Re: [LKP] [dmi] PANIC: early exception 0e rip 10:ffffffff81899e6b error 9 cr2 ffffffffff240000

From: Ard Biesheuvel
Date: Fri Nov 07 2014 - 04:04:12 EST

On 7 November 2014 09:46, Yuanhan Liu <yuanhan.liu@xxxxxxxxxxxxxxx> wrote:
> On Fri, Nov 07, 2014 at 09:23:56AM +0100, Ard Biesheuvel wrote:
>> On 7 November 2014 09:13, Yuanhan Liu <yuanhan.liu@xxxxxxxxxxxxxxx> wrote:
>> > On Fri, Nov 07, 2014 at 08:44:40AM +0100, Ard Biesheuvel wrote:
>> >> On 7 November 2014 08:37, Yuanhan Liu <yuanhan.liu@xxxxxxxxxxxxxxx> wrote:
>> >> > On Fri, Nov 07, 2014 at 08:17:36AM +0100, Ard Biesheuvel wrote:
>> >> >> On 7 November 2014 06:47, LKP <lkp@xxxxxx> wrote:
>> >> >> > FYI, we noticed the below changes on
>> >> >> >
>> >> >> > https://git.linaro.org/people/ard.biesheuvel/linux-arm efi-for-3.19
>> >> >> > commit aacdce6e880894acb57d71dcb2e3fc61b4ed4e96 ("dmi: add support for SMBIOS 3.0 64-bit entry point")
>> >> >> >
>> >> >> >
>> >> >> > +-----------------------+------------+------------+
>> >> >> > | | 2fa165a26c | aacdce6e88 |
>> >> >> > +-----------------------+------------+------------+
>> >> >> > | boot_successes | 20 | 10 |
>> >> >> > | early-boot-hang | 1 | |
>> >> >> > | boot_failures | 0 | 5 |
>> >> >> > | PANIC:early_exception | 0 | 5 |
>> >> >> > +-----------------------+------------+------------+
>> >> >> >
>> >> >> >
>> >> >> > [ 0.000000] BIOS-e820: [mem 0x0000000100000000-0x000000036fffffff] usable
>> >> >> > [ 0.000000] bootconsole [earlyser0] enabled
>> >> >> > [ 0.000000] NX (Execute Disable) protection: active
>> >> >> > PANIC: early exception 0e rip 10:ffffffff81899e6b error 9 cr2 ffffffffff240000
>> >> >> > [ 0.000000] CPU: 0 PID: 0 Comm: swapper Not tainted 3.18.0-rc2-gc5221e6 #1
>> >> >> > [ 0.000000] 0000000000000000 ffffffff82203d30 ffffffff819f0a6e 00000000000003f8
>> >> >> > [ 0.000000] ffffffffff240000 ffffffff82203e18 ffffffff823701b0 ffffffff82511401
>> >> >> > [ 0.000000] 0000000000000000 0000000000000ba3 0000000000000000 ffffffffff240000
>> >> >> > [ 0.000000] Call Trace:
>> >> >> > [ 0.000000] [<ffffffff819f0a6e>] dump_stack+0x4e/0x68
>> >> >> > [ 0.000000] [<ffffffff823701b0>] early_idt_handler+0x90/0xb7
>> >> >> > [ 0.000000] [<ffffffff823c80da>] ? dmi_save_one_device+0x81/0x81
>> >> >> > [ 0.000000] [<ffffffff81899e6b>] ? dmi_table+0x3f/0x94
>> >> >> > [ 0.000000] [<ffffffff81899e42>] ? dmi_table+0x16/0x94
>> >> >> > [ 0.000000] [<ffffffff823c80da>] ? dmi_save_one_device+0x81/0x81
>> >> >> > [ 0.000000] [<ffffffff823c80da>] ? dmi_save_one_device+0x81/0x81
>> >> >> > [ 0.000000] [<ffffffff823c7eff>] dmi_walk_early+0x44/0x69
>> >> >> > [ 0.000000] [<ffffffff823c88a2>] dmi_present+0x180/0x1ff
>> >> >> > [ 0.000000] [<ffffffff823c8ab3>] dmi_scan_machine+0x144/0x191
>> >> >> > [ 0.000000] [<ffffffff82370702>] ? loglevel+0x31/0x31
>> >> >> > [ 0.000000] [<ffffffff82377f52>] setup_arch+0x490/0xc73
>> >> >> > [ 0.000000] [<ffffffff819eef73>] ? printk+0x4d/0x4f
>> >> >> > [ 0.000000] [<ffffffff82370b90>] start_kernel+0x9c/0x43f
>> >> >> > [ 0.000000] [<ffffffff82370120>] ? early_idt_handlers+0x120/0x120
>> >> >> > [ 0.000000] [<ffffffff823704a2>] x86_64_start_reservations+0x2a/0x2c
>> >> >> > [ 0.000000] [<ffffffff823705df>] x86_64_start_kernel+0x13b/0x14a
>> >> >> > [ 0.000000] RIP 0x4
>> >> >> >
>> >> >>
>> >> >> This is most puzzling. Could anyone decode the exception?
>> >> >> This looks like the non-EFI path through dmi_scan_machine(), which
>> >> >> calls dmi_present() /after/ calling dmi_smbios3_present(), which
>> >> >> apparently has not found the _SM3_ header tag. Or could the call stack
>> >> >> be inaccurate?
>> >> >>
>> >> >> Anyway, it would be good to know the exact type of the platform,
>> >> >
>> >> > It's a Nehalem-EP machine, wht 16 CPU and 12G memory.
>> >> >
>> >> >> and
>> >> >> perhaps we could find out if there is an inadvertent _SM3_ tag
>> >> >> somewhere in the 0xF0000 - 0xFFFFF range?
>> >> >
>> >> > Sorry, how?
>> >> >
>> >>
>> >> That's not a brand new machine, so I suppose there wouldn't be a
>> >> SMBIOS 3.0 header lurking in there.
>> >>
>> >> Anyway, if you are in a position to try things, could you apply this
>> >>
>> >> --- a/drivers/firmware/dmi_scan.c
>> >> +++ b/drivers/firmware/dmi_scan.c
>> >> @@ -617,7 +617,7 @@ void __init dmi_scan_machine(void)
>> >> memset(buf, 0, 16);
>> >> for (q = p; q < p + 0x10000; q += 16) {
>> >> memcpy_fromio(buf + 16, q, 16);
>> >> - if (!dmi_smbios3_present(buf) || !dmi_present(buf)) {
>> >> + if (!dmi_present(buf)) {
>> >> dmi_available = 1;
>> >> dmi_early_unmap(p, 0x10000);
>> >> goto out;
>> >>
>> >> and try again?
>> >
>> > kernel boots perfectly with this patch applied.
>> >
>> > --yliu
>> >
>>
>> Thank you! Very useful to know
>>
>
> Sigh, I made a silly error, I speicified wrong commit while testing your
> patch. Sorry for that.
>
> And I tested it again, with your former patch, sorry, the panic still
> happens.
>
> --yliu
>

