remove(struct pci_dev *dev)kfree (card); } -static int rsxx_pci_suspend (struct pci_dev *dev,pm_message_t state) +static int __maybe_unused rsxx_pci_suspend ( + __attribute__((unused))struct device *dev) {/* We don't support suspend at this time. */ return -ENOSYS ; } +#define rsxx_pci_resume NULL +static void rsxx_pci_shutdown (struct pci_dev *dev) {struct rsxx_cardinfo *card =pci_get_drvdata (dev);
@@ -1071,12 +1074,14 @@ static const struct pci_device_id rsxx_pci_ids[] = {MODULE_DEVICE_TABLE (pci, rsxx_pci_ids); +static SIMPLE_DEV_PM_OPS(rsxx_pci_pm_ops, rsxx_pci_suspend, rsxx_pci_resume); +static struct pci_driver rsxx_pci_driver = { .name =DRIVER_NAME , .id_table = rsxx_pci_ids, .probe = rsxx_pci_probe, .remove = rsxx_pci_remove, - .suspend = rsxx_pci_suspend, + .driver.pm = &rsxx_pci_pm_ops, .shutdown = rsxx_pci_shutdown, .err_handler = &rsxx_err_handler, };
Signature · 1 line
2.27.0
PATCH v3 3/3 skd: use generic power management
In reply to Vaibhav GuptaDrivers using legacy power management .suspen()/.resume() callbacks have to manage PCI states and device's PM states themselves. They also need to take care of standard configuration registers.
Switch to generic power management framework using a single "struct dev_pm_ops" variable to take the unnecessary load from the driver. This also avoids the need for the driver to directly call most of the PCI helper functions and device power state control functions, as through the generic framework PCI Core takes care of the necessary operations, and drivers are required to do only device-specific jobs.
--- drivers/block/skd_main.c | 30 ++++++++---------------------- 1 file changed, 8 insertions(+), 22 deletions(-)
drivers/block/skd_main.c +8 -22
@@ -3315,10 +3315,11 @@ static void skd_pci_remove(struct pci_dev *pdev)return ; } -static int skd_pci_suspend (struct pci_dev *pdev,pm_message_t state) +static int __maybe_unused skd_pci_suspend (struct device *dev) {int i;struct skd_device *skdev; +struct pci_dev *pdev =to_pci_dev (dev); skdev =pci_get_drvdata (pdev);if (!skdev) {
@@ -3337,18 +3338,15 @@ static int skd_pci_suspend(struct pci_dev *pdev, pm_message_t state)if (skdev->pcie_error_reporting_is_enabled)pci_disable_pcie_error_reporting (pdev); -pci_release_regions (pdev); -pci_save_state (pdev); -pci_disable_device (pdev); -pci_set_power_state (pdev,pci_choose_state (pdev, state));return 0 ; } -static int skd_pci_resume (struct pci_dev *pdev) +static int __maybe_unused skd_pci_resume (struct device *dev) {int i;int rc =0 ;struct skd_device *skdev; +struct pci_dev *pdev =to_pci_dev (dev); skdev =pci_get_drvdata (pdev);if (!skdev) {
@@ -3356,16 +3354,8 @@ static int skd_pci_resume(struct pci_dev *pdev)return -1 ; } -pci_set_power_state (pdev,PCI_D0 ); -pci_enable_wake (pdev,PCI_D0 ,0 ); -pci_restore_state (pdev); +device_wakeup_disable (dev); - rc =pci_enable_device (pdev); -if (rc) -return rc; - rc =pci_request_regions (pdev,DRV_NAME ); -if (rc) - goto err_out; rc =dma_set_mask_and_coherent (&pdev->dev,DMA_BIT_MASK (64 ));if (rc) rc =dma_set_mask_and_coherent (&pdev->dev,DMA_BIT_MASK (32 ));
@@ -3374,7 +3364,6 @@ static int skd_pci_resume(struct pci_dev *pdev) goto err_out_regions; } -pci_set_master (pdev); rc =pci_enable_pcie_error_reporting (pdev);if (rc) { dev_err(&pdev->dev,
@@ -3427,10 +3416,6 @@ static int skd_pci_resume(struct pci_dev *pdev)pci_disable_pcie_error_reporting (pdev); err_out_regions: -pci_release_regions (pdev); - -err_out: -pci_disable_device (pdev);return rc; }
