On 27 Sep 2020, at 16:45, Bjorn Helgaas wrote:
On Tue, Sep 22, 2020 at 02:38:56PM -0700, Sean V Kelley wrote:
From: Qiuxu Zhuo <qiuxu.zhuo@xxxxxxxxx>
When attempting error recovery for an RCiEP associated with an RCEC
device,
there needs to be a way to update the Root Error Status, the
Uncorrectable
Error Status and the Uncorrectable Error Severity of the parent RCEC.
In some non-native cases in which there is no OS visible device
associated with the RCiEP, there is nothing to act upon as the
firmware
is acting before the OS. So add handling for the linked 'rcec' in
AER/ERR
while taking into account non-native cases.
Co-developed-by: Sean V Kelley <sean.v.kelley@xxxxxxxxx>
Signed-off-by: Sean V Kelley <sean.v.kelley@xxxxxxxxx>
Signed-off-by: Qiuxu Zhuo <qiuxu.zhuo@xxxxxxxxx>
---
drivers/pci/pcie/aer.c | 9 +++++----
drivers/pci/pcie/err.c | 38 ++++++++++++++++++++++++--------------
2 files changed, 29 insertions(+), 18 deletions(-)
diff --git a/drivers/pci/pcie/aer.c b/drivers/pci/pcie/aer.c
index 65dff5f3457a..dccdba60b5d9 100644
--- a/drivers/pci/pcie/aer.c
+++ b/drivers/pci/pcie/aer.c
@@ -1358,17 +1358,18 @@ static int aer_probe(struct pcie_device *dev)
static pci_ers_result_t aer_root_reset(struct pci_dev *dev)
{
int aer = dev->aer_cap;
+ int rc = 0;
u32 reg32;
- int rc;
-
/* Disable Root's interrupt in response to error messages */
pci_read_config_dword(dev, aer + PCI_ERR_ROOT_COMMAND, ®32);
reg32 &= ~ROOT_PORT_INTR_ON_MESG_MASK;
pci_write_config_dword(dev, aer + PCI_ERR_ROOT_COMMAND, reg32);
- rc = pci_bus_error_reset(dev);
- pci_info(dev, "Root Port link has been reset\n");
+ if (pci_pcie_type(dev) != PCI_EXP_TYPE_RC_EC) {
+ rc = pci_bus_error_reset(dev);
+ pci_info(dev, "Root Port link has been reset\n");
+ }
/* Clear Root Error Status */
pci_read_config_dword(dev, aer + PCI_ERR_ROOT_STATUS, ®32);
diff --git a/drivers/pci/pcie/err.c b/drivers/pci/pcie/err.c
index 5380ecc41506..a61a2518163a 100644
--- a/drivers/pci/pcie/err.c
+++ b/drivers/pci/pcie/err.c
@@ -149,7 +149,8 @@ static int report_resume(struct pci_dev *dev,
void *data)
/**
* pci_bridge_walk - walk bridges potentially AER affected
* @bridge bridge which may be an RCEC with associated RCiEPs,
- * an RCiEP associated with an RCEC, or a Port.
+ * or a Port.
+ * @dev an RCiEP lacking an associated RCEC.
* @cb callback to be called for each device found
* @userdata arbitrary pointer to be passed to callback.
*
@@ -160,13 +161,16 @@ static int report_resume(struct pci_dev *dev,
void *data)
* If the device provided has no subordinate bus, call the provided
* callback on the device itself.
*/
-static void pci_bridge_walk(struct pci_dev *bridge, int (*cb)(struct
pci_dev *, void *),
+static void pci_bridge_walk(struct pci_dev *bridge, struct pci_dev
*dev,
+ int (*cb)(struct pci_dev *, void *),
void *userdata)
{
- if (bridge->subordinate)
+ if (bridge && bridge->subordinate)
I *guess* this means there's a possibility that bridge is NULL? And I
guess that case is when pci_upstream_bridge(dev) in pcie_do_recovery()
was NULL?
I can't figure out what that means in practice. Oh, wait, I bet I
know -- this is the non-native case where there's no OS-visible
reporting device.
