Skip to content

Commit 760554a

Browse files
committed
Merge branch 'pds_core-pci-reset'
Shannon Nelson says: ==================== pds_core: add PCI reset handling Make sure pds_core can handle and recover from PCI function resets and similar PCI bus issues: add detection and handlers for PCI problems. ==================== Signed-off-by: David S. Miller <davem@davemloft.net>
2 parents 8d6198a + 1e18ec3 commit 760554a

File tree

5 files changed

+101
-11
lines changed

5 files changed

+101
-11
lines changed

drivers/net/ethernet/amd/pds_core/core.c

Lines changed: 34 additions & 9 deletions
Original file line numberDiff line numberDiff line change
@@ -445,12 +445,13 @@ int pdsc_setup(struct pdsc *pdsc, bool init)
445445
goto err_out_teardown;
446446

447447
/* Set up the VIFs */
448-
err = pdsc_viftypes_init(pdsc);
449-
if (err)
450-
goto err_out_teardown;
448+
if (init) {
449+
err = pdsc_viftypes_init(pdsc);
450+
if (err)
451+
goto err_out_teardown;
451452

452-
if (init)
453453
pdsc_debugfs_add_viftype(pdsc);
454+
}
454455

455456
clear_bit(PDSC_S_FW_DEAD, &pdsc->state);
456457
return 0;
@@ -469,8 +470,10 @@ void pdsc_teardown(struct pdsc *pdsc, bool removing)
469470
pdsc_qcq_free(pdsc, &pdsc->notifyqcq);
470471
pdsc_qcq_free(pdsc, &pdsc->adminqcq);
471472

472-
kfree(pdsc->viftype_status);
473-
pdsc->viftype_status = NULL;
473+
if (removing) {
474+
kfree(pdsc->viftype_status);
475+
pdsc->viftype_status = NULL;
476+
}
474477

475478
if (pdsc->intr_info) {
476479
for (i = 0; i < pdsc->nintrs; i++)
@@ -512,18 +515,21 @@ void pdsc_stop(struct pdsc *pdsc)
512515
PDS_CORE_INTR_MASK_SET);
513516
}
514517

515-
static void pdsc_fw_down(struct pdsc *pdsc)
518+
void pdsc_fw_down(struct pdsc *pdsc)
516519
{
517520
union pds_core_notifyq_comp reset_event = {
518521
.reset.ecode = cpu_to_le16(PDS_EVENT_RESET),
519522
.reset.state = 0,
520523
};
521524

522525
if (test_and_set_bit(PDSC_S_FW_DEAD, &pdsc->state)) {
523-
dev_err(pdsc->dev, "%s: already happening\n", __func__);
526+
dev_warn(pdsc->dev, "%s: already happening\n", __func__);
524527
return;
525528
}
526529

530+
if (pdsc->pdev->is_virtfn)
531+
return;
532+
527533
/* Notify clients of fw_down */
528534
if (pdsc->fw_reporter)
529535
devlink_health_report(pdsc->fw_reporter, "FW down reported", pdsc);
@@ -533,7 +539,7 @@ static void pdsc_fw_down(struct pdsc *pdsc)
533539
pdsc_teardown(pdsc, PDSC_TEARDOWN_RECOVERY);
534540
}
535541

536-
static void pdsc_fw_up(struct pdsc *pdsc)
542+
void pdsc_fw_up(struct pdsc *pdsc)
537543
{
538544
union pds_core_notifyq_comp reset_event = {
539545
.reset.ecode = cpu_to_le16(PDS_EVENT_RESET),
@@ -546,6 +552,11 @@ static void pdsc_fw_up(struct pdsc *pdsc)
546552
return;
547553
}
548554

