[PATCH] powerpc/pseries: avoid crash in PCI code if mem system not up
authorLinas Vepstas <linas@austin.ibm.com>
Thu, 27 Apr 2006 09:31:20 +0000 (02:31 -0700)
committerPaul Mackerras <paulus@samba.org>
Wed, 3 May 2006 13:06:40 +0000 (23:06 +1000)
The powerpc code is currently performing PCI setup before memory
initialization.  PCI setup touches PCI config space registers.  If the PCI
card is bad, this will evoke an error, which currrently can't be handled,
as the PCI error recovery code expects kmalloc() to be functional.  This
patch will cause the system to punt instead of crashing with

cpu 0x0: Vector: 300 (Data Access) at [c0000000004434d0]
    pc: c0000000000c06b4: .kmem_cache_alloc+0x8c/0xf4
    lr: c00000000004ad6c: .eeh_send_failure_event+0x48/0xfc

This patch will also print name of the offending pci device.

Signed-off-by: Linas Vepstas <linas@austin.ibm.com>
Cc: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Paul Mackerras <paulus@samba.org>
arch/powerpc/platforms/pseries/eeh_event.c

index a1bda6f96fd1ff2c4a50f7d2221779e3b6629124..40020c65c89e4da0e2365fac47316593ee712c12 100644 (file)
@@ -118,7 +118,15 @@ int eeh_send_failure_event (struct device_node *dn,
 {
        unsigned long flags;
        struct eeh_event *event;
+       char *location;
 
+       if (!mem_init_done) {
+               printk(KERN_ERR "EEH: event during early boot not handled\n");
+               location = (char *) get_property(dn, "ibm,loc-code", NULL);
+               printk(KERN_ERR "EEH: device node = %s\n", dn->full_name);
+               printk(KERN_ERR "EEH: PCI location = %s\n", location);
+               return 1;
+       }
        event = kmalloc(sizeof(*event), GFP_ATOMIC);
        if (event == NULL) {
                printk (KERN_ERR "EEH: out of memory, event not handled\n");