| Message ID | 20210529192527.2708-6-ameynarkhede03@gmail.com (mailing list archive) |
|---|---|
| State | Superseded |
| Delegated to: | Bjorn Helgaas |
| Headers | show |
| Series | Expose and manage PCI device reset | expand |
[+cc Rafael in case you have any ideas about acpi_bind_one() below] Mention ACPI in the subject, e.g., PCI: Add support for ACPI _RST reset method On Sun, May 30, 2021 at 12:55:25AM +0530, Amey Narkhede wrote: > From: Shanker Donthineni <sdonthineni@nvidia.com> > > The _RST is a standard method specified in the ACPI specification. It > provides a function level reset when it is described in the acpi_device > context associated with PCI-device. > > Implement a new reset function pci_dev_acpi_reset() for probing RST > method and execute if it is defined in the firmware. The ACPI binding > information is available only after calling device_add(). To consider > _RST method, move pci_init_reset_methods() to end of pci_device_add() > and craete two sysfs entries reset & reset_methond from > pci_create_sysfs_dev_files() s/craete/create/ s/reset_methond/reset_method/ > The default priority of the acpi reset is set to below device-specific > and above hardware resets. s/acpi/ACPI/ > Signed-off-by: Shanker Donthineni <sdonthineni@nvidia.com> > Reviewed-by: Sinan Kaya <okaya@kernel.org> > --- > drivers/pci/pci-sysfs.c | 23 ++++++++++++++++++++--- > drivers/pci/pci.c | 30 ++++++++++++++++++++++++++++++ > drivers/pci/probe.c | 2 +- > include/linux/pci.h | 2 +- > 4 files changed, 52 insertions(+), 5 deletions(-) > > diff --git a/drivers/pci/pci-sysfs.c b/drivers/pci/pci-sysfs.c > index 04b3d6565..b332d7923 100644 > --- a/drivers/pci/pci-sysfs.c > +++ b/drivers/pci/pci-sysfs.c > @@ -1482,12 +1482,30 @@ static const struct attribute_group pci_dev_reset_attr_group = { > .is_visible = pci_dev_reset_attr_is_visible, > }; > > +const struct attribute_group *pci_dev_reset_groups[] = { > + &pci_dev_reset_attr_group, > + &pci_dev_reset_method_attr_group, > + NULL, > +}; These should be static sysfs attributes if possible, e.g., see e1d3f3268b0e ("PCI/sysfs: Convert "config" to static attribute"). pci_create_sysfs_dev_files() will soon be removed completely. > int __must_check pci_create_sysfs_dev_files(struct pci_dev *pdev) > { > + int retval; > + > if (!sysfs_initialized) > return -EACCES; > > - return pci_create_resource_files(pdev); > + retval = pci_create_resource_files(pdev); > + if (retval) > + return retval; > + > + retval = device_add_groups(&pdev->dev, pci_dev_reset_groups); > + if (retval) { > + pci_remove_resource_files(pdev); > + return retval; > + } > + > + return 0; > } > > /** > @@ -1501,6 +1519,7 @@ void pci_remove_sysfs_dev_files(struct pci_dev *pdev) > if (!sysfs_initialized) > return; > > + device_remove_groups(&pdev->dev, pci_dev_reset_groups); > pci_remove_resource_files(pdev); > } > > @@ -1594,8 +1613,6 @@ const struct attribute_group *pci_dev_groups[] = { > &pci_dev_group, > &pci_dev_config_attr_group, > &pci_dev_rom_attr_group, > - &pci_dev_reset_attr_group, > - &pci_dev_reset_method_attr_group, > &pci_dev_vpd_attr_group, > #ifdef CONFIG_DMI > &pci_dev_smbios_attr_group, > diff --git a/drivers/pci/pci.c b/drivers/pci/pci.c > index bbed852d9..4a7019d0b 100644 > --- a/drivers/pci/pci.c > +++ b/drivers/pci/pci.c > @@ -5115,6 +5115,35 @@ static void pci_dev_restore(struct pci_dev *dev) > err_handler->reset_done(dev); > } > > +/** > + * pci_dev_acpi_reset - do a function level reset using _RST method > + * @dev: device to reset > + * @probe: check if _RST method is included in the acpi_device context. > + */ > +static int pci_dev_acpi_reset(struct pci_dev *dev, int probe) > +{ > +#ifdef CONFIG_ACPI > + acpi_handle handle = ACPI_HANDLE(&dev->dev); > + > + /* Return -ENOTTY if _RST method is not included in the dev context */ > + if (!handle || !acpi_has_method(handle, "_RST")) > + return -ENOTTY; > + > + /* Return 0 for probe phase indicating that we can reset this device */ > + if (probe) > + return 0; > + > + /* Invoke _RST() method to perform a function level reset */ Superfluous comment. Actually all the single-line comments here are superfluous. > + if (ACPI_FAILURE(acpi_evaluate_object(handle, "_RST", NULL, NULL))) { > + pci_warn(dev, "Failed to reset the device\n"); The message should mention the type of reset, e.g., "ACPI _RST failed ..." > + return -EINVAL; > + } > + return 0; > +#else > + return -ENOTTY; > +#endif > +} > + > /* > * The ordering for functions in pci_reset_fn_methods > * is required for reset_methods byte array defined > @@ -5122,6 +5151,7 @@ static void pci_dev_restore(struct pci_dev *dev) > */ > const struct pci_reset_fn_method pci_reset_fn_methods[] = { > { &pci_dev_specific_reset, .name = "device_specific" }, > + { &pci_dev_acpi_reset, .name = "acpi" }, > { &pcie_reset_flr, .name = "flr" }, > { &pci_af_flr, .name = "af_flr" }, > { &pci_pm_reset, .name = "pm" }, > diff --git a/drivers/pci/probe.c b/drivers/pci/probe.c > index 90fd4f61f..eeab791a0 100644 > --- a/drivers/pci/probe.c > +++ b/drivers/pci/probe.c > @@ -2404,7 +2404,6 @@ static void pci_init_capabilities(struct pci_dev *dev) > pci_rcec_init(dev); /* Root Complex Event Collector */ > > pcie_report_downtraining(dev); > - pci_init_reset_methods(dev); > } > > /* > @@ -2495,6 +2494,7 @@ void pci_device_add(struct pci_dev *dev, struct pci_bus *bus) > dev->match_driver = false; > ret = device_add(&dev->dev); > WARN_ON(ret < 0); > + pci_init_reset_methods(dev); This is a little sketchy. We shouldn't be doing device config stuff after device_add() because that's when it becomes available for drivers to bind to the device. If we do anything with the device after that point, we may interfere with a driver. I think the problem is that we don't call acpi_bind_one() until device_add(). There's some hackery in pci-acpi.c to deal with a similar problem for something else -- see acpi_pci_bridge_d3(). I don't know how to fix this yet. Here's the call graph that I think is relevant: pci_scan_single_device pci_scan_device pci_set_of_node dev->dev.of_node = of_pci_find_child_device() <-- set OF stuff pci_device_add device_add device_platform_notify acpi_platform_notify case KOBJ_ADD: acpi_device_notify acpi_bind_one ACPI_COMPANION_SET() <-- sets ACPI_COMPANION pci_init_reset_methods pci_dev_acpi_reset(PCI_RESET_PROBE) handle = ACPI_HANDLE(&dev->dev) <-- uses ACPI_COMPANION I think it's kind of a general problem that we currently don't have access to the ACPI stuff until *after* device_add(). I included pci_set_of_node() in the graph above because that seems sort of like an OF analogue of what acpi_bind_one() is doing. I would really like to do the ACPI_COMPANION setup earlier, maybe at the same time as pci_set_of_node(). But I don't know enough about what acpi_bind_one() does -- there's a lot going on in there. > } > > struct pci_dev *pci_scan_single_device(struct pci_bus *bus, int devfn) > diff --git a/include/linux/pci.h b/include/linux/pci.h > index 6e9bc4f9c..a7f063da2 100644 > --- a/include/linux/pci.h > +++ b/include/linux/pci.h > @@ -49,7 +49,7 @@ > PCI_STATUS_SIG_TARGET_ABORT | \ > PCI_STATUS_PARITY) > > -#define PCI_RESET_METHODS_NUM 5 > +#define PCI_RESET_METHODS_NUM 6 > > /* > * The PCI interface treats multi-function devices as independent > -- > 2.31.1 >
