by kernel test robot

[permalink] [raw]

Subject: Re: [PATCH v15 5/9] PCI/AER: Factor out error reporting from AER

Hi Oza,

Thank you for the patch! Yet something to improve:

[auto build test ERROR on pci/next]
[also build test ERROR on v4.17-rc3 next-20180503]
[if your patch is applied to the wrong git tree, please drop us a note to help improve the system]

url: https://github.com/0day-ci/linux/commits/Oza-Pawandeep/Address-error-and-recovery-for-AER-and-DPC/20180504-050017
base: https://git.kernel.org/pub/scm/linux/kernel/git/helgaas/pci.git next
config: x86_64-randconfig-x015-201817 (attached as .config)
compiler: gcc-7 (Debian 7.3.0-16) 7.3.0
reproduce:
# save the attached .config to linux build tree
make ARCH=x86_64

All errors (new ones prefixed by >>):

drivers/pci/pcie/err.c: In function 'report_error_detected':
>> drivers/pci/pcie/err.c:98:3: error: implicit declaration of function 'pci_uevent_ers'; did you mean 'pci_select_bars'? [-Werror=implicit-function-declaration]
pci_uevent_ers(dev, PCI_ERS_RESULT_NONE);
^~~~~~~~~~~~~~
pci_select_bars
drivers/pci/pcie/err.c: In function 'reset_link':
drivers/pci/pcie/err.c:203:5: warning: 'driver' is used uninitialized in this function [-Wuninitialized]
if (driver && driver->reset_link) {
^
cc1: some warnings being treated as errors

vim +98 drivers/pci/pcie/err.c

52
53 static int report_error_detected(struct pci_dev *dev, void *data)
54 {
55 pci_ers_result_t vote;
56 const struct pci_error_handlers *err_handler;
57 struct aer_broadcast_data *result_data;
58
59 result_data = (struct aer_broadcast_data *) data;
60
61 device_lock(&dev->dev);
62 dev->error_state = result_data->state;
63
64 if (!dev->driver ||
65 !dev->driver->err_handler ||
66 !dev->driver->err_handler->error_detected) {
67 if (result_data->state == pci_channel_io_frozen &&
68 dev->hdr_type != PCI_HEADER_TYPE_BRIDGE) {
69 /*
70 * In case of fatal recovery, if one of down-
71 * stream device has no driver. We might be
72 * unable to recover because a later insmod
73 * of a driver for this device is unaware of
74 * its hw state.
75 */
76 pci_printk(KERN_DEBUG, dev, "device has %s\n",
77 dev->driver ?
78 "no AER-aware driver" : "no driver");
79 }
80
81 /*
82 * If there's any device in the subtree that does not
83 * have an error_detected callback, returning
84 * PCI_ERS_RESULT_NO_AER_DRIVER prevents calling of
85 * the subsequent mmio_enabled/slot_reset/resume
86 * callbacks of "any" device in the subtree. All the
87 * devices in the subtree are left in the error state
88 * without recovery.
89 */
90
91 if (dev->hdr_type != PCI_HEADER_TYPE_BRIDGE)
92 vote = PCI_ERS_RESULT_NO_AER_DRIVER;
93 else
94 vote = PCI_ERS_RESULT_NONE;
95 } else {
96 err_handler = dev->driver->err_handler;
97 vote = err_handler->error_detected(dev, result_data->state);
> 98 pci_uevent_ers(dev, PCI_ERS_RESULT_NONE);
99 }
100
101 result_data->result = merge_result(result_data->result, vote);
102 device_unlock(&dev->dev);
103 return 0;
104 }
105

---
0-DAY kernel test infrastructure Open Source Technology Center
https://lists.01.org/pipermail/kbuild-all Intel Corporation

Attachments:

