[Crash-utility] [PATCH] x86_64: Fix that Particular kvaddr is converted to wrong paddr (RHEL6 x86_64)
Nakajima Akira
nakajima.akira at nttcom.co.jp
Tue Jan 5 09:46:32 UTC 2016
Thanks for your patch improvement.
+#define VSYSCALL_START 0xffffffffff600000
+#define VSYSCALL_END 0xffffffffffe00000
FIXADDR_START, FIXADDR_TOP look good for me.
But these are depended on kernel ver.
So VSYSCALL_START/END are reasonable.
2.6.9-89.ELsmp (RHEL 4.8)
2.6.18-398.el5 (RHEL 5.11)
FIXADDR_START = ffffffffff57c000
FIXADDR_TOP = ffffffffffdff000
2.6.32-71.el6.x86_64 (RHEL 6.0)
2.6.32-504.el6.x86_64 (RHEL 6.6)
FIXADDR_START = ffffffffff577000
FIXADDR_TOP = ffffffffffdff000
3.10.0-229.el7.x86_64 (RHEL 7.1)
3.10.0-329.el7.x86_64 (RHEL 7.2)
FIXADDR_START = ffffffffff525000
FIXADDR_TOP = ffffffffffdff000
4.2.3-300.fc23.x86_64 (Fedora 23)
FIXADDR_START = ffffffffff567000
FIXADDR_TOP = ffffffffff7ff000
On 2016/01/05 1:44, Dave Anderson wrote:
>
>
> ----- Original Message -----
>>> But because it is such a fundamental change that you are proposing, it might
>>> make sense to create a check/macro that is specific to the vsyscall region.
>>
>> What is check/macro? How do we check what?
>> Certainly, these addresses are vsyscall region.
>> After applying this patch, crash looks read vsyscall region correctly.
>
> The vsyscall-region check fits more naturally in the x86_64_IS_VMALLOC_ADDR() function,
> along with all of the other checks for non-unity-mapped addresses.
>
> Queued for crash-7.1.5:
>
> https://github.com/crash-utility/crash/commit/2e29d684a4583a45ae3c9b16014ab204eb6812fa
>
> Thanks for the report,
> Dave
>
>
>>
>>
>>
>> crash> sym vsyscall
>> // abbreviation //
>> ffffffffff600000 (t) .vsyscall_0
>> ffffffffff600140 (t) .vsyscall_fn
>> ffffffffff600180 (d) .vsyscall_gtod_data
>> ffffffffff600180 (D) __vsyscall_gtod_data
>> ffffffffff600400 (t) .vsyscall_1
>> ffffffffff600800 (t) .vsyscall_2
>> ffffffffff700680 (A) VDSO64_vsyscall_gtod_data
>>
>> crash> sym VDSO
>> symbol not found: VDSO
>> possible alternatives:
>> ffffffffff700000 (A) VDSO64_PRELINK
>> ffffffffff700670 (A) VDSO64_jiffies
>> ffffffffff700678 (A) VDSO64_vgetcpu_mode
>> ffffffffff700680 (A) VDSO64_vsyscall_gtod_data
>>
>>
>> //////// not patched /////////
>> crash> dis .vsyscall_0
>> 0xffffffffff600000 <.vsyscall_0>: add %al,(%rax)
>> 0xffffffffff600002 <vgettimeofday+2>: add %al,(%rax)
>> 0xffffffffff600004 <vgettimeofday+4>: add %al,(%rax)
>> 0xffffffffff600006 <vgettimeofday+6>: add %al,(%rax)
>> 0xffffffffff600008 <vgettimeofday+8>: add %al,(%rax)
>> 0xffffffffff60000a <vgettimeofday+10>: add %al,(%rax)
>> // abbreviation //
>>
>>
>> //////// after applying this patch /////////
>> crash> dis .vsyscall_0
>> 0xffffffffff600000 <.vsyscall_0>: push %rbp
>> 0xffffffffff600001 <vgettimeofday+1>: mov %rsp,%rbp
>> 0xffffffffff600004 <vgettimeofday+4>: push %r13
>> 0xffffffffff600006 <vgettimeofday+6>: push %r12
>> 0xffffffffff600008 <vgettimeofday+8>: mov %rdi,%r12
>> 0xffffffffff60000b <vgettimeofday+11>: push %rbx
>> 0xffffffffff60000c <vgettimeofday+12>: mov %rsi,%rbx
>> 0xffffffffff60000f <vgettimeofday+15>: sub $0x8,%rsp
>> 0xffffffffff600013 <vgettimeofday+19>: test %rdi,%rdi
>> 0xffffffffff600016 <vgettimeofday+22>: je 0xffffffffff6000d5
>> <vgettimeofday+213>
>> // abbreviation //
>>
>>
>> On 2015/12/22 22:42, Dave Anderson wrote:
>>>
>>>
>>> ----- Original Message -----
>>>> I didn't check XEN HYPER MODE, I don't have XEN.