Hi Bjorn, On 6/5/21 3:53 PM, Bjorn Helgaas wrote: > Mention ACPI in the subject, e.g., > > PCI: Add support for ACPI _RST reset method Will change in the next patch. > On Sun, May 30, 2021 at 12:55:25AM +0530, Amey Narkhede wrote: >> From: Shanker Donthineni <sdonthineni@nvidia.com> >> >> The _RST is a standard method specified in the ACPI specification. It >> provides a function level reset when it is described in the acpi_device >> context associated with PCI-device. >> >> Implement a new reset function pci_dev_acpi_reset() for probing RST >> method and execute if it is defined in the firmware. The ACPI binding >> information is available only after calling device_add(). To consider >> _RST method, move pci_init_reset_methods() to end of pci_device_add() >> and craete two sysfs entries reset & reset_methond from >> pci_create_sysfs_dev_files() > s/craete/create/ > s/reset_methond/reset_method/ Will fix it. >> The default priority of the acpi reset is set to below device-specific >> and above hardware resets. > s/acpi/ACPI/ Will fix it. >> Signed-off-by: Shanker Donthineni <sdonthineni@nvidia.com> >> Reviewed-by: Sinan Kaya <okaya@kernel.org> >> --- >> drivers/pci/pci-sysfs.c | 23 ++++++++++++++++++++--- >> drivers/pci/pci.c | 30 ++++++++++++++++++++++++++++++ >> drivers/pci/probe.c | 2 +- >> include/linux/pci.h | 2 +- >> 4 files changed, 52 insertions(+), 5 deletions(-) >> >> diff --git a/drivers/pci/pci-sysfs.c b/drivers/pci/pci-sysfs.c >> index 04b3d6565..b332d7923 100644 >> --- a/drivers/pci/pci-sysfs.c >> +++ b/drivers/pci/pci-sysfs.c >> @@ -1482,12 +1482,30 @@ static const struct attribute_group pci_dev_reset_attr_group = { >> .is_visible = pci_dev_reset_attr_is_visible, >> }; >> >> +const struct attribute_group *pci_dev_reset_groups[] = { >> + &pci_dev_reset_attr_group, >> + &pci_dev_reset_method_attr_group, >> + NULL, >> +}; > These should be static sysfs attributes if possible, e.g., see > e1d3f3268b0e ("PCI/sysfs: Convert "config" to static attribute"). > pci_create_sysfs_dev_files() will soon be removed completely. > >> int __must_check pci_create_sysfs_dev_files(struct pci_dev *pdev) >> { >> + int retval; >> + >> if (!sysfs_initialized) >> return -EACCES; >> >> - return pci_create_resource_files(pdev); >> + retval = pci_create_resource_files(pdev); >> + if (retval) >> + return retval; >> + >> + retval = device_add_groups(&pdev->dev, pci_dev_reset_groups); >> + if (retval) { >> + pci_remove_resource_files(pdev); >> + return retval; >> + } >> + >> + return 0; >> } >> >> /** >> @@ -1501,6 +1519,7 @@ void pci_remove_sysfs_dev_files(struct pci_dev *pdev) >> if (!sysfs_initialized) >> return; >> >> + device_remove_groups(&pdev->dev, pci_dev_reset_groups); >> pci_remove_resource_files(pdev); >> } >> >> @@ -1594,8 +1613,6 @@ const struct attribute_group *pci_dev_groups[] = { >> &pci_dev_group, >> &pci_dev_config_attr_group, >> &pci_dev_rom_attr_group, >> - &pci_dev_reset_attr_group, >> - &pci_dev_reset_method_attr_group, >> &pci_dev_vpd_attr_group, >> #ifdef CONFIG_DMI >> &pci_dev_smbios_attr_group, >> diff --git a/drivers/pci/pci.c b/drivers/pci/pci.c >> index