(No filename) (3.42 kB)
.config.gz (25.53 kB)
Download all attachments

2018-05-03 22:55:19

On 2018-05-10 18:52, Bjorn Helgaas wrote:
> On Thu, May 03, 2018 at 01:03:57AM -0400, Oza Pawandeep wrote:
>> Current DPC driver does not do recovery, e.g. calling end-point's
>> driver's
>> callbacks, which sanitize the sw.
>>
>> DPC driver implements link_reset callback, and calls
>> pci_do_recovery().
>>
>> Signed-off-by: Oza Pawandeep <[email protected]>
>>
>> diff --git a/drivers/pci/pcie/dpc.c b/drivers/pci/pcie/dpc.c
>> index 80ec384..aed7c9f 100644
>> --- a/drivers/pci/pcie/dpc.c
>> +++ b/drivers/pci/pcie/dpc.c
>> @@ -73,29 +73,21 @@ static void dpc_wait_link_inactive(struct dpc_dev
>> *dpc)
>> pcie_wait_for_link(pdev, false);
>> }
>>
>> -static void dpc_work(struct work_struct *work)
>> +static pci_ers_result_t dpc_reset_link(struct pci_dev *pdev)
>> {
>> - struct dpc_dev *dpc = container_of(work, struct dpc_dev, work);
>> - struct pci_dev *dev, *temp, *pdev = dpc->dev->port;
>> - struct pci_bus *parent = pdev->subordinate;
>> - u16 cap = dpc->cap_pos, ctl;
>> -
>> - pci_lock_rescan_remove();
>> - list_for_each_entry_safe_reverse(dev, temp, &parent->devices,
>> - bus_list) {
>> - pci_dev_get(dev);
>> - pci_dev_set_disconnected(dev, NULL);
>> - if (pci_has_subordinate(dev))
>> - pci_walk_bus(dev->subordinate,
>> - pci_dev_set_disconnected, NULL);
>> - pci_stop_and_remove_bus_device(dev);
>> - pci_dev_put(dev);
>> - }
>> - pci_unlock_rescan_remove();
>
> I think it would be good to have a comment here about why this
> "reset_link"
> function doesn't actually reset the link, e.g.,
>
> /*
> * DPC disables the Link automatically in hardware, so it has
> * already been reset by the time we get here.
> */
>
>> + struct dpc_dev *dpc;
>> + struct pcie_device *pciedev;
>> + struct device *devdpc;
>> + u16 cap, ctl;
>> +
>> + devdpc = pcie_port_find_device(pdev, PCIE_PORT_SERVICE_DPC);
>> + pciedev = to_pcie_device(devdpc);
>> + dpc = get_service_data(pciedev);
>> + cap = dpc->cap_pos;
>
> And maybe one about waiting until the link is inactive, then clearing
> DPC
> Trigger Status to allow the port to leave DPC.
>
>> dpc_wait_link_inactive(dpc);
>> if (dpc->rp_extensions && dpc_wait_rp_inactive(dpc))
>> - return;
>> + return PCI_ERS_RESULT_DISCONNECT;
>> if (dpc->rp_extensions && dpc->rp_pio_status) {
>> pci_write_config_dword(pdev, cap + PCI_EXP_DPC_RP_PIO_STATUS,
>> dpc->rp_pio_status);
>> @@ -108,6 +100,17 @@ static void dpc_work(struct work_struct *work)
>> pci_read_config_word(pdev, cap + PCI_EXP_DPC_CTL, &ctl);
>> pci_write_config_word(pdev, cap + PCI_EXP_DPC_CTL,
>> ctl | PCI_EXP_DPC_CTL_INT_EN);
>> +
>> + return PCI_ERS_RESULT_RECOVERED;
>> +}
>> +
>> +static void dpc_work(struct work_struct *work)
>> +{
>> + struct dpc_dev *dpc = container_of(work, struct dpc_dev, work);
>> + struct pci_dev *pdev = dpc->dev->port;
>> +
>> + /* From DPC point of view error is always FATAL. */
>> + pcie_do_recovery(pdev, DPC_FATAL);
>> }
>>
>> static void dpc_process_rp_pio_error(struct dpc_dev *dpc)
>> @@ -288,6 +291,7 @@ static struct pcie_port_service_driver dpcdriver =
>> {
>> .service = PCIE_PORT_SERVICE_DPC,
>> .probe = dpc_probe,
>> .remove = dpc_remove,
>> + .reset_link = dpc_reset_link,
>> };
>>
>> static int __init dpc_service_init(void)
>> diff --git a/drivers/pci/pcie/err.c b/drivers/pci/pcie/err.c
>> index 877785d..526aba8 100644
>> --- a/drivers/pci/pcie/err.c
>> +++ b/drivers/pci/pcie/err.c
>> @@ -181,11 +181,12 @@ static pci_ers_result_t
>> default_reset_link(struct pci_dev *dev)
>> return PCI_ERS_RESULT_RECOVERED;
>> }
>>
>> -static pci_ers_result_t reset_link(struct pci_dev *dev)
>> +static pci_ers_result_t reset_link(struct pci_dev *dev, int severity)
>> {
>> struct pci_dev *udev;
>> pci_ers_result_t status;
>> struct pcie_port_service_driver *driver;
>> + u32 service;
>>
>> if (dev->hdr_type == PCI_HEADER_TYPE_BRIDGE) {
>> /* Reset this port for all subordinates */
>> @@ -196,7 +197,12 @@ static pci_ers_result_t reset_link(struct pci_dev
>> *dev)
>> }
>>
>> /* Use the aer driver of the component firstly */
>> - driver = pcie_port_find_service(udev, PCIE_PORT_SERVICE_AER);
>> + if (severity == DPC_FATAL)
>> + service = PCIE_PORT_SERVICE_DPC;
>> + else
>> + service = PCIE_PORT_SERVICE_AER;
>> +
>> + driver = pcie_port_find_service(udev, service);
>
> This is where I was wondering about passing in "service" directly
> instead
> of "severity".