@@ -3450,13 +3435,14 @@ static void skd_pci_shutdown(struct pci_dev *pdev)skd_stop_device (skdev); } +static SIMPLE_DEV_PM_OPS(skd_pci_pm_ops, skd_pci_suspend, skd_pci_resume); +static struct pci_driver skd_driver = { .name =DRV_NAME , .id_table = skd_pci_tbl, .probe = skd_pci_probe, .remove = skd_pci_remove, - .suspend = skd_pci_suspend, - .resume = skd_pci_resume, + .driver.pm = &skd_pci_pm_ops, .shutdown = skd_pci_shutdown, };
Signature · 1 line
2.27.0
Re: [PATCH v3 3/3] skd: use generic power management
Quoted 99 lines · Vaibhav Gupta
Drivers using legacy power management .suspen()/.resume() callbacks have to manage PCI states and device's PM states themselves. They also need to take care of standard configuration registers.
Switch to generic power management framework using a single "struct
dev_pm_ops" variable to take the unnecessary load from the driver. This also avoids the need for the driver to directly call most of the PCI helper functions and device power state control functions, as through the generic framework PCI Core takes care of the necessary operations, and drivers are required to do only device-specific jobs.Signed-off-by Vaibhav Gupta <[email protected]>--- drivers/block/skd_main.c | 30 ++++++++---------------------- 1 file changed, 8 insertions(+), 22 deletions(-)drivers/block/skd_main.c +7
-20@@ -3315,10 +3315,11 @@ static void skd_pci_remove(struct pci_dev *pdev)return ; } -static int skd_pci_suspend (struct pci_dev *pdev,pm_message_t state) +static int __maybe_unused skd_pci_suspend (struct device *dev) {int i;struct skd_device *skdev; +struct pci_dev *pdev =to_pci_dev (dev); skdev =pci_get_drvdata (pdev);if (!skdev) {@@ -3337,18 +3338,15 @@ static int skd_pci_suspend(struct pci_dev *pdev, pm_message_t state)if (skdev->pcie_error_reporting_is_enabled)pci_disable_pcie_error_reporting (pdev); -pci_release_regions (pdev); -pci_save_state (pdev); -pci_disable_device (pdev); -pci_set_power_state (pdev,pci_choose_state (pdev, state));return 0 ; } -static int skd_pci_resume (struct pci_dev *pdev) +static int __maybe_unused skd_pci_resume (struct device *dev) {int i;int rc =0 ;struct skd_device *skdev; +struct pci_dev *pdev =to_pci_dev (dev); skdev =pci_get_drvdata (pdev);if (!skdev) {@@ -3356,16 +3354,8 @@ static int skd_pci_resume(struct pci_dev *pdev)return -1 ; } -pci_set_power_state (pdev,PCI_D0 ); -pci_enable_wake (pdev,PCI_D0 ,0 ); -pci_restore_state (pdev); +device_wakeup_disable (dev); - rc =pci_enable_device (pdev); -if (rc) -return rc; - rc =pci_request_regions (pdev,DRV_NAME ); -if (rc) - goto err_out; rc =dma_set_mask_and_coherent (&pdev->dev,DMA_BIT_MASK (64 ));if (rc) rc =dma_set_mask_and_coherent (&pdev->dev,DMA_BIT_MASK (32 ));@@ -3374,7 +3364,6 @@ static int skd_pci_resume(struct pci_dev *pdev) goto err_out_regions; } -pci_set_master (pdev); rc =pci_enable_pcie_error_reporting (pdev);if (rc) { dev_err(&pdev->dev,@@ -3427,10 +3416,6 @@ static int skd_pci_resume(struct pci_dev *pdev)pci_disable_pcie_error_reporting (pdev); err_out_regions: -pci_release_regions (pdev); - -err_out: -pci_disable_device (pdev);return rc; }@@ -3450,13 +3435,14 @@ static void skd_pci_shutdown(struct pci_dev *pdev)skd_stop_device (skdev); } +static SIMPLE_DEV_PM_OPS(skd_pci_pm_ops, skd_pci_suspend, skd_pci_resume); +static struct pci_driver skd_driver = {
drivers/block/skd_main.c static void skd_pci_shutdown(struct pci_dev *pdev) · line 3448
Quoted 10 lines
drivers/block/skd_main.c +1
-2.name = DRV_NAME, .id_table = skd_pci_tbl, .probe = skd_pci_probe, .remove = skd_pci_remove, - .suspend = skd_pci_suspend, - .resume = skd_pci_resume, + .driver.pm = &skd_pci_pm_ops, .shutdown = skd_pci_shutdown, };
Signature · 2 lines
Damien Le Moal Western Digital Research
Quoted 37 lines · Vaibhav Gupta
Linux Kernel Mentee: Remove Legacy Power Management.