bbed852d9..4a7019d0b 100644 >> --- a/drivers/pci/pci.c >> +++ b/drivers/pci/pci.c >> @@ -5115,6 +5115,35 @@ static void pci_dev_restore(struct pci_dev *dev) >> err_handler->reset_done(dev); >> } >> >> +/** >> + * pci_dev_acpi_reset - do a function level reset using _RST method >> + * @dev: device to reset >> + * @probe: check if _RST method is included in the acpi_device context. >> + */ >> +static int pci_dev_acpi_reset(struct pci_dev *dev, int probe) >> +{ >> +#ifdef CONFIG_ACPI >> + acpi_handle handle = ACPI_HANDLE(&dev->dev); >> + >> + /* Return -ENOTTY if _RST method is not included in the dev context */ >> + if (!handle || !acpi_has_method(handle, "_RST")) >> + return -ENOTTY; >> + >> + /* Return 0 for probe phase indicating that we can reset this device */ >> + if (probe) >> + return 0; >> + >> + /* Invoke _RST() method to perform a function level reset */ > Superfluous comment. Actually all the single-line comments here are > superfluous. Will remove in the next patch. >> + if (ACPI_FAILURE(acpi_evaluate_object(handle, "_RST", NULL, NULL))) { >> + pci_warn(dev, "Failed to reset the device\n"); > The message should mention the type of reset, e.g., "ACPI _RST failed ..." > Will change to pci_warn(dev, "ACPI _RST failed\n"); >> + return -EINVAL; >> + } >> + return 0; >> +#else >> + return -ENOTTY; >> +#endif >> +} >> + >> /* >> * The ordering for functions in pci_reset_fn_methods >> * is required for reset_methods byte array defined >> @@ -5122,6 +5151,7 @@ static void pci_dev_restore(struct pci_dev *dev) >> */ >> const struct pci_reset_fn_method pci_reset_fn_methods[] = { >> { &pci_dev_specific_reset, .name = "device_specific" }, >> + { &pci_dev_acpi_reset, .name = "acpi" }, >> { &pcie_reset_flr, .name = "flr" }, >> { &pci_af_flr, .name = "af_flr" }, >> { &pci_pm_reset, .name = "pm" }, >> diff --git a/drivers/pci/probe.c b/drivers/pci/probe.c >> index 90fd4f61f..eeab791a0 100644 >> --- a/drivers/pci/probe.c >> +++ b/drivers/pci/probe.c >> @@ -2404,7 +2404,6 @@ static void pci_init_capabilities(struct pci_dev *dev) >> pci_rcec_init(dev); /* Root Complex Event Collector */ >> >> pcie_report_downtraining(dev); >> - pci_init_reset_methods(dev); >> } >> >> /* >> @@ -2495,6 +2494,7 @@ void pci_device_add(struct pci_dev *dev, struct pci_bus *bus) >> dev->match_driver = false; >> ret = device_add(&dev->dev); >> WARN_ON(ret < 0); >> + pci_init_reset_methods(dev); > This is a little sketchy. We shouldn't be doing device config stuff > after device_add() because that's when it becomes available for > drivers to bind to the device. If we do anything with the device > after that point, we may interfere with a driver. The reason I did PCI driver attach/bind is happening from pci_bus_add_device() after setting 'dev->match_driver = true'. I thought it's safe to update reset methods after calling device_add() and before driver bind happens. void pci_bus_add_device(struct pci_dev *dev) { int retval; /* * Can not put in pci_device_add yet because resources * are not assigned yet for some devices. */ pcibios_bus_add_device(dev); pci_fixup_device(pci_fixup_final, dev); pci_create_sysfs_dev_files(dev); pci_proc_attach_device(dev); pci_bridge_d3_update(dev); dev->match_driver = true; retval = device_attach(&dev->dev); ---- > PCI driver bind call > I think the problem is that we don't call acpi_bind_one() until > device_add(). There's some hackery in pci-acpi.c to deal with a > similar problem for something else -- see