I will take care of all your comments made on all the patches.
but this one I do not understand.

passing service directly instead of severity ? (I do not think you meant
to write severity there)
perhaps do you mean following ?

if (severity == DPC_FATAL)
pcie_port_find_service(udev, PCIE_PORT_SERVICE_DPC);
else
pcie_port_find_service(udev, PCIE_PORT_SERVICE_AER);

>
>> if (driver && driver->reset_link) {
>> status = driver->reset_link(udev);
>> @@ -302,7 +308,7 @@ static pci_ers_result_t do_fatal_recovery(struct
>> pci_dev *dev, int severity)
>> pci_dev_put(pdev);
>> }
>>
>> - result = reset_link(udev);
>> + result = reset_link(udev, severity);
>> if (result == PCI_ERS_RESULT_RECOVERED)
>> if (pcie_wait_for_link(udev, true))
>> pci_rescan_bus(udev->bus);
>> @@ -326,7 +332,8 @@ void pcie_do_recovery(struct pci_dev *dev, int
>> severity)
>> pci_ers_result_t status;
>> enum pci_channel_state state;
>>
>> - if (severity == AER_FATAL) {
>> + if ((severity == AER_FATAL) ||
>> + (severity == DPC_FATAL)) {
>> status = do_fatal_recovery(dev, severity);
>> if (status != PCI_ERS_RESULT_RECOVERED)
>> goto failed;
>> diff --git a/include/linux/aer.h b/include/linux/aer.h
>> index 8f87bbe..0c506fe 100644
>> --- a/include/linux/aer.h
>> +++ b/include/linux/aer.h
>> @@ -14,6 +14,7 @@
>> #define AER_NONFATAL 0
>> #define AER_FATAL 1
>> #define AER_CORRECTABLE 2
>> +#define DPC_FATAL 4
>>
>> struct pci_dev;
>>
>> --
>> 2.7.4
>>

2018-05-10 16:30:06

by Bjorn Helgaas

[permalink] [raw]

Subject: Re: [PATCH v15 8/9] PCI/DPC: Unify and plumb error handling into DPC

On Thu, May 10, 2018 at 07:56:00PM +0530, [email protected] wrote:
> On 2018-05-10 18:52, Bjorn Helgaas wrote:
> > On Thu, May 03, 2018 at 01:03:57AM -0400, Oza Pawandeep wrote:
> ...

> > > -static pci_ers_result_t reset_link(struct pci_dev *dev)
> > > +static pci_ers_result_t reset_link(struct pci_dev *dev, int severity)
> > > {
> > > struct pci_dev *udev;
> > > pci_ers_result_t status;
> > > struct pcie_port_service_driver *driver;
> > > + u32 service;
> > >
> > > if (dev->hdr_type == PCI_HEADER_TYPE_BRIDGE) {
> > > /* Reset this port for all subordinates */
> > > @@ -196,7 +197,12 @@ static pci_ers_result_t reset_link(struct
> > > pci_dev *dev)
> > > }
> > >
> > > /* Use the aer driver of the component firstly */
> > > - driver = pcie_port_find_service(udev, PCIE_PORT_SERVICE_AER);
> > > + if (severity == DPC_FATAL)
> > > + service = PCIE_PORT_SERVICE_DPC;
> > > + else
> > > + service = PCIE_PORT_SERVICE_AER;
> > > +
> > > + driver = pcie_port_find_service(udev, service);
> >
> > This is where I was wondering about passing in "service" directly
> > instead
> > of "severity".
>
> passing service directly instead of severity ? (I do not think you meant to
> write severity there)

I did mean "severity".

> perhaps do you mean following ?
>
> if (severity == DPC_FATAL)
> pcie_port_find_service(udev, PCIE_PORT_SERVICE_DPC);
> else
> pcie_port_find_service(udev, PCIE_PORT_SERVICE_AER);

No, my thought was that most of the places that use "severity" only
use it to decide between PCIE_PORT_SERVICE_DPC and
PCIE_PORT_SERVICE_AER, so if you just passed in "service" directly,
you wouldn't need this "if" statement.

Bjorn