The purpose of this patch series is to upgrade power management in block drivers. This has been done by upgrading .suspend() and .resume() callbacks.
The upgrade makes sure that the involvement of PCI Core does not change the order of operations executed in a driver. Thus, does not change its behavior.
In general, drivers with legacy PM, .suspend() and .resume() make use of PCI helper functions like pci_enable/disable_device_mem(),
pci_set_power_state(), pci_save/restore_state(), pci_enable/disable_device(), etc. to complete their job.The conversion requires the removal of those function calls, change the callbacks' definition accordingly and make use of
dev_pm_opsstructure.All patches are compile-tested only.
Test tools:
- Compiler: gcc (GCC) 10.1.0
- allmodconfig build: make -j$(nproc) W=1 allv3: v2 needed some changes in commit messages.
Vaibhav Gupta (3):
mtip32xx: use generic power management
rsxx: use generic power management
skd: use generic power managementdrivers/block/mtip32xx/mtip32xx.c | 54 +++++++------------------------ drivers/block/rsxx/core.c | 9 ++++-- drivers/block/skd_main.c | 30 +++++------------ 3 files changed, 27 insertions(+), 66 deletions(-)Signature · 1 line
2.27.0
v4 cover block: use generic power management
In reply to Vaibhav GuptaLinux Kernel Mentee: Remove Legacy Power Management.
The purpose of this patch series is to upgrade power management in block drivers. This has been done by upgrading .suspend() and .resume() callbacks.
The upgrade makes sure that the involvement of PCI Core does not change the order of operations executed in a driver. Thus, does not change its behavior.
In general, drivers with legacy PM, .suspend() and .resume() make use of PCI helper functions like pci_enable/disable_device_mem(), pci_set_power_state(), pci_save/restore_state(), pci_enable/disable_device(), etc. to complete their job.
The conversion requires the removal of those function calls, change the callbacks' definition accordingly and make use of dev_pm_ops structure.
All patches are compile-tested only.
Test tools:
- Compiler: gcc (GCC) 10.2.0
- allmodconfig build: make -j$(nproc) W=1 all
v4: patch-set rebased.
Vaibhav Gupta (3):
mtip32xx: use generic power management
rsxx: use generic power management
skd: use generic power management
drivers/block/mtip32xx/mtip32xx.c | 54 +++++++------------------------ drivers/block/rsxx/core.c | 9 ++++-- drivers/block/skd_main.c | 30 +++++------------ 3 files changed, 27 insertions(+), 66 deletions(-)
Signature · 6 lines
2.30.0 _______________________________________________ Linux-kernel-mentees mailing list [email protected] https://lists.linuxfoundation.org/mailman/listinfo/linux-kernel-mentees
PATCH v4 1/3 mtip32xx: use generic power management
Drivers using legacy power management .suspen()/.resume() callbacks have to manage PCI states and device's PM states themselves. They also need to take care of standard configuration registers.