acpi_pci_bridge_d3(). > > I don't know how to fix this yet. Here's the call graph that I think > is relevant: > > pci_scan_single_device > pci_scan_device > pci_set_of_node > dev->dev.of_node = of_pci_find_child_device() <-- set OF stuff > pci_device_add > device_add > device_platform_notify > acpi_platform_notify > case KOBJ_ADD: > acpi_device_notify > acpi_bind_one > ACPI_COMPANION_SET() <-- sets ACPI_COMPANION > pci_init_reset_methods > pci_dev_acpi_reset(PCI_RESET_PROBE) > handle = ACPI_HANDLE(&dev->dev) <-- uses ACPI_COMPANION > > I think it's kind of a general problem that we currently don't have > access to the ACPI stuff until *after* device_add(). I included > pci_set_of_node() in the graph above because that seems sort of > like an OF analogue of what acpi_bind_one() is doing. > > I would really like to do the ACPI_COMPANION setup earlier, maybe > at the same time as pci_set_of_node(). But I don't know enough about > what acpi_bind_one() does -- there's a lot going on in there. > Yes, it's a general problem ACPI binding information is available only after calling device_platform_notify(). There are no exported functions to set the ACPI_COMPANION like OF_NODE. Another approach: It simplifies the code logic if we update reset methods when creating sysfs entries 'reset' and 'reset_method'. I've verified this code and getting an expected behavior. root@jetson:~# cat /sys/bus/pci/devices/0009\:01\:00.0/reset_method acpi,flr --- a/drivers/pci/pci-sysfs.c +++ b/drivers/pci/pci-sysfs.c @@ -1427,7 +1427,7 @@ static umode_t pci_dev_reset_method_attr_is_visible(struct kobject *kobj, { struct pci_dev *pdev = to_pci_dev(kobj_to_dev(kobj)); - if (!pci_reset_supported(pdev)) + if (!pci_reset_supported(pdev) && !pci_init_reset_methods(pdev)) return 0; return a->mode; @@ -1471,7 +1471,7 @@ static umode_t pci_dev_reset_attr_is_visible(struct kobject *kobj, { struct pci_dev *pdev = to_pci_dev(kobj_to_dev(kobj)); - if (!pci_reset_supported(pdev)) + if (!pci_reset_supported(pdev) && !pci_init_reset_methods(pdev)) return 0; return a->mode; diff --git a/drivers/pci/pci.c b/drivers/pci/pci.c index bbed852d977f1..13654255fa3dc 100644 --- a/drivers/pci/pci.c +++ b/drivers/pci/pci.c @@ -5115,6 +5115,32 @@ static void pci_dev_restore(struct pci_dev *dev) err_handler->reset_done(dev); } +/** + * pci_dev_acpi_reset - do a function level reset using _RST method + * @dev: device to reset + * @probe: check if _RST method is included in the acpi_device context. + */ +static int pci_dev_acpi_reset(struct pci_dev *dev, int probe) +{ +#ifdef CONFIG_ACPI + acpi_handle handle = ACPI_HANDLE(&dev->dev); + + if (!handle || !acpi_has_method(handle, "_RST")) + return -ENOTTY; + + if (probe) + return 0; + + if (ACPI_FAILURE(acpi_evaluate_object(handle, "_RST", NULL, NULL))) { + pci_warn(dev, "ACPI _RST failed\n"); + return -EINVAL; + } + return 0; +#else + return -ENOTTY; +#endif +} + /* * The ordering for functions in pci_reset_fn_methods * is required for reset_methods byte array defined @@ -5122,6 +5148,7 @@ static void pci_dev_restore(struct pci_dev *dev) */ const struct pci_reset_fn_method pci_reset_fn_methods[] = { { &pci_dev_specific_reset, .name = "device_specific" }, + { &pci_dev_acpi_reset, .name = "acpi" }, { &pcie_reset_flr, .name = "flr" }, { &pci_af_flr, .name = "af_flr" }, { &pci_pm_reset, .name = "pm" }, @@ -5191,7 +5218,7 @@ EXPORT_SYMBOL_GPL(__pci_reset_function_locked); * Stores reset mechanisms supported by device in reset_methods