Date:   Wed, 19 Jan 2022 14:40:11 +0800
From:   Shuai Xue <xueshuai@...ux.alibaba.com>
To:     Bjorn Helgaas <helgaas@...nel.org>,
        "Rafael J. Wysocki" <rafael@...nel.org>
Cc:     bp@...en8.de, tony.luck@...el.com, james.morse@....com,
        lenb@...nel.org, rjw@...ysocki.net, bhelgaas@...gle.com,
        zhangliguang@...ux.alibaba.com, zhuo.song@...ux.alibaba.com,
        linux-kernel@...r.kernel.org, linux-acpi@...r.kernel.org,
        linux-arm-kernel@...ts.infradead.org, linux-pci@...r.kernel.org
Subject: Re: [PATCH v5] ACPI: Move sdei_init and ghes_init ahead to handle
 platform errors earlier

[+to Rafael, question about HEST/GHES/SDEI init]

Hi, Bjorn,

Thank you for your comments and quick reply.

在 2022/1/19 AM6:49, Bjorn Helgaas 写道:
> On Sun, Jan 16, 2022 at 04:43:10PM +0800, Shuai Xue wrote:
>> On an ACPI system, ACPI is initialised very early from a
>> subsys_initcall(), while SDEI is not ready until a
>> subsys_initcall_sync(). This patch is to reduce the time before GHES
>> initialization.
>>
>> The SDEI driver provides functions (e.g. apei_sdei_register_ghes(),
>> apei_sdei_unregister_ghes()) to register or unregister event callback
>> for dispatcher in firmware. When the GHES driver probing, it registers
>> the corresponding callback according to the notification type specified
>> by GHES. If the GHES notification type is SDEI, the GHES driver will
>> call apei_sdei_register_ghes() to register event call.
>>
>> When the firmware emits an event, it migrates the handling of the event
>> into the kernel at the registered entry-point __sdei_asm_handler. And
>> finally, the kernel will call the registered event callback and return
>> status_code to indicate the status of event handling. SDEI_EV_FAILED
>> indicates that the kernel failed to handle the event.
>>
>> Consequently, when an error occurs during kernel booting, the kernel is
>> unable to handle and report errors until the GHES driver is initialized
>> by device_initcall(), in which the event callback is registered.  For
>> example, when the kernel booting, the console logs many times from
>> firmware before GHES drivers init in our platform:
>>
>> 	Trip in MM PCIe RAS handle(Intr:910)
>>   	Clean PE[1.1.1] ERR_STS:0x4000100 -> 0 INT_STS:F0000000
>> 	Find RP(98:1.0)
>> 	--Walk dev(98:1.0) CE:0 UCE:4000
>> 	...
>> 	ERROR:   sdei_dispatch_event(32a) ret:-1
>> 	--handler(910) end
> 
> If I understand correctly, the firmware noticed an error, tried to
> report it to the kernel, and is complaining because the kernel isn't
> ready to handle it yet.  And the reason for this patch is to reduce
> these complaints from the firmware.

My thoughts exactly :)

> That doesn't seem like a very good reason for this patch.  There is
> *always* a window before the kernel is ready to handle events from the
> firmware.

Yes, there is always a window. But if we could do better in kernel that
reduces the window by 90% (from 33 seconds to 3 second), why not?

> Why is the firmware noticing these errors in the first place?  If
> you're seeing these complaints regularly, my guess is that either you
> have some terrible hardware or (more likely) the firmware isn't
> clearing some expected error condition correctly.  For example, maybe
> the Unsupported Request errors that happen while enumerating PCIe
> devices are being reported.
>
> If you register the callback function, the kernel will now start
> seeing these error reports.  What happens then?  Does the kernel log
> the errors somewhere?  Is that better than the current situation where
> the firmware logs them?

Yep, it is a hardware issue. The firmware only logs in console (ttyAMA0) and
we can not see it in kernel side. After the kernel starts seeing these error
reports, we could see EDAC/ghes and efi/cper detailed logs in dmesg. We did not
notice the problem until we check the console log, which inspired us to reduce
the window when kernel startup, so that we can see the message clearly and
properly. I think the intuition is to check the log of dmesg, not the console.

> However, I DO think that:
> 
>   - Removing acpi_hest_init() from acpi_pci_root_init(), and
> 
>   - Converting ghes_init() and sdei_init() from initcalls to explicit
>     calls
> 
> are very good reasons to do something like this patch because HEST is
> not PCI-specific, and IMO, explicit calls are better than initcalls
> because initcall ordering is implicit and not well-defined within a
> level.

Haha, if the above reasons still don't convince you, I would like to accept yours :)
Should we do it in one patch or separate it into two patches?

