Re: [BUG bisected]: apei_hest_parse explosion

["Rafael J. Wysocki" <rjw@sisk.pl>]

On Friday, February 22, 2013 02:40:58 AM Rafael J. Wysocki wrote:
> On Thursday, February 21, 2013 05:23:19 PM Yinghai Lu wrote:
> > On Thu, Feb 21, 2013 at 4:15 PM, Rafael J. Wysocki <rjw@sisk.pl> wrote:
> > > On Friday, February 22, 2013 12:56:01 AM Thomas Gleixner wrote:
> > >> Just booted linus head on one of my old kvm based testing systems and
> > >> got the following splat:
> > >>
> > >> [    0.531910] BUG: unable to handle kernel NULL pointer dereference at 0000000000000024
> > >> [    0.532173] IP: [<ffffffff812f1bd2>] apei_hest_parse+0xd2/0xf0
> > >> [    0.532173] PGD 0
> > >> [    0.532173] Oops: 0000 [#1] PREEMPT SMP
> > >> [    0.532173] Modules linked in:
> > >> [    0.532173] CPU 3
> > >> [    0.532173] Pid: 1, comm: swapper/0 Not tainted 3.8.0+ #195 Bochs Bochs
> > >> [    0.532173] RIP: 0010:[<ffffffff812f1bd2>]  [<ffffffff812f1bd2>] apei_hest_parse+0xd2/0xf0
> > >> [    0.532173] RSP: 0000:ffff88021685be78  EFLAGS: 00010206
> > >> [    0.532173] RAX: 0000000000000000 RBX: 0000000000000000 RCX: 0000000000000000
> > >> [    0.532173] RDX: 0000000000000000 RSI: 0000000000000000 RDI: 0000000000000028
> > >> [    0.532173] RBP: ffff88021685be98 R08: 00000000804efb06 R09: 0000000000000000
> > >> [    0.532173] R10: 0000000000000001 R11: 0000000000000000 R12: ffffffff81af3d40
> > >> [    0.532173] R13: ffffffff812bcb23 R14: 0000000000000000 R15: 0000000000000000
> > >> [    0.532173] FS:  0000000000000000(0000) GS:ffff88021fcc0000(0000) knlGS:0000000000000000
> > >> [    0.532173] CS:  0010 DS: 0000 ES: 0000 CR0: 0000000080050033
> > >> [    0.532173] CR2: 0000000000000024 CR3: 0000000001a0c000 CR4: 00000000000406a0
> > >> [    0.532173] DR0: 0000000000000000 DR1: 0000000000000000 DR2: 0000000000000000
> > >> [    0.532173] DR3: 0000000000000000 DR6: 00000000ffff0ff0 DR7: 0000000000000400
> > >> [    0.532173] Process swapper/0 (pid: 1, threadinfo ffff88021685a000, task ffff8802168c0000)
> > >> [    0.532173] Stack:
> > >> [    0.532173]  00000000fffffffa ffffffff81af3d40 0000000000000000 000000001fb3f55a
> > >> [    0.532173]  ffff88021685bea8 ffffffff812bcbe4 ffff88021685bec8 ffffffff81af3d5c
> > >> [    0.532173]  0000000000000030 0000000000000030 ffff88021685bef8 ffffffff8100023b
> > >> [    0.532173] Call Trace:
> > >> [    0.532173]  [<ffffffff81af3d40>] ? pcie_portdrv_init+0x7a/0x7a
> > >> [    0.532173]  [<ffffffff812bcbe4>] aer_acpi_firmware_first+0x20/0x30
> > >> [    0.532173]  [<ffffffff81af3d5c>] aer_service_init+0x1c/0x35
> > >> [    0.532173]  [<ffffffff8100023b>] do_one_initcall+0x57/0x13c
> > >> [    0.532173]  [<ffffffff81ad2d39>] kernel_init_freeable+0x142/0x1d1
> > >> [    0.532173]  [<ffffffff81ad254f>] ? loglevel+0x31/0x31
> > >> [    0.532173]  [<ffffffff815a8cef>] ? rest_init+0xc3/0xc3
> > >> [    0.532173]  [<ffffffff815a8cfd>] kernel_init+0xe/0xd6
> > >> [    0.532173]  [<ffffffff815bc73c>] ret_from_fork+0x7c/0xb0
> > >> [    0.532173]  [<ffffffff815a8cef>] ? rest_init+0xc3/0xc3
> > >> [    0.532173] Code: 14 0f b7 77 02 31 c0 48 c7 c7 65 c1 83 81 e8 35 1a 2c 00 eb 23 4c 89 f6 41 ff d5 85 c0 75 1e 4c 89 e7 ff c3 48 8b 05 a6 5b 7c 00 <3b> 58 24 0f 82 5a ff ff ff 31 c0 eb 05 b8 ea ff ff ff 5b 41 5c
> > >>
> > >> git bisect identified the following commit:
> > >>
> > >> commit 92ef2a25c763338905dce8344a0584606f842920
> > >> Author: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
> > >> Date:   Fri Dec 21 00:36:40 2012 +0100
> > >>
> > >>     ACPI: Change the ordering of PCI root bridge driver registrarion
> > >>
> > >> A revert is impossible w/o major surgery, so I'm going to bed and let
> > >> the involved folks think about it.
> > >
> > > Can you please post a dmesg boot log from v3.8 on the same system?
> > >
> > 
> > looks at the code flow, acpi_hest_init() is called early correctly.
> > 
> > later aer_acpi_firmware_first() should not hit hest_tab is NULL, but
> > hest_disable is not set.
> > 
> > could be some data corruption.. near hest_disable or hest_tab.
> 
> Yes, that's really odd.
> 
> It looks like the hest_tab memory mapping is unmapped between acpi_hest_init()
> and aer_acpi_firmware_first(), but I have no idea what may be responsible for
> that.
> 
> And the only relevant difference between now and before the commit above seems
> to be the change of the acpi_hest_init() ordering (which now is called earlier).

