Groups | Search | Server Info | Keyboard shortcuts | Login | Register [http] [https] [nntp] [nntps]
Groups > linux.kernel > #1580138 > unrolled thread
| Started by | "Baicar, Tyler" <tbaicar@codeaurora.org> |
|---|---|
| First post | 2017-02-13 23:50 +0100 |
| Last post | 2017-02-15 18:10 +0100 |
| Articles | 3 — 2 participants |
Back to article view | Back to linux.kernel
This discussion starts older than the indexed window; earlier articles aren't shown. The article labeled Started by
below is the oldest one visible, not the original post.
Re: [PATCH V8 06/10] acpi: apei: panic OS with fatal error status block "Baicar, Tyler" <tbaicar@codeaurora.org> - 2017-02-13 23:50 +0100
Re: [PATCH V8 06/10] acpi: apei: panic OS with fatal error status block James Morse <james.morse@arm.com> - 2017-02-15 13:20 +0100
Re: [PATCH V8 06/10] acpi: apei: panic OS with fatal error status block "Baicar, Tyler" <tbaicar@codeaurora.org> - 2017-02-15 18:10 +0100
| From | "Baicar, Tyler" <tbaicar@codeaurora.org> |
|---|---|
| Date | 2017-02-13 23:50 +0100 |
| Subject | Re: [PATCH V8 06/10] acpi: apei: panic OS with fatal error status block |
| Message-ID | <taxBw-1wx-17@gated-at.bofh.it> |
Hello James,
On 2/9/2017 3:48 AM, James Morse wrote:
> Hi Jonathan, Tyler,
>
> On 01/02/17 17:16, Tyler Baicar wrote:
>> From: "Jonathan (Zhixiong) Zhang" <zjzhang@codeaurora.org>
>>
>> Even if an error status block's severity is fatal, the kernel does not
>> honor the severity level and panic.
>>
>> With the firmware first model, the platform could inform the OS about a
>> fatal hardware error through the non-NMI GHES notification type. The OS
>> should panic when a hardware error record is received with this
>> severity.
>>
>> Call panic() after CPER data in error status block is printed if
>> severity is fatal, before each error section is handled.
>> diff --git a/drivers/acpi/apei/ghes.c b/drivers/acpi/apei/ghes.c
>> index 8756172..86c1f15 100644
>> --- a/drivers/acpi/apei/ghes.c
>> +++ b/drivers/acpi/apei/ghes.c
>> @@ -687,6 +689,13 @@ static int ghes_ack_error(struct acpi_hest_generic_v2 *generic_v2)
>> return rc;
>> }
>>
>> +static void __ghes_call_panic(void)
>> +{
>> + if (panic_timeout == 0)
>> + panic_timeout = ghes_panic_timeout;
>> + panic("Fatal hardware error!");
>> +}
>> +
> __ghes_panic() also has:
>> __ghes_print_estatus(KERN_EMERG, ghes->generic, ghes->estatus);
> Which prints this estatus regardless of rate limiting and cache-ing.
>
> [ ... ]
>
>> @@ -698,6 +707,10 @@ static int ghes_proc(struct ghes *ghes)
>> if (ghes_print_estatus(NULL, ghes->generic, ghes->estatus))
> ghes_print_estatus() uses some custom rate limiting '2 messages every 5
> seconds', GHES_SEV_PANIC shares the same limit as GHES_SEV_RECOVERABLE.
>
> I think its possible to get 2 recoverable messages, then a panic in a 5 second
> window. The rate limit will kick in to stop the panic estatus block being
> printed, but we still go on to call panic() without the real reason being printed...
>
> (the caching thing only seems to consider identical messages, given we would
> never see two panic messages, I don't think that will cause any problems.)
>
>> ghes_estatus_cache_add(ghes->generic, ghes->estatus);
>> }
>> + if (ghes_severity(ghes->estatus->error_severity) >= GHES_SEV_PANIC) {
>> + __ghes_call_panic();
>> + }
>> +
> I think this ghes_severity() then panic() should go above the:
>> if (!ghes_estatus_cached(ghes->estatus)) {
> and we should call __ghes_print_estatus() here too, to make sure the message
> definitely got out!
