-------------------------------diff file-----------------------------------------
diff --git a/drivers/pci/hotplug/pciehp_hpc.c b/drivers/pci/hotplug/pciehp_hpc.c
index bb5a8d9f03ad..c9f3ed86a084 100644
--- a/drivers/pci/hotplug/pciehp_hpc.c
+++ b/drivers/pci/hotplug/pciehp_hpc.c
@@ -700,6 +700,7 @@ static irqreturn_t pciehp_ist(int irq, void *dev_id)
irqreturn_t ret;
u32 events;
+ printk("==== %s %d start running\n", __func__, __LINE__);
ctrl->ist_running = true;
pci_config_pm_runtime_get(pdev);
@@ -755,6 +756,7 @@ static irqreturn_t pciehp_ist(int irq, void *dev_id)
pci_config_pm_runtime_put(pdev);
ctrl->ist_running = false;
wake_up(&ctrl->requester);
+ printk("==== %s %d stop running\n", __func__, __LINE__);
return ret;
}
diff --git a/drivers/pci/pci.c b/drivers/pci/pci.c
index 661f98c6c63a..ffa58f389456 100644
--- a/drivers/pci/pci.c
+++ b/drivers/pci/pci.c
@@ -4784,6 +4784,7 @@ static bool pcie_wait_for_link_delay(struct pci_dev *pdev, bool active,
if (active)
msleep(20);
rc = pcie_wait_for_link_status(pdev, false, active);
+ printk("======%s %d,wait for linksta:%d\n", __func__, __LINE__, rc);
if (active) {
if (rc)
rc = pcie_failed_link_retrain(pdev);
diff --git a/drivers/pci/pci.h b/drivers/pci/pci.h
index 2e40fc63ba31..b7e5af859517 100644
--- a/drivers/pci/pci.h
+++ b/drivers/pci/pci.h
@@ -337,12 +337,13 @@ void pci_bus_put(struct pci_bus *bus);
#define PCIE_LNKCAP_SLS2SPEED(lnkcap) \
({ \
- ((lnkcap) == PCI_EXP_LNKCAP_SLS_64_0GB ? PCIE_SPEED_64_0GT : \
- (lnkcap) == PCI_EXP_LNKCAP_SLS_32_0GB ? PCIE_SPEED_32_0GT : \
- (lnkcap) == PCI_EXP_LNKCAP_SLS_16_0GB ? PCIE_SPEED_16_0GT : \
- (lnkcap) == PCI_EXP_LNKCAP_SLS_8_0GB ? PCIE_SPEED_8_0GT : \
- (lnkcap) == PCI_EXP_LNKCAP_SLS_5_0GB ? PCIE_SPEED_5_0GT : \
- (lnkcap) == PCI_EXP_LNKCAP_SLS_2_5GB ? PCIE_SPEED_2_5GT : \
+ u32 __lnkcap = (lnkcap) & PCI_EXP_LNKCAP_SLS; \
+ (__lnkcap == PCI_EXP_LNKCAP_SLS_64_0GB ? PCIE_SPEED_64_0GT : \
+ __lnkcap == PCI_EXP_LNKCAP_SLS_32_0GB ? PCIE_SPEED_32_0GT : \
+ __lnkcap == PCI_EXP_LNKCAP_SLS_16_0GB ? PCIE_SPEED_16_0GT : \
+ __lnkcap == PCI_EXP_LNKCAP_SLS_8_0GB ? PCIE_SPEED_8_0GT : \
+ __lnkcap == PCI_EXP_LNKCAP_SLS_5_0GB ? PCIE_SPEED_5_0GT : \
+ __lnkcap == PCI_EXP_LNKCAP_SLS_2_5GB ? PCIE_SPEED_2_5GT : \
PCI_SPEED_UNKNOWN); \
})
@@ -357,13 +358,16 @@ void pci_bus_put(struct pci_bus *bus);
PCI_SPEED_UNKNOWN)
#define PCIE_LNKCTL2_TLS2SPEED(lnkctl2) \
- ((lnkctl2) == PCI_EXP_LNKCTL2_TLS_64_0GT ? PCIE_SPEED_64_0GT : \
- (lnkctl2) == PCI_EXP_LNKCTL2_TLS_32_0GT ? PCIE_SPEED_32_0GT : \
- (lnkctl2) == PCI_EXP_LNKCTL2_TLS_16_0GT ? PCIE_SPEED_16_0GT : \
- (lnkctl2) == PCI_EXP_LNKCTL2_TLS_8_0GT ? PCIE_SPEED_8_0GT : \
- (lnkctl2) == PCI_EXP_LNKCTL2_TLS_5_0GT ? PCIE_SPEED_5_0GT : \
- (lnkctl2) == PCI_EXP_LNKCTL2_TLS_2_5GT ? PCIE_SPEED_2_5GT : \