Switch to generic power management framework using a single "struct dev_pm_ops" variable to take the unnecessary load from the driver. This also avoids the need for the driver to directly call most of the PCI helper functions and device power state control functions, as through the generic framework PCI Core takes care of the necessary operations, and drivers are required to do only device-specific jobs.
--- drivers/block/mtip32xx/mtip32xx.c | 54 +++++++------------------------ 1 file changed, 12 insertions(+), 42 deletions(-)
drivers/block/mtip32xx/mtip32xx.c +12 -42
@@ -4232,14 +4232,13 @@ static void mtip_pci_remove(struct pci_dev *pdev)* 0 Success * <0 Error */ -static int mtip_pci_suspend (struct pci_dev *pdev,pm_message_t mesg) +static int __maybe_unused mtip_pci_suspend (struct device *dev) {int rv =0 ; -struct driver_data *dd =pci_get_drvdata (pdev); +struct driver_data *dd =dev_get_drvdata (dev);if (!dd) { -dev_err (&pdev->dev, -"Driver private datastructure is NULL\n" ); +dev_err (dev,"Driver private datastructure is NULL\n" );return -EFAULT ; }
@@ -4247,21 +4246,8 @@ static int mtip_pci_suspend(struct pci_dev *pdev, pm_message_t mesg)/* Disable ports & interrupts then send standby immediate */ rv =mtip_block_suspend (dd); -if (rv <0 ) { -dev_err (&pdev->dev, -"Failed to suspend controller\n" ); -return rv; - } - -/* -* Save the pci config space to pdev structure & -* disable the device -*/ -pci_save_state (pdev); -pci_disable_device (pdev); - -/* Move to Low power state*/ -pci_set_power_state (pdev, PCI_D3hot); +if (rv <0 ) +dev_err (dev,"Failed to suspend controller\n" );return rv; }
@@ -4273,42 +4259,25 @@ static int mtip_pci_suspend(struct pci_dev *pdev, pm_message_t mesg)* 0 Success * <0 Error */ -static int mtip_pci_resume (struct pci_dev *pdev) +static int __maybe_unused mtip_pci_resume (struct device *dev) {int rv =0 ;struct driver_data *dd; - dd =pci_get_drvdata (pdev); + dd =dev_get_drvdata (dev);if (!dd) { -dev_err (&pdev->dev, -"Driver private datastructure is NULL\n" ); +dev_err (dev,"Driver private datastructure is NULL\n" );return -EFAULT ; } -/* Move the device to active State */ -pci_set_power_state (pdev,PCI_D0 ); - -/* Restore PCI configuration space */ -pci_restore_state (pdev); - -/* Enable the PCI device*/ - rv =pcim_enable_device (pdev); -if (rv <0 ) { -dev_err (&pdev->dev, -"Failed to enable card during resume\n" ); - goto err; - } -pci_set_master (pdev); -/* * Calls hbaReset, initPort, & startPort function * then enables interrupts */ rv =mtip_block_resume (dd);if (rv <0 ) -dev_err (&pdev->dev,"Unable to resume\n" ); +dev_err (dev,"Unable to resume\n" ); -err:clear_bit (MTIP_DDF_RESUME_BIT , &dd->dd_flag);return rv;
@@ -4339,14 +4308,15 @@ static const struct pci_device_id mtip_pci_tbl[] = { { 0 } }; +static SIMPLE_DEV_PM_OPS(mtip_pci_pm_ops, mtip_pci_suspend, mtip_pci_resume); +/* Structure that describes the PCI driver functions. */ static struct pci_driver mtip_pci_driver = { .name =MTIP_DRV_NAME , .id_table = mtip_pci_tbl, .probe = mtip_pci_probe, .remove = mtip_pci_remove, - .suspend = mtip_pci_suspend, - .resume = mtip_pci_resume, + .driver.pm = &mtip_pci_pm_ops, .shutdown = mtip_pci_shutdown, };
Signature · 6 lines
2.30.0 _______________________________________________ Linux-kernel-mentees mailing list [email protected] https://lists.linuxfoundation.org/mailman/listinfo/linux-kernel-mentees
PATCH v4 2/3 rsxx: use generic power management
In reply to Vaibhav GuptaDrivers using legacy power management .suspen()/.resume() callbacks have to manage PCI states and device's PM states themselves. They also need to take care of standard configuration registers.