Okay, that makes sense. If we move this up, is there a problem with
calling __ghes_panic() instead of making the __ghes_print_estatus() and
__ghes_call_panic() calls here? It looks like that will just add a call
to oops_begin() and ghes_print_queued_estatus() as well, but this is
what ghes_notify_nmi() does if the severity is panic.
Thanks,
Tyler
> With that,
> Reviewed-by: James Morse <james.morse@arm.com>
>
>
> Thanks,
>
> James
--
Qualcomm Datacenter Technologies, Inc. as an affiliate of Qualcomm Technologies, Inc.
Qualcomm Technologies, Inc. is a member of the Code Aurora Forum,
a Linux Foundation Collaborative Project.
[toc] | [next] | [standalone]
| From | James Morse <james.morse@arm.com> |
|---|---|
| Date | 2017-02-15 13:20 +0100 |
| Message-ID | <tb6IV-8jM-1@gated-at.bofh.it> |
| In reply to | #1580138 |
Hi Tyler,
On 13/02/17 22:45, Baicar, Tyler wrote:
> On 2/9/2017 3:48 AM, James Morse wrote:
>> On 01/02/17 17:16, Tyler Baicar wrote:
>>> From: "Jonathan (Zhixiong) Zhang" <zjzhang@codeaurora.org>
>>>
>>> Even if an error status block's severity is fatal, the kernel does not
>>> honor the severity level and panic.
>>>
>>> With the firmware first model, the platform could inform the OS about a
>>> fatal hardware error through the non-NMI GHES notification type. The OS
>>> should panic when a hardware error record is received with this
>>> severity.
>>>
>>> Call panic() after CPER data in error status block is printed if
>>> severity is fatal, before each error section is handled.
>>> diff --git a/drivers/acpi/apei/ghes.c b/drivers/acpi/apei/ghes.c
>>> index 8756172..86c1f15 100644
>>> --- a/drivers/acpi/apei/ghes.c
>>> +++ b/drivers/acpi/apei/ghes.c
>>> @@ -687,6 +689,13 @@ static int ghes_ack_error(struct acpi_hest_generic_v2
>>> *generic_v2)
>>> return rc;
>>> }
>>> +static void __ghes_call_panic(void)
>>> +{
>>> + if (panic_timeout == 0)
>>> + panic_timeout = ghes_panic_timeout;
>>> + panic("Fatal hardware error!");
>>> +}
>>> +
>> __ghes_panic() also has:
>>> __ghes_print_estatus(KERN_EMERG, ghes->generic, ghes->estatus);
>> Which prints this estatus regardless of rate limiting and cache-ing.
[...]
>>> ghes_estatus_cache_add(ghes->generic, ghes->estatus);
>>> }
>>> + if (ghes_severity(ghes->estatus->error_severity) >= GHES_SEV_PANIC) {
>>> + __ghes_call_panic();
>>> + }
>> I think this ghes_severity() then panic() should go above the:
>>> if (!ghes_estatus_cached(ghes->estatus)) {
>> and we should call __ghes_print_estatus() here too, to make sure the message
>> definitely got out!
> Okay, that makes sense. If we move this up, is there a problem with calling
> __ghes_panic() instead of making the __ghes_print_estatus() and
> __ghes_call_panic() calls here? It looks like that will just add a call to
> oops_begin() and ghes_print_queued_estatus() as well, but this is what
> ghes_notify_nmi() does if the severity is panic.
I don't think the queued stuff is relevant, isn't that just for x86-NMI messages
that it doesn't print out directly?
A quick grep shows arm64 doesn't have oops_begin(), you may have to add some
equivalent mechanism. Lets try and avoid that rabbit hole!