555+
if (pdsc->pdev->is_virtfn) {
556+
clear_bit(PDSC_S_FW_DEAD, &pdsc->state);
557+
return;
558+
}
559+
549560
err = pdsc_setup(pdsc, PDSC_SETUP_RECOVERY);
550561
if (err)
551562
goto err_out;
@@ -567,6 +578,18 @@ static void pdsc_fw_up(struct pdsc *pdsc)
567578
pdsc_teardown(pdsc, PDSC_TEARDOWN_RECOVERY);
568579
}
569580

581+
static void pdsc_check_pci_health(struct pdsc *pdsc)
582+
{
583+
u8 fw_status = ioread8(&pdsc->info_regs->fw_status);
584+
585+
/* is PCI broken? */
586+
if (fw_status != PDS_RC_BAD_PCI)
587+
return;
588+
589+
pdsc_reset_prepare(pdsc->pdev);
590+
pdsc_reset_done(pdsc->pdev);
591+
}
592+
570593
void pdsc_health_thread(struct work_struct *work)
571594
{
572595
struct pdsc *pdsc = container_of(work, struct pdsc, health_work);
@@ -593,6 +616,8 @@ void pdsc_health_thread(struct work_struct *work)
593616
pdsc_fw_down(pdsc);
594617
}
595618

619+
pdsc_check_pci_health(pdsc);
620+
596621
pdsc->fw_generation = pdsc->fw_status & PDS_CORE_FW_STS_F_GENERATION;
597622

598623
out_unlock:

drivers/net/ethernet/amd/pds_core/core.h

Lines changed: 7 additions & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -283,6 +283,9 @@ int pdsc_devcmd_reset(struct pdsc *pdsc);
283283
int pdsc_dev_reinit(struct pdsc *pdsc);
284284
int pdsc_dev_init(struct pdsc *pdsc);
285285

286+
void pdsc_reset_prepare(struct pci_dev *pdev);
287+
void pdsc_reset_done(struct pci_dev *pdev);
288+
286289
int pdsc_intr_alloc(struct pdsc *pdsc, char *name,
287290
irq_handler_t handler, void *data);
288291
void pdsc_intr_free(struct pdsc *pdsc, int index);
@@ -309,4 +312,8 @@ irqreturn_t pdsc_adminq_isr(int irq, void *data);
309312

310313
int pdsc_firmware_update(struct pdsc *pdsc, const struct firmware *fw,
311314
struct netlink_ext_ack *extack);
315+
316+
void pdsc_fw_down(struct pdsc *pdsc);
317+
void pdsc_fw_up(struct pdsc *pdsc);
318+
312319
#endif /* _PDSC_H_ */

drivers/net/ethernet/amd/pds_core/dev.c

Lines changed: 9 additions & 2 deletions
Original file line numberDiff line numberDiff line change
@@ -42,6 +42,8 @@ int pdsc_err_to_errno(enum pds_core_status_code code)
4242
return -ERANGE;
4343
case PDS_RC_BAD_ADDR:
4444
return -EFAULT;
45+
case PDS_RC_BAD_PCI:
46+
return -ENXIO;
4547
case PDS_RC_EOPCODE:
4648
case PDS_RC_EINTR:
4749
case PDS_RC_DEV_CMD:
@@ -62,7 +64,7 @@ bool pdsc_is_fw_running(struct pdsc *pdsc)
6264
/* Firmware is useful only if the running bit is set and
6365
* fw_status != 0xff (bad PCI read)
6466
*/
65-
return (pdsc->fw_status != 0xff) &&
67+
return (pdsc->fw_status != PDS_RC_BAD_PCI) &&
6668
(pdsc->fw_status & PDS_CORE_FW_STS_F_RUNNING);
6769
}
6870

@@ -128,6 +130,7 @@ static int pdsc_devcmd_wait(struct pdsc *pdsc, u8 opcode, int max_seconds)
128130
unsigned long max_wait;
129131
unsigned long duration;
130132
int timeout = 0;
133+
bool running;
131134
int done = 0;
132135
int err = 0;
133136
int status;
@@ -136,6 +139,10 @@ static int pdsc_devcmd_wait(struct pdsc *pdsc, u8 opcode, int max_seconds)
136139
max_wait = start_time + (max_seconds * HZ);
137140