Switch to generic power management framework using a single "struct dev_pm_ops" variable to take the unnecessary load from the driver. This also avoids the need for the driver to directly call most of the PCI helper functions and device power state control functions, as through the generic framework PCI Core takes care of the necessary operations, and drivers are required to do only device-specific jobs.
--- drivers/block/rsxx/core.c | 9 +++++++-- 1 file changed, 7 insertions(+), 2 deletions(-)
drivers/block/rsxx/core.c +7 -2
@@ -1032,12 +1032,15 @@ static void rsxx_pci_remove(struct pci_dev *dev)kfree (card); } -static int rsxx_pci_suspend (struct pci_dev *dev,pm_message_t state) +static int __maybe_unused rsxx_pci_suspend ( + __attribute__((unused))struct device *dev) {/* We don't support suspend at this time. */ return -ENOSYS ; } +#define rsxx_pci_resume NULL +static void rsxx_pci_shutdown (struct pci_dev *dev) {struct rsxx_cardinfo *card =pci_get_drvdata (dev);
@@ -1073,12 +1076,14 @@ static const struct pci_device_id rsxx_pci_ids[] = {MODULE_DEVICE_TABLE (pci, rsxx_pci_ids); +static SIMPLE_DEV_PM_OPS(rsxx_pci_pm_ops, rsxx_pci_suspend, rsxx_pci_resume); +static struct pci_driver rsxx_pci_driver = { .name =DRIVER_NAME , .id_table = rsxx_pci_ids, .probe = rsxx_pci_probe, .remove = rsxx_pci_remove, - .suspend = rsxx_pci_suspend, + .driver.pm = &rsxx_pci_pm_ops, .shutdown = rsxx_pci_shutdown, .err_handler = &rsxx_err_handler, };
Signature · 6 lines
2.30.0 _______________________________________________ Linux-kernel-mentees mailing list [email protected] https://lists.linuxfoundation.org/mailman/listinfo/linux-kernel-mentees
PATCH v4 3/3 skd: use generic power management
In reply to Vaibhav GuptaDrivers using legacy power management .suspen()/.resume() callbacks have to manage PCI states and device's PM states themselves. They also need to take care of standard configuration registers.
Switch to generic power management framework using a single "struct dev_pm_ops" variable to take the unnecessary load from the driver. This also avoids the need for the driver to directly call most of the PCI helper functions and device power state control functions, as through the generic framework PCI Core takes care of the necessary operations, and drivers are required to do only device-specific jobs.