Given __ghes_panic() calls __ghes_print_estatus() too, you could try moving that
into your new __ghes_call_panic().... or whatever results in the least lines
changed!
Thanks,
James
[toc] | [prev] | [next] | [standalone]
| From | "Baicar, Tyler" <tbaicar@codeaurora.org> |
|---|---|
| Date | 2017-02-15 18:10 +0100 |
| Message-ID | <tbbfA-2La-25@gated-at.bofh.it> |
| In reply to | #1581265 |
On 2/15/2017 5:13 AM, James Morse wrote:
> Hi Tyler,
>
> On 13/02/17 22:45, Baicar, Tyler wrote:
>> On 2/9/2017 3:48 AM, James Morse wrote:
>>> On 01/02/17 17:16, Tyler Baicar wrote:
>>>> From: "Jonathan (Zhixiong) Zhang" <zjzhang@codeaurora.org>
>>>>
>>>> Even if an error status block's severity is fatal, the kernel does not
>>>> honor the severity level and panic.
>>>>
>>>> With the firmware first model, the platform could inform the OS about a
>>>> fatal hardware error through the non-NMI GHES notification type. The OS
>>>> should panic when a hardware error record is received with this
>>>> severity.
>>>>
>>>> Call panic() after CPER data in error status block is printed if
>>>> severity is fatal, before each error section is handled.
>>>> diff --git a/drivers/acpi/apei/ghes.c b/drivers/acpi/apei/ghes.c
>>>> index 8756172..86c1f15 100644
>>>> --- a/drivers/acpi/apei/ghes.c
>>>> +++ b/drivers/acpi/apei/ghes.c
>>>> @@ -687,6 +689,13 @@ static int ghes_ack_error(struct acpi_hest_generic_v2
>>>> *generic_v2)
>>>> return rc;
>>>> }
>>>> +static void __ghes_call_panic(void)
>>>> +{
>>>> + if (panic_timeout == 0)
>>>> + panic_timeout = ghes_panic_timeout;
>>>> + panic("Fatal hardware error!");
>>>> +}
>>>> +
>>> __ghes_panic() also has:
>>>> __ghes_print_estatus(KERN_EMERG, ghes->generic, ghes->estatus);
>>> Which prints this estatus regardless of rate limiting and cache-ing.
> [...]
>
>>>> ghes_estatus_cache_add(ghes->generic, ghes->estatus);
>>>> }
>>>> + if (ghes_severity(ghes->estatus->error_severity) >= GHES_SEV_PANIC) {
>>>> + __ghes_call_panic();
>>>> + }
>>> I think this ghes_severity() then panic() should go above the:
>>>> if (!ghes_estatus_cached(ghes->estatus)) {
>>> and we should call __ghes_print_estatus() here too, to make sure the message
>>> definitely got out!
>
>> Okay, that makes sense. If we move this up, is there a problem with calling
>> __ghes_panic() instead of making the __ghes_print_estatus() and
>> __ghes_call_panic() calls here? It looks like that will just add a call to
>> oops_begin() and ghes_print_queued_estatus() as well, but this is what
>> ghes_notify_nmi() does if the severity is panic.
>
> I don't think the queued stuff is relevant, isn't that just for x86-NMI messages
> that it doesn't print out directly?
>
> A quick grep shows arm64 doesn't have oops_begin(), you may have to add some
> equivalent mechanism. Lets try and avoid that rabbit hole!
>
> Given __ghes_panic() calls __ghes_print_estatus() too, you could try moving that
> into your new __ghes_call_panic().... or whatever results in the least lines
> changed!
Sounds good, I will just use __ghes_print_estatus() and __ghes_call_panic().
Thanks,
Tyler
--
Qualcomm Datacenter Technologies, Inc. as an affiliate of Qualcomm Technologies, Inc.
Qualcomm Technologies, Inc. is a member of the Code Aurora Forum,
a Linux Foundation Collaborative Project.
[toc] | [prev] | [standalone]
Back to top | Article view | linux.kernel
csiph-web