- PCI_SPEED_UNKNOWN)
+({ \
+ u16 __lnkctl2 = (lnkctl2) & PCI_EXP_LNKCTL2_TLS; \
+ (__lnkctl2 == PCI_EXP_LNKCTL2_TLS_64_0GT ? PCIE_SPEED_64_0GT : \
+ __lnkctl2 == PCI_EXP_LNKCTL2_TLS_32_0GT ? PCIE_SPEED_32_0GT : \
+ __lnkctl2 == PCI_EXP_LNKCTL2_TLS_16_0GT ? PCIE_SPEED_16_0GT : \
+ __lnkctl2 == PCI_EXP_LNKCTL2_TLS_8_0GT ? PCIE_SPEED_8_0GT : \
+ __lnkctl2 == PCI_EXP_LNKCTL2_TLS_5_0GT ? PCIE_SPEED_5_0GT : \
+ __lnkctl2 == PCI_EXP_LNKCTL2_TLS_2_5GT ? PCIE_SPEED_2_5GT : \
+ PCI_SPEED_UNKNOWN); \
+})
/* PCIe speed to Mb/s reduced by encoding overhead */
#define PCIE_SPEED2MBS_ENC(speed) \
diff --git a/drivers/pci/pcie/bwctrl.c b/drivers/pci/pcie/bwctrl.c
index b59cacc740fa..a8ce09f67d3b 100644
--- a/drivers/pci/pcie/bwctrl.c
+++ b/drivers/pci/pcie/bwctrl.c
@@ -168,8 +168,10 @@ int pcie_set_target_speed(struct pci_dev *port, enum pci_bus_speed speed_req,
if (WARN_ON_ONCE(!pcie_valid_speed(speed_req)))
return -EINVAL;
- if (bus && bus->cur_bus_speed == speed_req)
+ if (bus && bus->cur_bus_speed == speed_req) {
+ printk("========== %s %d, speed has been set\n", __func__, __LINE__);
return 0;
+ }
target_speed = pcie_bwctrl_select_speed(port, speed_req);
@@ -184,6 +186,7 @@ int pcie_set_target_speed(struct pci_dev *port, enum pci_bus_speed speed_req,
mutex_lock(&data->set_speed_mutex);
ret = pcie_bwctrl_change_speed(port, target_speed, use_lt);
+ printk("========== %s %d, bwctl change speed ret:0x%x\n", __func__, __LINE__,ret);
if (data)
mutex_unlock(&data->set_speed_mutex);
@@ -209,8 +212,10 @@ static void pcie_bwnotif_enable(struct pcie_device *srv)
/* Count LBMS seen so far as one */
ret = pcie_capability_read_word(port, PCI_EXP_LNKSTA, &link_status);
- if (ret == PCIBIOS_SUCCESSFUL && link_status & PCI_EXP_LNKSTA_LBMS)
+ if (ret == PCIBIOS_SUCCESSFUL && link_status & PCI_EXP_LNKSTA_LBMS) {
+ printk("==== %s %d lbms_count++\n", __func__, __LINE__);
atomic_inc(&data->lbms_count);
+ }
pcie_capability_set_word(port, PCI_EXP_LNKCTL,
PCI_EXP_LNKCTL_LBMIE | PCI_EXP_LNKCTL_LABIE);
@@ -239,6 +244,7 @@ static irqreturn_t pcie_bwnotif_irq(int irq, void *context)
int ret;
ret = pcie_capability_read_word(port, PCI_EXP_LNKSTA, &link_status);
+ printk("==== %s %d(start running),link_status:0x%x\n", __func__, __LINE__,link_status);
if (ret != PCIBIOS_SUCCESSFUL)
return IRQ_NONE;
@@ -246,8 +252,10 @@ static irqreturn_t pcie_bwnotif_irq(int irq, void *context)
if (!events)
return IRQ_NONE;
- if (events & PCI_EXP_LNKSTA_LBMS)
+ if (events & PCI_EXP_LNKSTA_LBMS) {
+ printk("==== %s %d lbms_count++\n", __func__, __LINE__);
atomic_inc(&data->lbms_count);
+ }
pcie_capability_write_word(port, PCI_EXP_LNKSTA, events);
@@ -258,6 +266,7 @@ static irqreturn_t pcie_bwnotif_irq(int irq, void *context)
* cleared to avoid missing link speed changes.