byte array * which is a member of struct pci_dev. */ -void pci_init_reset_methods(struct pci_dev *dev) +bool pci_init_reset_methods(struct pci_dev *dev) { int i, rc; u8 prio = PCI_RESET_METHODS_NUM; @@ -5209,6 +5236,7 @@ void pci_init_reset_methods(struct pci_dev *dev) break; } memcpy(dev->reset_methods, reset_methods, sizeof(reset_methods)); + return pci_reset_supported(dev); } /** diff --git a/drivers/pci/pci.h b/drivers/pci/pci.h index 13ec6bd6f4f76..3e871a5a21bbd 100644 --- a/drivers/pci/pci.h +++ b/drivers/pci/pci.h @@ -33,7 +33,7 @@ enum pci_mmap_api { int pci_mmap_fits(struct pci_dev *pdev, int resno, struct vm_area_struct *vmai, enum pci_mmap_api mmap_api); -void pci_init_reset_methods(struct pci_dev *dev); +bool pci_init_reset_methods(struct pci_dev *dev); int pci_bridge_secondary_bus_reset(struct pci_dev *dev); int pci_bus_error_reset(struct pci_dev *dev); -void pci_init_reset_methods(struct pci_dev *dev); +bool pci_init_reset_methods(struct pci_dev *dev); int pci_bridge_secondary_bus_reset(struct pci_dev *dev); int pci_bus_error_reset(struct pci_dev *dev); diff --git a/drivers/pci/probe.c b/drivers/pci/probe.c index 90fd4f61f3802..275a067d7a282 100644 --- a/drivers/pci/probe.c +++ b/drivers/pci/probe.c @@ -2404,7 +2404,6 @@ static void pci_init_capabilities(struct pci_dev *dev) pci_rcec_init(dev); /* Root Complex Event Collector */ pcie_report_downtraining(dev); - pci_init_reset_methods(dev); }
Hi Bjorn, On 6/5/21 3:53 PM, Bjorn Helgaas wrote: > This is a little sketchy. We shouldn't be doing device config stuff > after device_add() because that's when it becomes available for > drivers to bind to the device. If we do anything with the device > after that point, we may interfere with a driver. > > I think the problem is that we don't call acpi_bind_one() until > device_add(). There's some hackery in pci-acpi.c to deal with a > similar problem for something else -- see acpi_pci_bridge_d3(). > > I don't know how to fix this yet. Here's the call graph that I think > is relevant: I've refactored pci_dev_acpi_reset() to avoid dependency on acpi_bind_one(). It can be called any time after creating pci_dev object. The code logic is not exactly same as acpi_pci_bridge_d3() but similar flow. No need to set ACPI_COMPANION since it would be updated eventually after probing the reset methods. Please review the below code and provide suggestions for the next step. Updated patch: [PATCH v5 5/7] PCI: Add support for ACPI _RST reset method The _RST is a standard method specified in the ACPI specification. It provides a function level reset when it is described in the acpi_device context associated with PCI-device. Implement a new reset function pci_dev_acpi_reset() for probing RST method and execute if it is defined in the firmware. The ACPI binding information is available only after calling device_add(). Since the ACPI_COMPANION was not done before calling pci_init_reset_methods(), use acpi_pci_find_companion() to know the ACPI binding. The default priority of the ACPI reset is set to below device-specific and above hardware resets. Signed-off-by: Shanker Donthineni <sdonthineni@nvidia.com> Suggested-by: Alex Williamson <alex.williamson@redhat.com> Reviewed-by: Sinan Kaya <okaya@kernel.org> --- drivers/pci/pci-acpi.c | 30 ++++++++++++++++++++++++++++++ drivers/pci/pci.c | 1 + drivers/pci/pci.h | 6 ++++++ include/linux/pci.h | 2 +- 4 files changed, 38 insertions(+), 1 deletion(-) diff --git a/drivers/pci/pci-acpi.c