Skip to contentSkip to conversation

: resumable error decoding

SPARC64 Kirill Korotaev 3 participants 4 messages Nov 21, 2006 – Nov 30, 2006

lore
Share & tools
lore
#0 lore raw

David,

Running stress tests on OpenVZ 2.6.18 sparc64 kernel we hit the following:
------- cut --------

[285401.094964] RESUMABLE ERROR: Reporting on cpu 0
[285401.626736] RESUMABLE ERROR: err_handle[410000000000c6f] err_stick[103921ee2007c] err_type[00000004:warning resumable]
[285402.869015] RESUMABLE ERROR: err_attrs[00000020:       ]
[285403.491920] RESUMABLE ERROR: err_raddr[0000000000000000] err_size[0] err_cpu[0]
[285404.347508] TSTATE: 0000004480001602 TPC: 000000000041931c TNPC: 0000000000419320 Y: 00000000    Not tainted
[285405.496613] TPC: <cpu_idle+0x84/0xc0>
[285405.892615] g0: 00000000006e2531 g1: 0000000000000016 g2: 0000000000000014 g3: 00000000006def80
[285406.550536] g4: 00000000006e2f80 g5: fffff8000449bd40 g6: 00000000006def80 g7: 0000038000004000
[285406.884717] o0: 0000000000000000 o1: 00000000006def88 o2: 0000000000004000 o3: 4000000000000000
[285407.214724] o4: 0000000000001290 o5: 0000000000000012 sp: 00000000006e2531 ret_pc: 0000000000419308
[285407.562135] RPC: <cpu_idle+0x70/0xc0>
[285407.701342] l0: 00000000006de800 l1: 0000000000000027 l2: 0000000000000000 l3: 00000001ff000000
[285408.029282] l4: 0000000040004110 l5: 00000000fff74080 l6: 00000000fff4d701 l7: 00000000f0254040
[285408.348195] i0: 0000000100000000 i1: 0000000000000000 i2: 0000000000000000 i3: 0000000100000000
[285408.681920] i4: 0000000000000080 i5: 0000000000000080 i6: 00000000006e25f1 i7: 00000000007a67ec
[285409.010870] I7: <start_kernel+0x294/0x300>

------- cut --------

it looks like the hardware reports some problem and the most interesting field is err_attrs...

        u32             err_attrs;
#define SUN4V_ERR_ATTRS_PROCESSOR       0x00000001
#define SUN4V_ERR_ATTRS_MEMORY          0x00000002
#define SUN4V_ERR_ATTRS_PIO             0x00000004
#define SUN4V_ERR_ATTRS_INT_REGISTERS   0x00000008
#define SUN4V_ERR_ATTRS_FPU_REGISTERS   0x00000010
#define SUN4V_ERR_ATTRS_USER_MODE       0x01000000
#define SUN4V_ERR_ATTRS_PRIV_MODE       0x02000000
#define SUN4V_ERR_ATTRS_RES_QUEUE_FULL  0x80000000

.. which should explain what subsystem is faulty. However, 2.6.18 kernel knows nothing about the value 0x20 :/ I also didn't find anything in available documenation about this. Can you sched some light on this please?
A link to the doc or some hint would be very much appreciated.

Thanks,
Kirill

#1 lore raw

From: Kirill Korotaev <[email protected]>
Date: Tue, 21 Nov 2006 16:42:47 +0300

Quoted 6 lines

Running stress tests on OpenVZ 2.6.18 sparc64 kernel we hit the following:
------- cut --------

[285401.094964] RESUMABLE ERROR: Reporting on cpu 0
[285401.626736] RESUMABLE ERROR: err_handle[410000000000c6f] err_stick[103921ee2007c] err_type[00000004:warning resumable]
[285402.869015] RESUMABLE ERROR: err_attrs[00000020:       ]
[285403.491920] RESUMABLE ERROR: err_raddr[0000000000000000] err_size[0] err_cpu[0]

This is a power-off request, did someone push the power-off button or give the power-off command from the System Controller console?

I should add proper support for this, this report is a good reminder :-)

All resumable errors of type 0x4 are power-off requests. Unfortunately these encodings are not in any of the publicly published documents.

#2 lore raw
Quoted 10 lines
Quoted 6 lines

Running stress tests on OpenVZ 2.6.18 sparc64 kernel we hit the following:
------- cut --------

[285401.094964] RESUMABLE ERROR: Reporting on cpu 0
[285401.626736] RESUMABLE ERROR: err_handle[410000000000c6f] err_stick[103921ee2007c] err_type[00000004:warning resumable]
[285402.869015] RESUMABLE ERROR: err_attrs[00000020:       ]
[285403.491920] RESUMABLE ERROR: err_raddr[0000000000000000] err_size[0] err_cpu[0]

This is a power-off request, did someone push the power-off button or give the power-off command from the System Controller console?

ahh, looks like this :)
one of our users reproduced an issue which causes both mainstream and 2.6.18 OVZ kerenls to hang on sparc :/ will investigate... probably he reset the box after the hang :)

Quoted 2 lines

I should add proper support for this, this report is a good reminder :-)

would be nice :@)

Quoted 3 lines

All resumable errors of type 0x4 are power-off requests. Unfortunately these encodings are not in any of the publicly published documents.

thanks a lot for the explanation!

Thanks,
Kirill

#3 lore raw

From: Kirill Korotaev <[email protected]>
Date: Wed, 22 Nov 2006 13:19:28 +0300

Quoted 3 lines
Quoted 2 lines

I should add proper support for this, this report is a good reminder :-)

would be nice :@)

I tested the following patch and it worked fine for me on a T2000, let me know if it works for you too:

commit 035f09edbbc921b9688a65ec58c0f49b822e605c

Author: David S. Miller <[email protected]>
Date:   Wed Nov 29 21:16:21 2006 -0800

[SPARC64]: Run ctrl-alt-del action for sun4v powerdown request.

arch/sparc64/kernel/traps.c +11 -1
index ec7a601..ad67784 100644
@@ -10,7 +10,7 @@
  */
 
 #include <linux/module.h>
-#include <linux/sched.h>  /* for jiffies */
+#include <linux/sched.h>
 #include <linux/kernel.h>
 #include <linux/kallsyms.h>
 #include <linux/signal.h>
@@ -1873,6 +1873,16 @@ void sun4v_resum_error(struct pt_regs *r
 
 	put_cpu();
 
+	if (ent->err_type == SUN4V_ERR_TYPE_WARNING_RES) {
+		/* If err_type is 0x4, it's a powerdown request.  Do
+		 * not do the usual resumable error log because that
+		 * makes it look like some abnormal error.
+		 */
+		printk(KERN_INFO "Power down request...\n");
+		kill_cad_pid(SIGINT, 1);
+		return;
+	}
+
 	sun4v_log_error(regs, &local_copy, cpu,
 			KERN_ERR "RESUMABLE ERROR",
 			&sun4v_resum_oflow_cnt);