Yes, this is for the non-native case.
We need some sort of comment because this is really not a top-of-mind
situation unless you happen to be working on one of the few systems
like that.
Will do.
Not too sure this scenario is really a good fit for the
pci_bridge_walk() model, but I think it's OK for now with a hint so we
can remember this no RCEC, no Root Port model.
Understood.
pci_walk_bus(bridge->subordinate, cb, userdata);
- else
+ else if (bridge)
cb(bridge, userdata);
+ else
+ cb(dev, userdata);
}
static pci_ers_result_t flr_on_rciep(struct pci_dev *dev)
@@ -196,16 +200,24 @@ pci_ers_result_t pcie_do_recovery(struct
pci_dev *dev,
type = pci_pcie_type(dev);
if (type == PCI_EXP_TYPE_ROOT_PORT ||
type == PCI_EXP_TYPE_DOWNSTREAM ||
- type == PCI_EXP_TYPE_RC_EC ||
- type == PCI_EXP_TYPE_RC_END)
+ type == PCI_EXP_TYPE_RC_EC)
bridge = dev;
+ else if (type == PCI_EXP_TYPE_RC_END)
+ bridge = dev->rcec;
else
bridge = pci_upstream_bridge(dev);
pci_dbg(dev, "broadcast error_detected message\n");
if (state == pci_channel_io_frozen) {
- pci_bridge_walk(bridge, report_frozen_detected, &status);
+ pci_bridge_walk(bridge, dev, report_frozen_detected, &status);
if (type == PCI_EXP_TYPE_RC_END) {
+ /*
+ * The callback only clears the Root Error Status
+ * of the RCEC (see aer.c).
+ */
+ if (bridge)
+ reset_subordinate_devices(bridge);
It's unfortunate to add yet another special case in this code, and I'm
not thrilled about the one in aer_root_reset() either. It's just not
obvious why they should be there. I'm sure if I spent 30 minutes
decoding things, it would all make sense. Guess I'm just griping
because I don't have a better suggestion.
I could also add some more detail here referencing the non-native case
where there may not even exist an RCEC. Aside from comments, I could
employ a descriptive flag such as “native_bridge”.
Thanks,
Sean
status = flr_on_rciep(dev);
if (status != PCI_ERS_RESULT_RECOVERED) {
pci_warn(dev, "function level reset failed\n");
@@ -219,13 +231,13 @@ pci_ers_result_t pcie_do_recovery(struct
pci_dev *dev,
}
}
} else {
- pci_bridge_walk(bridge, report_normal_detected, &status);
+ pci_bridge_walk(bridge, dev, report_normal_detected, &status);
}
if (status == PCI_ERS_RESULT_CAN_RECOVER) {
status = PCI_ERS_RESULT_RECOVERED;
pci_dbg(dev, "broadcast mmio_enabled message\n");
- pci_bridge_walk(bridge, report_mmio_enabled, &status);
+ pci_bridge_walk(bridge, dev, report_mmio_enabled, &status);
}
if (status == PCI_ERS_RESULT_NEED_RESET) {
@@ -236,18 +248,16 @@ pci_ers_result_t pcie_do_recovery(struct
pci_dev *dev,
*/
status = PCI_ERS_RESULT_RECOVERED;
pci_dbg(dev, "broadcast slot_reset message\n");
- pci_bridge_walk(bridge, report_slot_reset, &status);
+ pci_bridge_walk(bridge, dev, report_slot_reset, &status);
}
if (status != PCI_ERS_RESULT_RECOVERED)
goto failed;
pci_dbg(dev, "broadcast resume message\n");
- pci_bridge_walk(bridge, report_resume, &status);
+ pci_bridge_walk(bridge, dev, report_resume, &status);
- if (type == PCI_EXP_TYPE_ROOT_PORT ||
- type == PCI_EXP_TYPE_DOWNSTREAM ||
- type == PCI_EXP_TYPE_RC_EC) {
+ if (bridge) {
if (pcie_aer_is_native(bridge))
pcie_clear_device_status(bridge);
pci_aer_clear_nonfatal_status(bridge);
--
2.28.0