>>>> If we need similar statement "if (kvaddr < MODULES_END)"
>>>> please add inside in "if (XEN_HYPER_MODE())" (1859 at x86_64_kvtop)
>>>>
>>>>
>>>>
>>>>
>>>> >From ed300b74998e0923313e4fd14b9a41e305942b44 Mon Sep 17 00:00:00 2001
>>>> From: Nakajima Akira <nakajima.akira at nttcom.co.jp>
>>>> Date: Tue, 22 Dec 2015 15:46:42 +0900
>>>> Subject: [PATCH] Fix that particular kvaddr is converted to wrong paddr
>>>>
>>>> BUG INFO
>>>> Particular kvaddr is converted to wrong paddr.
>>>> You can see this bug on RHEL6_x86_64. (at present only RHEL6)
>>>> (I checked RHEL5, RHEL7, Fedora21, Fedora23)
>>>
>>>
>>> I am out of the office until January 4th (only checking email until then),
>>> and will not
>>> be able to look into this until then. But because it is such a fundamental
>>> change that
>>> you are proposing, it might make sense to create a check/macro that is
>>> specific to
>>> the vsyscall region.
>>>
>>> Thanks,
>>> Dave
>>>
>>>>
>>>> from /proc/kallsyms
>>>> ffffffffff6008c0 D __jiffies
>>>>
>>>> /////////// wrong ///////////
>>>> crash> vtop ffffffffff6008c0
>>>> VIRTUAL PHYSICAL
>>>> ffffffffff6008c0 7f6008c0
>>>>
>>>> PAGE PHYSICAL MAPPING INDEX CNT FLAGS
>>>> ffffea00000623b8 1c11000 0 0 1 20000000000400
>>>> reserved
>>>>
>>>>
>>>> crash> rd ffffffffff6008c0
>>>> ffffffffff6008c0: 0000000000000000 ........
>>>>
>>>>
>>>> /////////// correct ///////////
>>>> crash> vtop ffffffffff6008c0
>>>> VIRTUAL PHYSICAL
>>>> ffffffffff6008c0 1c118c0
>>>>
>>>> PAGE PHYSICAL MAPPING INDEX CNT FLAGS
>>>> ffffea00000623b8 1c11000 0 0 1 20000000000400
>>>> reserved
>>>>
>>>>
>>>> crash> rd ffffffffff6008c0
>>>> ffffffffff6008c0: 00000000ffffe43a :.......
>>>>
>>>> Reported-by: Nakajima Akira <nakajima.akira at nttcom.co.jp>
>>>> Signed-off-by: Nakajima Akira <nakajima.akira at nttcom.co.jp>
>>>>
>>>> ---
>>>> x86_64.c | 28 +++++++++++++++-------------
>>>> 1 files changed, 15 insertions(+), 13 deletions(-)
>>>>
>>>> diff --git a/x86_64.c b/x86_64.c
>>>> index ff6fdd5..dab4d43 100644
>>>> --- a/x86_64.c
>>>> +++ b/x86_64.c
>>>> @@ -1872,19 +1872,21 @@ x86_64_kvtop(struct task_context *tc, ulong
>>>> kvaddr,
>>>> physaddr_t *paddr, int verbo
>>>> fprintf(fp, "PAGE DIRECTORY: %lx\n", *pml4);
>>>> }
>>>> } else {
>>>> - if (!vt->vmalloc_start) {
>>>> - *paddr = x86_64_VTOP(kvaddr);
>>>> - return TRUE;
>>>> - }
>>>> -
>>>> - if (!IS_VMALLOC_ADDR(kvaddr)) {
>>>> - *paddr = x86_64_VTOP(kvaddr);
>>>> - if (!verbose)
>>>> - return TRUE;
>>>> - }
>>>> -
>>>> - if (XEN() && (kt->xen_flags & WRITABLE_PAGE_TABLES))
>>>> - return (x86_64_kvtop_xen_wpt(tc, kvaddr, paddr, verbose));
>>>> + if (kvaddr < MODULES_END) {
>>>> + if (!vt->vmalloc_start) {
>>>> + *paddr = x86_64_VTOP(kvaddr);
>>>> + return TRUE;
>>>> + }
>>>> +
>>>> + if (!IS_VMALLOC_ADDR(kvaddr)) {
>>>> + *paddr = x86_64_VTOP(kvaddr);
>>>> + if (!verbose)
>>>> + return TRUE;
>>>> + }
>>>> +
>>>> + if (XEN() && (kt->xen_flags & WRITABLE_PAGE_TABLES))
>>>> + return (x86_64_kvtop_xen_wpt(tc, kvaddr, paddr, verbose));
>>>> + }
>>>>
>>>> /*
>>>> * pgd = pgd_offset_k(addr);
>>>> --
>>>> 1.7.1
>>>>
>>>>
>>>> --
>>>> Crash-utility mailing list
>>>> Crash-utility at redhat.com
>>>> https://www.redhat.com/mailman/listinfo/crash-utility
>>>
>>> --
>>> Crash-utility mailing list
>>> Crash-utility at redhat.com
>>> https://www.redhat.com/mailman/listinfo/crash-utility
>>> .
>>>
>>
>> --
>> Crash-utility mailing list
>> Crash-utility at redhat.com
>> https://www.redhat.com/mailman/listinfo/crash-utility
>>
>
> --
> Crash-utility mailing list
> Crash-utility at redhat.com
> https://www.redhat.com/mailman/listinfo/crash-utility
> .
>
More information about the Crash-utility
mailing list