OK, no worries.

Could you please try the attached patch? On my ARM system, it produces
something like this

====== Decoding _DMI_ header:
5f 44 4d 49 5f 89 62 02 00 c0 8a fe 0c 00 27 cf
====== Remapped SMBIOS table 0xfe8ac000 at ffffff800001e000, size 0x262, num 0xc
====== Processing SMBIOS table entry at ffffff800001e000, type 0x0, length 0x18
====== Processing SMBIOS table entry at ffffff800001e043, type 0x1, length 0x1b
====== Processing SMBIOS table entry at ffffff800001e09d, type 0x2, length 0x11
====== Processing SMBIOS table entry at ffffff800001e105, type 0x3, length 0x18
====== Processing SMBIOS table entry at ffffff800001e155, type 0x4, length 0x2a
====== Processing SMBIOS table entry at ffffff800001e19a, type 0x7, length 0x13
====== Processing SMBIOS table entry at ffffff800001e1b5, type 0x9, length 0x11
====== Processing SMBIOS table entry at ffffff800001e1cf, type 0x10, length 0x17
====== Processing SMBIOS table entry at ffffff800001e1e8, type 0x11, length 0x28
====== Processing SMBIOS table entry at ffffff800001e22e, type 0x13, length 0x1f
====== Processing SMBIOS table entry at ffffff800001e24f, type 0x20, length 0xb
====== Processing SMBIOS table entry at ffffff800001e25c, type 0x7f, length 0x4
SMBIOS 2.7 present.
DMI: ARM Arm Versatile Express/Arm Versatile Express, BIOS 16:20:46 Oct 28 2014

That should help us pinpoint what is going on here.

--
Ard.
diff --git a/drivers/firmware/dmi_scan.c b/drivers/firmware/dmi_scan.c
index c5f7b4e9eb6c..0f7bc9db3d0d 100644
--- a/drivers/firmware/dmi_scan.c
+++ b/drivers/firmware/dmi_scan.c
@@ -92,6 +92,9 @@ static void dmi_table(u8 *buf, int len, int num,
while ((i < num) && (data - buf + sizeof(struct dmi_header)) <= len) {
const struct dmi_header *dm = (const struct dmi_header *)data;

+ pr_err("====== Processing SMBIOS table entry at %p, type 0x%x, length 0x%x\n",
+ data, dm->type, dm->length);
+
/*
* 7.45 End-of-Table (Type 127) [SMBIOS reference spec v3.0.0]
*/
@@ -126,6 +129,9 @@ static int __init dmi_walk_early(void (*decode)(const struct dmi_header *,
if (buf == NULL)
return -1;

+ pr_err("====== Remapped SMBIOS table 0x%llx at %p, size 0x%x, num 0x%x\n",
+ dmi_base, buf, dmi_len, dmi_num);
+
dmi_table(buf, dmi_len, dmi_num, decode, NULL);

add_device_randomness(buf, dmi_len);
@@ -495,10 +501,17 @@ static int __init dmi_present(const u8 *buf)
buf += 16;

if (memcmp(buf, "_DMI_", 5) == 0 && dmi_checksum(buf, 15)) {
+ int i;
+
dmi_num = get_unaligned_le16(buf + 12);
dmi_len = get_unaligned_le16(buf + 6);
dmi_base = get_unaligned_le32(buf + 8);

+ pr_err("====== Decoding _DMI_ header:\n");
+ for (i = 0; i < 16; i++)
+ pr_cont("%02x ", buf[i]);
+ pr_cont("\n");
+
if (dmi_walk_early(dmi_decode) == 0) {
if (smbios_ver) {
dmi_ver = smbios_ver;
@@ -617,7 +630,7 @@ void __init dmi_scan_machine(void)
memset(buf, 0, 16);
for (q = p; q < p + 0x10000; q += 16) {
memcpy_fromio(buf + 16, q, 16);
- if (!dmi_smbios3_present(buf) || !dmi_present(buf)) {
+ if (/*!dmi_smbios3_present(buf) ||*/ !dmi_present(buf)) {
dmi_available = 1;
dmi_early_unmap(p, 0x10000);
goto out;