b/drivers/pci/pci-acpi.c index 36bc23e217592..c344c33f5c910 100644 --- a/drivers/pci/pci-acpi.c +++ b/drivers/pci/pci-acpi.c @@ -934,6 +934,36 @@ static pci_power_t acpi_pci_choose_state(struct pci_dev *pdev) static struct acpi_device *acpi_pci_find_companion(struct device *dev); +/** + * pci_dev_acpi_reset - do a function level reset using _RST method + * @dev: device to reset + * @probe: check if _RST method is included in the acpi_device context. + */ +int pci_dev_acpi_reset(struct pci_dev *dev, int probe) +{ + acpi_handle handle = ACPI_HANDLE(&dev->dev); + + /* Find out ACPI_HANDLE if not available in the device context */ + if (!handle) { + handle = acpi_device_handle(acpi_pci_find_companion(&dev->dev)); + if (!handle) + return -ENOTTY; + } + + if (!acpi_has_method(handle, "_RST")) + return -ENOTTY; + + if (probe) + return 0; + + if (ACPI_FAILURE(acpi_evaluate_object(handle, "_RST", NULL, NULL))) { + pci_warn(dev, "ACPI _RST failed\n"); + return -EINVAL; + } + + return 0; +} + static bool acpi_pci_bridge_d3(struct pci_dev *dev) { const struct fwnode_handle *fwnode; diff --git a/drivers/pci/pci.c b/drivers/pci/pci.c index bbed852d977f1..5726d120b70a2 100644 --- a/drivers/pci/pci.c +++ b/drivers/pci/pci.c @@ -5122,6 +5122,7 @@ static void pci_dev_restore(struct pci_dev *dev) */ const struct pci_reset_fn_method pci_reset_fn_methods[] = { { &pci_dev_specific_reset, .name = "device_specific" }, + { &pci_dev_acpi_reset, .name = "acpi" }, { &pcie_reset_flr, .name = "flr" }, { &pci_af_flr, .name = "af_flr" }, { &pci_pm_reset, .name = "pm" }, diff --git a/drivers/pci/pci.h b/drivers/pci/pci.h index 13ec6bd6f4f76..f3974ed1a99c2 100644 --- a/drivers/pci/pci.h +++ b/drivers/pci/pci.h @@ -703,7 +703,13 @@ static inline int pci_aer_raw_clear_status(struct pci_dev *dev) { return -EINVAL #ifdef CONFIG_ACPI int pci_acpi_program_hp_params(struct pci_dev *dev); extern const struct attribute_group pci_dev_acpi_attr_group; +int pci_dev_acpi_reset(struct pci_dev *dev, int probe); #else +static inline int pci_dev_acpi_reset(struct pci_dev *dev, int probe) +{ + return -ENOTTY; +} + static inline int pci_acpi_program_hp_params(struct pci_dev *dev) { return -ENODEV; diff --git a/include/linux/pci.h b/include/linux/pci.h index 6e9bc4f9cdab4..a7f063da2fe5f 100644 --- a/include/linux/pci.h +++ b/include/linux/pci.h @@ -49,7 +49,7 @@ PCI_STATUS_SIG_TARGET_ABORT | \ PCI_STATUS_PARITY) -#define PCI_RESET_METHODS_NUM 5 +#define PCI_RESET_METHODS_NUM 6
Hi Bjorn, On 6/5/21 3:53 PM, Bjorn Helgaas wrote: > This is a little sketchy. We shouldn't be doing device config stuff > after device_add() because that's when it becomes available for > drivers to bind to the device. If we do anything with the device > after that point, we may interfere with a driver. > > I think the problem is that we don't call acpi_bind_one() until > device_add(). There's some hackery in pci-acpi.c to deal with a > similar problem for something else -- see acpi_pci_bridge_d3(). Thanks for pointing to reference code. I've added a new patch for setting ACPI_COMPANION before device_add(). Please review the updated patch series v7 'Expose and manage PCI device reset' and provide feedback. -Shanker