*/
pcie_update_link_speed(port->subordinate);
+ printk("==== %s %d(stop running),link_status:0x%x\n", __func__, __LINE__,link_status);
return IRQ_HANDLED;
}
@@ -268,8 +277,10 @@ void pcie_reset_lbms_count(struct pci_dev *port)
guard(rwsem_read)(&pcie_bwctrl_lbms_rwsem);
data = port->link_bwctrl;
- if (data)
+ if (data) {
+ printk("==== %s %d lbms_count set to 0\n", __func__, __LINE__);
atomic_set(&data->lbms_count, 0);
+ }
else
pcie_capability_write_word(port, PCI_EXP_LNKSTA,
PCI_EXP_LNKSTA_LBMS);
diff --git a/drivers/pci/quirks.c b/drivers/pci/quirks.c
index 76f4df75b08a..a602f9aa5d6a 100644
--- a/drivers/pci/quirks.c
+++ b/drivers/pci/quirks.c
@@ -41,8 +41,11 @@ static bool pcie_lbms_seen(struct pci_dev *dev, u16 lnksta)
int ret;
ret = pcie_lbms_count(dev, &count);
- if (ret < 0)
+ if (ret < 0) {
+ printk("==== %s %d lnksta(0x%x) & LBMS\n", __func__, __LINE__, lnksta);
return lnksta & PCI_EXP_LNKSTA_LBMS;
+ }
+ printk("==== %s %d count:0x%lx\n", __func__, __LINE__, count);
return count > 0;
}
@@ -110,6 +113,8 @@ int pcie_failed_link_retrain(struct pci_dev *dev)
pcie_capability_read_word(dev, PCI_EXP_LNKCTL2, &lnkctl2);
pcie_capability_read_word(dev, PCI_EXP_LNKSTA, &lnksta);
+ pci_info(dev, "============ %s %d, lnkctl2:0x%x, lnksta:0x%x\n",
+ __func__, __LINE__, lnkctl2, lnksta);
if (!(lnksta & PCI_EXP_LNKSTA_DLLLA) && pcie_lbms_seen(dev, lnksta)) {
u16 oldlnkctl2 = lnkctl2;
@@ -121,9 +126,14 @@ int pcie_failed_link_retrain(struct pci_dev *dev)
pcie_set_target_speed(dev, PCIE_LNKCTL2_TLS2SPEED(oldlnkctl2),
true);
return ret;
+ } else {
+ pci_info(dev, "retraining sucessfully, but now is in Gen 1\n");
}
+ pcie_capability_read_word(dev, PCI_EXP_LNKCTL2, &lnkctl2);
pcie_capability_read_word(dev, PCI_EXP_LNKSTA, &lnksta);
+ pci_info(dev, "============ %s %d, oldlnkctl2:0x%x,newlnkctl2:0x%x,newlnksta:0x%x\n",
+ __func__, __LINE__, oldlnkctl2, lnkctl2, lnksta);
}
if ((lnksta & PCI_EXP_LNKSTA_DLLLA) &&
-------------------------------diff file-----------------------------------------
Based on the information in the log from 566.755596 to 567.801220, the issue
has been reproduced. Between 566 and 567 seconds, the pcie_bwnotif_irq interrupt
was triggered 4 times, this indicates that during this period, the NVMe drive
was plugged and unplugged multiple times.
Thanks,
Regards,
Jiwei
> didn't explain LBMS (nor DLLLA) in the above sequence so it's hard to
> follow what is going on here. LBMS in particular is of high interest here
> because I'm trying to understand if something should clear it on the
> hotplug side (there's already one call to clear it in remove_board()).
>
> In step 2 (pcie_set_target_speed() in step 1 succeeded),
> pcie_failed_link_retrain() attempts to restore >2.5GT/s speed, this only
> occurs when pci_match_id() matches. I guess you're trying to say that step
> 2 is not taken because pci_match_id() is not matching but the wording
> above is very confusing.
>
> Overall, I failed to understand the scenario here fully despite trying to
> think it through over these few days.
>
>> the target link speed
>> field of the Link Control
>> 2 Register will keep 0x1.
>>
>> In order to fix the issue, don't do the retraining work except ASMedia
>> ASM2824.
>>
>> Fixes: a89c82249c37 ("PCI: Work around PCIe link training failures")
>> Reported-by: Adrian Huang <ahuang12@xxxxxxxxxx>
>> Signed-off-by: Jiwei Sun <sunjw10@xxxxxxxxxx>
>> ---
>> drivers/pci/quirks.c | 6 ++++--
>> 1 file changed, 4 insertions(+), 2 deletions(-)
>>
>> diff --git a/drivers/pci/quirks.c b/drivers/pci/quirks.c
>> index 605628c810a5..ff04ebd9ae16 100644
>> --- a/drivers/pci/quirks.c
>> +++ b/drivers/pci/quirks.c
>> @@ -104,6 +104,9 @@ int pcie_failed_link_retrain(struct pci_dev *dev)
>> u16 lnksta, lnkctl2;
>> int ret = -ENOTTY;
>>
>> + if (!pci_match_id(ids, dev))
>> + return 0;
>> +
>> if (!pci_is_pcie(dev) || !pcie_downstream_port(dev) ||
>> !pcie_cap_has_lnkctl2(dev) || !dev->link_active_reporting)
>> return ret;
>> @@ -129,8 +132,7 @@ int pcie_failed_link_retrain(struct pci_dev *dev)
>> }
>>
>> if ((lnksta & PCI_EXP_LNKSTA_DLLLA) &&
>> - (lnkctl2 & PCI_EXP_LNKCTL2_TLS) == PCI_EXP_LNKCTL2_TLS_2_5GT &&
>> - pci_match_id(ids, dev)) {
>> + (lnkctl2 & PCI_EXP_LNKCTL2_TLS) == PCI_EXP_LNKCTL2_TLS_2_5GT) {
>> u32 lnkcap;
>>
>> pci_info(dev, "removing 2.5GT/s downstream link speed restriction\n");
>>
>