--- drivers/block/skd_main.c | 30 ++++++++---------------------- 1 file changed, 8 insertions(+), 22 deletions(-)
drivers/block/skd_main.c +8 -22
@@ -3317,10 +3317,11 @@ static void skd_pci_remove(struct pci_dev *pdev)return ; } -static int skd_pci_suspend (struct pci_dev *pdev,pm_message_t state) +static int __maybe_unused skd_pci_suspend (struct device *dev) {int i;struct skd_device *skdev; +struct pci_dev *pdev =to_pci_dev (dev); skdev =pci_get_drvdata (pdev);if (!skdev) {
@@ -3339,18 +3340,15 @@ static int skd_pci_suspend(struct pci_dev *pdev, pm_message_t state)if (skdev->pcie_error_reporting_is_enabled)pci_disable_pcie_error_reporting (pdev); -pci_release_regions (pdev); -pci_save_state (pdev); -pci_disable_device (pdev); -pci_set_power_state (pdev,pci_choose_state (pdev, state));return 0 ; } -static int skd_pci_resume (struct pci_dev *pdev) +static int __maybe_unused skd_pci_resume (struct device *dev) {int i;int rc =0 ;struct skd_device *skdev; +struct pci_dev *pdev =to_pci_dev (dev); skdev =pci_get_drvdata (pdev);if (!skdev) {
@@ -3358,16 +3356,8 @@ static int skd_pci_resume(struct pci_dev *pdev)return -1 ; } -pci_set_power_state (pdev,PCI_D0 ); -pci_enable_wake (pdev,PCI_D0 ,0 ); -pci_restore_state (pdev); +device_wakeup_disable (dev); - rc =pci_enable_device (pdev); -if (rc) -return rc; - rc =pci_request_regions (pdev,DRV_NAME ); -if (rc) - goto err_out; rc =dma_set_mask_and_coherent (&pdev->dev,DMA_BIT_MASK (64 ));if (rc) rc =dma_set_mask_and_coherent (&pdev->dev,DMA_BIT_MASK (32 ));
@@ -3376,7 +3366,6 @@ static int skd_pci_resume(struct pci_dev *pdev) goto err_out_regions; } -pci_set_master (pdev); rc =pci_enable_pcie_error_reporting (pdev);if (rc) { dev_err(&pdev->dev,
@@ -3429,10 +3418,6 @@ static int skd_pci_resume(struct pci_dev *pdev)pci_disable_pcie_error_reporting (pdev); err_out_regions: -pci_release_regions (pdev); - -err_out: -pci_disable_device (pdev);return rc; }
@@ -3452,13 +3437,14 @@ static void skd_pci_shutdown(struct pci_dev *pdev)skd_stop_device (skdev); } +static SIMPLE_DEV_PM_OPS(skd_pci_pm_ops, skd_pci_suspend, skd_pci_resume); +static struct pci_driver skd_driver = { .name =DRV_NAME , .id_table = skd_pci_tbl, .probe = skd_pci_probe, .remove = skd_pci_remove, - .suspend = skd_pci_suspend, - .resume = skd_pci_resume, + .driver.pm = &skd_pci_pm_ops, .shutdown = skd_pci_shutdown, };
Signature · 6 lines
2.30.0 _______________________________________________ Linux-kernel-mentees mailing list [email protected] https://lists.linuxfoundation.org/mailman/listinfo/linux-kernel-mentees
Re: [PATCH v2 3/3] skd: use generic power management
In reply to Vaibhav GuptaQuoted 3 lines · Vaibhav Gupta
Okay. I will improve on it. Just inform me after testing that if any other changes are required. I guess [PATCH 1/3] and [PATCH 2/3] are okay, so I will only send v3 of [PATCH 3/3] after suggested changes.
FWIW, there's a recent conversation on [email protected] about updating individual patches in a series (sorry, can't find a link to it). But the gist of it was that posting only [v3 3/3] leads to confusion because
- we can't tell whether [v3 1/3] and [v3 2/3] got lost en-route, and
- collecting things from v2 and v3 is more work for the maintainer.
Bottom line: repost the whole series, even if some patches haven't changed.
Bjorn
Re: [PATCH v2 3/3] skd: use generic power management
Quoted 10 lines · Bjorn Helgaas
Quoted 3 lines · Vaibhav Gupta
Okay. I will improve on it. Just inform me after testing that if any other changes are required. I guess [PATCH 1/3] and [PATCH 2/3] are okay, so I will only send v3 of [PATCH 3/3] after suggested changes.
FWIW, there's a recent conversation on [email protected] about updating individual patches in a series (sorry, can't find a link to it). But the gist of it was that posting only [v3 3/3] leads to confusion because
Quoted 6 lines
- we can't tell whether [v3 1/3] and [v3 2/3] got lost en-route, and
- collecting things from v2 and v3 is more work for the maintainer.
Bottom line: repost the whole series, even if some patches haven't changed.
Thanks a lot for updating me with the info :D .
I sent the whole series as v3 hours ago.
--Vaibhav Gupta
Quoted 2 lines
Bjorn