powrpc/powernv: Reset PHB in kdump kernel

In the kdump scenario, the first kerenl doesn't shutdown PCI devices and the kdump kerenl clean PHB IODA table at the early probe time. That means the kdump kerenl can't support PCI transactions piled by the first kerenl. Otherwise, lots of EEH errors and frozen PEs will be detected. In order to avoid the EEH errors, the PHB is resetted to drop all PCI transaction from the first kerenl. Signed-off-by: Gavin Shan <gwshan@linux.vnet.ibm.com> Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>

powrpc/powernv: Reset PHB in kdump kernel
In the kdump scenario, the first kerenl doesn't shutdown PCI devices and the kdump kerenl clean PHB IODA table at the early probe time. That means the kdump kerenl can't support PCI transactions piled by the first kerenl. Otherwise, lots of EEH errors and frozen PEs will be detected. In order to avoid the EEH errors, the PHB is resetted to drop all PCI transaction from the first kerenl. Signed-off-by: Gavin Shan <gwshan@linux.vnet.ibm.com> Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Gavin Shan · Benjamin Herrenschmidt · Eric Lee · Eric Lee · Eric Lee · Eric Lee
1 parent d92a208d08
Showing 3 changed files with 24 additions and 4 deletions Side-by-side Diff
arch/powerpc/platforms/powernv/eeh-ioda.c
arch/powerpc/platforms/powernv/pci-ioda.c
arch/powerpc/platforms/powernv/pci.h
@@ -388,13 +388,16 @@
 		if (rc <= 0)
 			break;
  
-		msleep(rc);
+		if (system_state < SYSTEM_RUNNING)
+			udelay(1000 * rc);
+		else
+			msleep(rc);
 	}
  
 	return rc;
 }
  
-static int ioda_eeh_phb_reset(struct pci_controller *hose, int option)
+int ioda_eeh_phb_reset(struct pci_controller *hose, int option)
 {
 	struct pnv_phb *phb = hose->private_data;
 	s64 rc = OPAL_HARDWARE;
@@ -422,8 +425,12 @@
 	 * need the PCI bus settlement delay.
 	 */
 	rc = ioda_eeh_phb_poll(phb);
-	if (option == EEH_RESET_DEACTIVATE)
-		msleep(EEH_PE_RST_SETTLE_TIME);
+	if (option == EEH_RESET_DEACTIVATE) {
+		if (system_state < SYSTEM_RUNNING)
+			udelay(1000 * EEH_PE_RST_SETTLE_TIME);
+		else
+			msleep(EEH_PE_RST_SETTLE_TIME);
+	}
 out:
 	if (rc != OPAL_SUCCESS)
 		return -EIO;
@@ -13,6 +13,7 @@
  
 #include <linux/kernel.h>
 #include <linux/pci.h>
+#include <linux/crash_dump.h>
 #include <linux/debugfs.h>
 #include <linux/delay.h>
 #include <linux/string.h>
@@ -1393,6 +1394,17 @@
 	rc = opal_pci_reset(phb_id, OPAL_PCI_IODA_TABLE_RESET, OPAL_ASSERT_RESET);
 	if (rc)
 		pr_warning("  OPAL Error %ld performing IODA table reset !\n", rc);
+
+	/* If we're running in kdump kerenl, the previous kerenl never
+	 * shutdown PCI devices correctly. We already got IODA table
+	 * cleaned out. So we have to issue PHB reset to stop all PCI
+	 * transactions from previous kerenl.
+	 */
+	if (is_kdump_kernel()) {
+		pr_info("  Issue PHB reset ...\n");
+		ioda_eeh_phb_reset(hose, EEH_RESET_FUNDAMENTAL);
+		ioda_eeh_phb_reset(hose, OPAL_DEASSERT_RESET);
+	}
 }
  
 void __init pnv_pci_init_ioda2_phb(struct device_node *np)
@@ -205,6 +205,7 @@
 extern void pnv_pci_ioda_tce_invalidate(struct iommu_table *tbl,
 					__be64 *startp, __be64 *endp, bool rm);
 extern void pnv_pci_reset_secondary_bus(struct pci_dev *dev);
+extern int ioda_eeh_phb_reset(struct pci_controller *hose, int option);
  
 #endif /* __POWERNV_PCI_H */
...	...	@@ -388,13 +388,16 @@
388	388	if (rc <= 0)
389	389	break;
390	390
391		- msleep(rc);
	391	+ if (system_state < SYSTEM_RUNNING)
	392	+ udelay(1000 * rc);
	393	+ else
	394	+ msleep(rc);
392	395	}
393	396
394	397	return rc;
395	398	}
396	399
397		-static int ioda_eeh_phb_reset(struct pci_controller *hose, int option)
	400	+int ioda_eeh_phb_reset(struct pci_controller *hose, int option)
398	401	{
399	402	struct pnv_phb *phb = hose->private_data;
400	403	s64 rc = OPAL_HARDWARE;
...	...	@@ -422,8 +425,12 @@
422	425	* need the PCI bus settlement delay.
423	426	*/
424	427	rc = ioda_eeh_phb_poll(phb);
425		- if (option == EEH_RESET_DEACTIVATE)
426		- msleep(EEH_PE_RST_SETTLE_TIME);
	428	+ if (option == EEH_RESET_DEACTIVATE) {
	429	+ if (system_state < SYSTEM_RUNNING)
	430	+ udelay(1000 * EEH_PE_RST_SETTLE_TIME);
	431	+ else
	432	+ msleep(EEH_PE_RST_SETTLE_TIME);
	433	+ }
427	434	out:
428	435	if (rc != OPAL_SUCCESS)
429	436	return -EIO;
...	...	@@ -13,6 +13,7 @@
13	13
14	14	#include <linux/kernel.h>
15	15	#include <linux/pci.h>
	16	+#include <linux/crash_dump.h>
16	17	#include <linux/debugfs.h>
17	18	#include <linux/delay.h>
18	19	#include <linux/string.h>
...	...	@@ -1393,6 +1394,17 @@
1393	1394	rc = opal_pci_reset(phb_id, OPAL_PCI_IODA_TABLE_RESET, OPAL_ASSERT_RESET);
1394	1395	if (rc)
1395	1396	pr_warning(" OPAL Error %ld performing IODA table reset !\n", rc);
	1397	+
	1398	+ /* If we're running in kdump kerenl, the previous kerenl never
	1399	+ * shutdown PCI devices correctly. We already got IODA table
	1400	+ * cleaned out. So we have to issue PHB reset to stop all PCI
	1401	+ * transactions from previous kerenl.
	1402	+ */
	1403	+ if (is_kdump_kernel()) {
	1404	+ pr_info(" Issue PHB reset ...\n");
	1405	+ ioda_eeh_phb_reset(hose, EEH_RESET_FUNDAMENTAL);
	1406	+ ioda_eeh_phb_reset(hose, OPAL_DEASSERT_RESET);
	1407	+ }
1396	1408	}
1397	1409
1398	1410	void __init pnv_pci_init_ioda2_phb(struct device_node *np)
...	...	@@ -205,6 +205,7 @@
205	205	extern void pnv_pci_ioda_tce_invalidate(struct iommu_table *tbl,
206	206	__be64 startp, __be64 endp, bool rm);
207	207	extern void pnv_pci_reset_secondary_bus(struct pci_dev *dev);
	208	+extern int ioda_eeh_phb_reset(struct pci_controller *hose, int option);
208	209
209	210	#endif /* __POWERNV_PCI_H */