jcs.org
qemu with hax to log dma reads & writes
jcs.org/2018/11/12/vfio
ppc/spapr: Don't kill the guest if a recovered FWNMI machine check delivery fails
Try to be tolerant of FWNMI delivery errors if the machine check had been
recovered by the host.
Signed-off-by: Nicholas Piggin <npiggin@gmail.com>
Message-Id: <20200325142906.221248-5-npiggin@gmail.com>
Reviewed-by: Greg Kurz <groug@kaod.org>
[dwg: Updated comment at Greg's suggestion]
Signed-off-by: David Gibson <david@gibson.dropbear.id.au>
authored by
Nicholas Piggin
and committed by
David Gibson
4f7a11f9
b90b9ecb
+25
-5
+25
-5
hw/ppc/spapr_events.c
+25
-5
hw/ppc/spapr_events.c
···
833
833
/* get rtas addr from fdt */
834
834
rtas_addr = spapr_get_rtas_addr();
835
835
if (!rtas_addr) {
836
-
error_report(
836
+
if (!recovered) {
837
+
error_report(
837
838
"FWNMI: Unable to deliver machine check to guest: rtas_addr not found.");
838
-
qemu_system_guest_panicked(NULL);
839
+
qemu_system_guest_panicked(NULL);
840
+
} else {
841
+
warn_report(
842
+
"FWNMI: Unable to deliver machine check to guest: rtas_addr not found. "
843
+
"Machine check recovered.");
844
+
}
839
845
g_free(ext_elog);
840
846
return;
841
847
}
842
848
849
+
/*
850
+
* By taking the interlock, we assume that the MCE will be
851
+
* delivered to the guest. CAUTION: don't add anything that could
852
+
* prevent the MCE to be delivered after this line, otherwise the
853
+
* guest won't be able to release the interlock and ultimately
854
+
* hang/crash?
855
+
*/
856
+
spapr->fwnmi_machine_check_interlock = cpu->vcpu_id;
857
+
843
858
stq_be_phys(&address_space_memory, rtas_addr + RTAS_ERROR_LOG_OFFSET,
844
859
env->gpr[3]);
845
860
cpu_physical_memory_write(rtas_addr + RTAS_ERROR_LOG_OFFSET +
···
876
891
* that CPU called "ibm,nmi-interlock")
877
892
*/
878
893
if (spapr->fwnmi_machine_check_interlock == cpu->vcpu_id) {
879
-
error_report(
894
+
if (!recovered) {
895
+
error_report(
880
896
"FWNMI: Unable to deliver machine check to guest: nested machine check.");
881
-
qemu_system_guest_panicked(NULL);
897
+
qemu_system_guest_panicked(NULL);
898
+
} else {
899
+
warn_report(
900
+
"FWNMI: Unable to deliver machine check to guest: nested machine check. "
901
+
"Machine check recovered.");
902
+
}
882
903
return;
883
904
}
884
905
qemu_cond_wait_iothread(&spapr->fwnmi_machine_check_interlock_cond);
···
906
927
warn_report("Received a fwnmi while migration was in progress");
907
928
}
908
929
909
-
spapr->fwnmi_machine_check_interlock = cpu->vcpu_id;
910
930
spapr_mce_dispatch_elog(cpu, recovered);
911
931
}
912
932