IB/mthca: Recover from catastrophic errors

Trigger device remove and then add when a catastrophic error is
detected in hardware.  This, in turn, will cause a device reset, which
we hope will recover from the catastrophic condition.

Since this might interefere with debugging the root cause, add a
module option to suppress this behaviour.

Signed-off-by: Jack Morgenstein <jackm@mellanox.co.il>
Signed-off-by: Michael S. Tsirkin <mst@mellanox.co.il>
Signed-off-by: Roland Dreier <rolandd@cisco.com>
This commit is contained in:
Jack Morgenstein 2006-08-15 21:11:18 +03:00 committed by Roland Dreier
parent 07eeec0627
commit b3b30f5e8a
3 changed files with 136 additions and 21 deletions

View File

@ -34,6 +34,7 @@
#include <linux/jiffies.h>
#include <linux/timer.h>
#include <linux/workqueue.h>
#include "mthca_dev.h"
@ -48,9 +49,41 @@ enum {
static DEFINE_SPINLOCK(catas_lock);
static LIST_HEAD(catas_list);
static struct workqueue_struct *catas_wq;
static struct work_struct catas_work;
static int catas_reset_disable;
module_param_named(catas_reset_disable, catas_reset_disable, int, 0644);
MODULE_PARM_DESC(catas_reset_disable, "disable reset on catastrophic event if nonzero");
static void catas_reset(void *work_ptr)
{
struct mthca_dev *dev, *tmpdev;
LIST_HEAD(tlist);
int ret;
mutex_lock(&mthca_device_mutex);
spin_lock_irq(&catas_lock);
list_splice_init(&catas_list, &tlist);
spin_unlock_irq(&catas_lock);
list_for_each_entry_safe(dev, tmpdev, &tlist, catas_err.list) {
ret = __mthca_restart_one(dev->pdev);
if (ret)
mthca_err(dev, "Reset failed (%d)\n", ret);
else
mthca_dbg(dev, "Reset succeeded\n");
}
mutex_unlock(&mthca_device_mutex);
}
static void handle_catas(struct mthca_dev *dev)
{
struct ib_event event;
unsigned long flags;
const char *type;
int i;
@ -82,6 +115,14 @@ static void handle_catas(struct mthca_dev *dev)
for (i = 0; i < dev->catas_err.size; ++i)
mthca_err(dev, " buf[%02x]: %08x\n",
i, swab32(readl(dev->catas_err.map + i)));
if (catas_reset_disable)
return;
spin_lock_irqsave(&catas_lock, flags);
list_add(&dev->catas_err.list, &catas_list);
queue_work(catas_wq, &catas_work);
spin_unlock_irqrestore(&catas_lock, flags);
}
static void poll_catas(unsigned long dev_ptr)
@ -135,6 +176,7 @@ void mthca_start_catas_poll(struct mthca_dev *dev)
dev->catas_err.timer.data = (unsigned long) dev;
dev->catas_err.timer.function = poll_catas;
dev->catas_err.timer.expires = jiffies + MTHCA_CATAS_POLL_INTERVAL;
INIT_LIST_HEAD(&dev->catas_err.list);
add_timer(&dev->catas_err.timer);
}
@ -153,4 +195,24 @@ void mthca_stop_catas_poll(struct mthca_dev *dev)
dev->catas_err.addr),
dev->catas_err.size * 4);
}
spin_lock_irq(&catas_lock);
list_del(&dev->catas_err.list);
spin_unlock_irq(&catas_lock);
}
int __init mthca_catas_init(void)
{
INIT_WORK(&catas_work, catas_reset, NULL);
catas_wq = create_singlethread_workqueue("mthca_catas");
if (!catas_wq)
return -ENOMEM;
return 0;
}
void mthca_catas_cleanup(void)
{
destroy_workqueue(catas_wq);
}

View File

@ -45,6 +45,7 @@
#include <linux/dma-mapping.h>
#include <linux/timer.h>
#include <linux/mutex.h>
#include <linux/list.h>
#include <asm/semaphore.h>
@ -283,8 +284,11 @@ struct mthca_catas_err {
unsigned long stop;
u32 size;
struct timer_list timer;
struct list_head list;
};
extern struct mutex mthca_device_mutex;
struct mthca_dev {
struct ib_device ib_dev;
struct pci_dev *pdev;
@ -450,6 +454,9 @@ void mthca_unregister_device(struct mthca_dev *dev);
void mthca_start_catas_poll(struct mthca_dev *dev);
void mthca_stop_catas_poll(struct mthca_dev *dev);
int __mthca_restart_one(struct pci_dev *pdev);
int mthca_catas_init(void);
void mthca_catas_cleanup(void);
int mthca_uar_alloc(struct mthca_dev *dev, struct mthca_uar *uar);
void mthca_uar_free(struct mthca_dev *dev, struct mthca_uar *uar);