138141
while (!done && !timeout) {
142+
running = pdsc_is_fw_running(pdsc);
143+
if (!running)
144+
break;
145+
139146
done = pdsc_devcmd_done(pdsc);
140147
if (done)
141148
break;
@@ -152,7 +159,7 @@ static int pdsc_devcmd_wait(struct pdsc *pdsc, u8 opcode, int max_seconds)
152159
dev_dbg(dev, "DEVCMD %d %s after %ld secs\n",
153160
opcode, pdsc_devcmd_str(opcode), duration / HZ);
154161

155-
if (!done || timeout) {
162+
if ((!done || timeout) && running) {
156163
dev_err(dev, "DEVCMD %d %s timeout, done %d timeout %d max_seconds=%d\n",
157164
opcode, pdsc_devcmd_str(opcode), done, timeout,
158165
max_seconds);

drivers/net/ethernet/amd/pds_core/main.c

Lines changed: 50 additions & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -445,12 +445,62 @@ static void pdsc_remove(struct pci_dev *pdev)
445445
devlink_free(dl);
446446
}
447447

448+
void pdsc_reset_prepare(struct pci_dev *pdev)
449+
{
450+
struct pdsc *pdsc = pci_get_drvdata(pdev);
451+
452+
pdsc_fw_down(pdsc);
453+
454+
pci_free_irq_vectors(pdev);
455+
pdsc_unmap_bars(pdsc);
456+
pci_release_regions(pdev);
457+
pci_disable_device(pdev);
458+
}
459+
460+
void pdsc_reset_done(struct pci_dev *pdev)
461+
{
462+
struct pdsc *pdsc = pci_get_drvdata(pdev);
463+
struct device *dev = pdsc->dev;
464+
int err;
465+
466+
err = pci_enable_device(pdev);
467+
if (err) {
468+
dev_err(dev, "Cannot enable PCI device: %pe\n", ERR_PTR(err));
469+
return;
470+
}
471+
pci_set_master(pdev);
472+
473+
if (!pdev->is_virtfn) {
474+
pcie_print_link_status(pdsc->pdev);
475+
476+
err = pci_request_regions(pdsc->pdev, PDS_CORE_DRV_NAME);
477+
if (err) {
478+
dev_err(pdsc->dev, "Cannot request PCI regions: %pe\n",
479+
ERR_PTR(err));
480+
return;
481+
}
482+
483+
err = pdsc_map_bars(pdsc);
484+
if (err)
485+
return;
486+
}
487+
488+
pdsc_fw_up(pdsc);
489+
}
490+
491+
static const struct pci_error_handlers pdsc_err_handler = {
492+
/* FLR handling */
493+
.reset_prepare = pdsc_reset_prepare,
494+
.reset_done = pdsc_reset_done,
495+
};
496+
448497
static struct pci_driver pdsc_driver = {
449498
.name = PDS_CORE_DRV_NAME,
450499
.id_table = pdsc_id_table,
451500
.probe = pdsc_probe,
452501
.remove = pdsc_remove,
453502
.sriov_configure = pdsc_sriov_configure,
503+
.err_handler = &pdsc_err_handler,
454504
};
455505

456506
void *pdsc_get_pf_struct(struct pci_dev *vf_pdev)

include/linux/pds/pds_core_if.h

Lines changed: 1 addition & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -79,6 +79,7 @@ enum pds_core_status_code {
7979
PDS_RC_EVFID = 31, /* VF ID does not exist */
8080
PDS_RC_BAD_FW = 32, /* FW file is invalid or corrupted */
8181
PDS_RC_ECLIENT = 33, /* No such client id */
82+
PDS_RC_BAD_PCI = 255, /* Broken PCI when reading status */
8283
};
8384

8485
/**

0 commit comments

Comments
 (0)