We actually don't really need to do that thing so early, I think.  It looks like
we only need to make it available early enough for the AER driver to be able to
use it, so I wonder if moving the acpi_hest_init() to a separate
subsys_initcall() will work around the problem.  That is, something like the
patch below.

But even if this helps, I will be wanting to understand what's up here.

Thanks,
Rafael


---
 drivers/acpi/apei/hest.c |   17 +++++++++++------
 drivers/acpi/pci_root.c  |    2 --
 include/acpi/apei.h      |    6 ------
 3 files changed, 11 insertions(+), 14 deletions(-)

Index: test/drivers/acpi/apei/hest.c
===================================================================
--- test.orig/drivers/acpi/apei/hest.c
+++ test/drivers/acpi/apei/hest.c
@@ -205,7 +205,7 @@ static int __init setup_hest_disable(cha
 
 __setup("hest_disable", setup_hest_disable);
 
-void __init acpi_hest_init(void)
+static int __init acpi_hest_init(void)
 {
 	acpi_status status;
 	int rc = -ENODEV;
@@ -213,7 +213,7 @@ void __init acpi_hest_init(void)
 
 	if (hest_disable) {
 		pr_info(HEST_PFX "Table parsing disabled.\n");
-		return;
+		return 0;
 	}
 
 	if (acpi_disabled)
@@ -221,9 +221,10 @@ void __init acpi_hest_init(void)
 
 	status = acpi_get_table(ACPI_SIG_HEST, 0,
 				(struct acpi_table_header **)&hest_tab);
-	if (status == AE_NOT_FOUND)
+	if (status == AE_NOT_FOUND) {
+		pr_info(HEST_PFX "Table not found.\n");
 		goto err;
-	else if (ACPI_FAILURE(status)) {
+	} else if (ACPI_FAILURE(status)) {
 		const char *msg = acpi_format_exception(status);
 		pr_err(HEST_PFX "Failed to get table, %s\n", msg);
 		rc = -EINVAL;
@@ -240,7 +241,11 @@ void __init acpi_hest_init(void)
 	}
 
 	pr_info(HEST_PFX "Table parsing has been initialized.\n");
-	return;
-err:
+	return 0;
+
+ err:
 	hest_disable = 1;
+	return rc;
 }
+
+subsys_initcall(acpi_hest_init);
Index: test/drivers/acpi/pci_root.c
===================================================================
--- test.orig/drivers/acpi/pci_root.c
+++ test/drivers/acpi/pci_root.c
@@ -656,8 +656,6 @@ static void acpi_pci_root_remove(struct
 
 void __init acpi_pci_root_init(void)
 {
-	acpi_hest_init();
-
 	if (!acpi_pci_disabled) {
 		pci_acpi_crs_quirks();
 		acpi_scan_add_handler(&pci_root_handler);
Index: test/include/acpi/apei.h
===================================================================
--- test.orig/include/acpi/apei.h
+++ test/include/acpi/apei.h
@@ -24,12 +24,6 @@ extern bool ghes_disable;
 #define ghes_disable 1
 #endif
 
-#ifdef CONFIG_ACPI_APEI
-void __init acpi_hest_init(void);
-#else
-static inline void acpi_hest_init(void) { return; }
-#endif
-
 typedef int (*apei_hest_func_t)(struct acpi_hest_header *hest_hdr, void *data);
 int apei_hest_parse(apei_hest_func_t func, void *data);
 


-- 
I speak only for myself.
Rafael J. Wysocki, Intel Open Source Technology Center.