View File

@ -80,6 +80,8 @@ static int tune_pci = 0;
module_param(tune_pci, int, 0444);
MODULE_PARM_DESC(tune_pci, "increase PCI burst from the default set by BIOS if nonzero");
struct mutex mthca_device_mutex;
static const char mthca_version[] __devinitdata =
DRV_NAME ": Mellanox InfiniBand HCA driver v"
DRV_VERSION " (" DRV_RELDATE ")\n";
@ -978,28 +980,15 @@ static struct {
MTHCA_FLAG_SINAI_OPT }
};
static int __devinit mthca_init_one(struct pci_dev *pdev,
const struct pci_device_id *id)
static int __mthca_init_one(struct pci_dev *pdev, int hca_type)
{
static int mthca_version_printed = 0;
int ddr_hidden = 0;
int err;
struct mthca_dev *mdev;
if (!mthca_version_printed) {
printk(KERN_INFO "%s", mthca_version);
++mthca_version_printed;
}
printk(KERN_INFO PFX "Initializing %s\n",
pci_name(pdev));
if (id->driver_data >= ARRAY_SIZE(mthca_hca_table)) {
printk(KERN_ERR PFX "%s has invalid driver data %lx\n",
pci_name(pdev), id->driver_data);
return -ENODEV;
}
err = pci_enable_device(pdev);
if (err) {
dev_err(&pdev->dev, "Cannot enable PCI device, "
@ -1065,7 +1054,7 @@ static int __devinit mthca_init_one(struct pci_dev *pdev,
mdev->pdev = pdev;
mdev->mthca_flags = mthca_hca_table[id->driver_data].flags;
mdev->mthca_flags = mthca_hca_table[hca_type].flags;
if (ddr_hidden)
mdev->mthca_flags |= MTHCA_FLAG_DDR_HIDDEN;
@ -1099,13 +1088,13 @@ static int __devinit mthca_init_one(struct pci_dev *pdev,
if (err)
goto err_cmd;
if (mdev->fw_ver < mthca_hca_table[id->driver_data].latest_fw) {
if (mdev->fw_ver < mthca_hca_table[hca_type].latest_fw) {
mthca_warn(mdev, "HCA FW version %d.%d.%d is old (%d.%d.%d is current).\n",
(int) (mdev->fw_ver >> 32), (int) (mdev->fw_ver >> 16) & 0xffff,
(int) (mdev->fw_ver & 0xffff),
(int) (mthca_hca_table[id->driver_data].latest_fw >> 32),
(int) (mthca_hca_table[id->driver_data].latest_fw >> 16) & 0xffff,
(int) (mthca_hca_table[id->driver_data].latest_fw & 0xffff));
(int) (mthca_hca_table[hca_type].latest_fw >> 32),
(int) (mthca_hca_table[hca_type].latest_fw >> 16) & 0xffff,
(int) (mthca_hca_table[hca_type].latest_fw & 0xffff));
mthca_warn(mdev, "If you have problems, try updating your HCA FW.\n");
}
@ -1122,6 +1111,7 @@ static int __devinit mthca_init_one(struct pci_dev *pdev,
goto err_unregister;
pci_set_drvdata(pdev, mdev);
mdev->hca_type = hca_type;
return 0;
@ -1166,7 +1156,7 @@ err_disable_pdev:
return err;
}
static void __devexit mthca_remove_one(struct pci_dev *pdev)
static void __mthca_remove_one(struct pci_dev *pdev)
{
struct mthca_dev *mdev = pci_get_drvdata(pdev);
u8 status;
@ -1211,6 +1201,51 @@ static void __devexit mthca_remove_one(struct pci_dev *pdev)
}
}
int __mthca_restart_one(struct pci_dev *pdev)
{
struct mthca_dev *mdev;
mdev = pci_get_drvdata(pdev);
if (!mdev)
return -ENODEV;
__mthca_remove_one(pdev);
return __mthca_init_one(pdev, mdev->hca_type);
}
static int __devinit mthca_init_one(struct pci_dev *pdev,
const struct pci_device_id *id)
{
static int mthca_version_printed = 0;
int ret;
mutex_lock(&mthca_device_mutex);
if (!mthca_version_printed) {
printk(KERN_INFO "%s", mthca_version);
++mthca_version_printed;
}
if (id->driver_data >= ARRAY_SIZE(mthca_hca_table)) {
printk(KERN_ERR PFX "%s has invalid driver data %lx\n",
pci_name(pdev), id->driver_data);
mutex_unlock(&mthca_device_mutex);
return -ENODEV;
}
ret = __mthca_init_one(pdev, id->driver_data);
mutex_unlock(&mthca_device_mutex);
return ret;
}
static void __devexit mthca_remove_one(struct pci_dev *pdev)
{
mutex_lock(&mthca_device_mutex);
__mthca_remove_one(pdev);
mutex_unlock(&mthca_device_mutex);
}
static struct pci_device_id mthca_pci_table[] = {
{ PCI_DEVICE(PCI_VENDOR_ID_MELLANOX, PCI_DEVICE_ID_MELLANOX_TAVOR),
.driver_data = TAVOR },
@ -1248,13 +1283,24 @@ static int __init mthca_init(void)
{
int ret;
mutex_init(&mthca_device_mutex);
ret = mthca_catas_init();
if (ret)
return ret;
ret = pci_register_driver(&mthca_driver);
return ret < 0 ? ret : 0;
if (ret < 0) {
mthca_catas_cleanup();
return ret;
}
return 0;
}
static void __exit mthca_cleanup(void)
{
pci_unregister_driver(&mthca_driver);
mthca_catas_cleanup();
}
module_init(mthca_init);