diff --git a/drivers/pci/pci-sysfs.c b/drivers/pci/pci-sysfs.c index 04b3d6565..b332d7923 100644 --- a/drivers/pci/pci-sysfs.c +++ b/drivers/pci/pci-sysfs.c @@ -1482,12 +1482,30 @@ static const struct attribute_group pci_dev_reset_attr_group = { .is_visible = pci_dev_reset_attr_is_visible, }; +const struct attribute_group *pci_dev_reset_groups[] = { + &pci_dev_reset_attr_group, + &pci_dev_reset_method_attr_group, + NULL, +}; + int __must_check pci_create_sysfs_dev_files(struct pci_dev *pdev) { + int retval; + if (!sysfs_initialized) return -EACCES; - return pci_create_resource_files(pdev); + retval = pci_create_resource_files(pdev); + if (retval) + return retval; + + retval = device_add_groups(&pdev->dev, pci_dev_reset_groups); + if (retval) { + pci_remove_resource_files(pdev); + return retval; + } + + return 0; } /** @@ -1501,6 +1519,7 @@ void pci_remove_sysfs_dev_files(struct pci_dev *pdev) if (!sysfs_initialized) return; + device_remove_groups(&pdev->dev, pci_dev_reset_groups); pci_remove_resource_files(pdev); } @@ -1594,8 +1613,6 @@ const struct attribute_group *pci_dev_groups[] = { &pci_dev_group, &pci_dev_config_attr_group, &pci_dev_rom_attr_group, - &pci_dev_reset_attr_group, - &pci_dev_reset_method_attr_group, &pci_dev_vpd_attr_group, #ifdef CONFIG_DMI &pci_dev_smbios_attr_group, diff --git a/drivers/pci/pci.c b/drivers/pci/pci.c index bbed852d9..4a7019d0b 100644 --- a/drivers/pci/pci.c +++ b/drivers/pci/pci.c @@ -5115,6 +5115,35 @@ static void pci_dev_restore(struct pci_dev *dev) err_handler->reset_done(dev); } +/** + * pci_dev_acpi_reset - do a function level reset using _RST method + * @dev: device to reset + * @probe: check if _RST method is included in the acpi_device context. + */ +static int pci_dev_acpi_reset(struct pci_dev *dev, int probe) +{ +#ifdef CONFIG_ACPI + acpi_handle handle = ACPI_HANDLE(&dev->dev); + + /* Return -ENOTTY if _RST method is not included in the dev context */ + if (!handle || !acpi_has_method(handle, "_RST")) + return -ENOTTY; + + /* Return 0 for probe phase indicating that we can reset this device */ + if (probe) + return 0; + + /* Invoke _RST() method to perform a function level reset */ + if (ACPI_FAILURE(acpi_evaluate_object(handle, "_RST", NULL, NULL))) { + pci_warn(dev, "Failed to reset the device\n"); + return -EINVAL; + } + return 0; +#else + return -ENOTTY; +#endif +} + /* * The ordering for functions in pci_reset_fn_methods * is required for reset_methods byte array defined @@ -5122,6 +5151,7 @@ static void pci_dev_restore(struct pci_dev *dev) */ const struct pci_reset_fn_method pci_reset_fn_methods[] = { { &pci_dev_specific_reset, .name = "device_specific" }, + { &pci_dev_acpi_reset, .name = "acpi" }, { &pcie_reset_flr, .name = "flr" }, { &pci_af_flr, .name = "af_flr" }, { &pci_pm_reset, .name = "pm" }, diff --git a/drivers/pci/probe.c b/drivers/pci/probe.c index 90fd4f61f..eeab791a0 100644 --- a/drivers/pci/probe.c +++ b/drivers/pci/probe.c @@ -2404,7 +2404,6 @@ static void pci_init_capabilities(struct pci_dev *dev) pci_rcec_init(dev); /* Root Complex Event Collector */ pcie_report_downtraining(dev); - pci_init_reset_methods(dev); } /* @@ -2495,6 +2494,7 @@ void pci_device_add(struct pci_dev *dev, struct pci_bus *bus) dev->match_driver = false; ret = device_add(&dev->dev); WARN_ON(ret < 0); + pci_init_reset_methods(dev); } struct pci_dev *pci_scan_single_device(struct pci_bus *bus, int devfn) diff --git a/include/linux/pci.h b/include/linux/pci.h index 6e9bc4f9c..a7f063da2 100644 --- a/include/linux/pci.h +++ b/include/linux/pci.h @@ -49,7 +49,7 @@ PCI_STATUS_SIG_TARGET_ABORT | \ PCI_STATUS_PARITY) -#define PCI_RESET_METHODS_NUM 5 +#define PCI_RESET_METHODS_NUM 6 /* * The PCI interface treats multi-function devices as independent