>> This is because the callback function has not been registered yet.  Due
>> to the poor design of firmware, previously reported errors will be
>> overwritten by new ones. Therefore, all errors that occurred before GHES
>> initialization are missed and there is no chance to report and find them
>> again.
>>
>> From commit e147133a42cb ("ACPI / APEI: Make hest.c manage the estatus
>> memory pool") was merged, ghes_init() relies on acpi_hest_init() to manage
>> the estatus memory pool. On the other hand, ghes_init() relies on
>> sdei_init() to detect the SDEI version and the framework for registering
>> and unregistering events. The dependencies are as follows
>>
>> 	ghes_init() => acpi_hest_init() => acpi_bus_init() => acpi_init()
>> 	ghes_init() => sdei_init()
>>
>> Based on above, move sdei_init and ghes_init as far ahead as possible,
>> right after acpi_iort_init() in acpi_init(). After this patch, there are
>> still errors missing, but we have to tolerate it until the software
>> reporting capability is built. And this patch merely reduce the time
>> before GHES initialization. The boot dmesg before this patch:
>>
>> 	[    3.688586] HEST: Table parsing has been initialized.
>> 	...
>> 	[   33.204340] calling  sdei_init+0x0/0x120 @ 1
>> 	[   33.208645] sdei: SDEIv1.0 (0x0) detected in firmware.
>> 	...
>> 	[   36.005390] calling  ghes_init+0x0/0x11c @ 1
>> 	[   36.190021] GHES: APEI firmware first mode is enabled by APEI bit and WHEA _OSC.
>>
>> After this patch, the boot dmesg like bellow:
>>
>> 	[    3.688664] HEST: Table parsing has been initialized.
>> 	[    3.688691] sdei: SDEIv1.0 (0x0) detected in firmware.
>> 	[    3.694557] GHES: APEI firmware first mode is enabled by APEI bit and WHEA _OSC.
>>
>> As we can see, the initialization of GHES is advanced by 33 seconds.
>> (It should be noted that the effect of optimization varies with the
>> platform.)
>>
>> Signed-off-by: Shuai Xue <xueshuai@...ux.alibaba.com>
>> ---
>>  drivers/acpi/apei/ghes.c    | 18 ++++++++----------
>>  drivers/acpi/bus.c          |  4 ++++
>>  drivers/acpi/pci_root.c     |  4 ++--
>>  drivers/firmware/arm_sdei.c | 13 ++-----------
>>  include/acpi/apei.h         |  2 ++
>>  include/linux/arm_sdei.h    |  2 ++
>>  6 files changed, 20 insertions(+), 23 deletions(-)
>>
>> diff --git a/drivers/acpi/apei/ghes.c b/drivers/acpi/apei/ghes.c
>> index 0c5c9acc6254..64f24a9f213f 100644
>> --- a/drivers/acpi/apei/ghes.c
>> +++ b/drivers/acpi/apei/ghes.c
>> @@ -1457,27 +1457,26 @@ static struct platform_driver ghes_platform_driver = {
>>  	.remove		= ghes_remove,
>>  };
>>  
>> -static int __init ghes_init(void)
>> +void __init ghes_init(void)
>>  {
>>  	int rc;
>>  
>>  	if (acpi_disabled)
>> -		return -ENODEV;
>> +		return;
>>  
>>  	switch (hest_disable) {
>>  	case HEST_NOT_FOUND:
>> -		return -ENODEV;
>> +		pr_info(GHES_PFX "HEST is not found!\n");
> 
> I don't know whether this "HEST is not found" message is worthwhile or
> not.  I don't think lack of an HEST is an error, and users may be
> alarmed.  But this is an ACPI thing, so up to you and Rafael.

If we explicit call ghes_init(), we can't tell if ghes is initialized
successfully based on the return value of initcall. So I add a info message.
What's your opinion, Rafeal?

>> +		return;
>>  	case HEST_DISABLED:
>>  		pr_info(GHES_PFX "HEST is not enabled!\n");
>> -		return -EINVAL;
>> +		return;
>>  	default:
>>  		break;
>>  	}
>>  
>> -	if (ghes_disable) {
>> +	if (ghes_disable)
>>  		pr_info(GHES_PFX "GHES is not enabled!\n");
>> -		return -EINVAL;
>> -	}
>>  
>>  	ghes_nmi_init_cxt();
> 
> Previously, if "ghes_disable", we returned immediately.  After this
> patch, if "ghes_disable", we print "GHES is not enabled!\n", but then
> we go on to call ghes_nmi_init_cxt() and register the
> &ghes_platform_driver.  That looks wrong.  Maybe you meant to keep a
> "return" here?

Sorry, it's my wrong. I will fix it in next version.

>> @@ -1495,8 +1494,7 @@ static int __init ghes_init(void)
>>  	else
>>  		pr_info(GHES_PFX "Failed to enable APEI firmware first mode.\n");
>>  
>> -	return 0;
>> +	return;
>>  err:
>> -	return rc;
>> +	ghes_disable = 1;
> 
> Why do you set ghes_disable here?  As far as I can tell, we will never
> look at it again.  The places we do look at it are:
> 
>   - ghes_init(): earlier in this function, so we've already done that,
> 
>   - acpi_hest_init(): we've already called that, too, and
> 
>   - acpi_bus_osc_negotiate_platform_control(): called from
>     acpi_bus_init(), which we've already called.

I add it for future potential usage. Thank you for pointing it out.
If you think it is not necessary, I will delete it in next version.


>>  }
>> -device_initcall(ghes_init);
>> diff --git a/drivers/acpi/bus.c b/drivers/acpi/bus.c
>> index 07f604832fd6..1dcd71df2cd5 100644
>> --- a/drivers/acpi/bus.c
>> +++ b/drivers/acpi/bus.c
>> @@ -30,6 +30,7 @@
>>  #include <linux/acpi_viot.h>
>>  #include <linux/pci.h>
>>  #include <acpi/apei.h>
>> +#include <linux/arm_sdei.h>
>>  #include <linux/suspend.h>
>>  #include <linux/prmt.h>
>>  
>> @@ -1331,6 +1332,9 @@ static int __init acpi_init(void)
>>  
>>  	pci_mmcfg_late_init();
>>  	acpi_iort_init();
>> +	acpi_hest_init();
>> +	sdei_init();
>> +	ghes_init();
> 
> sdei_init() and ghes_init() stick out here because they're the only
> functions without an "acpi_" prefix.  Maybe a separate patch to rename
> them?

Great idea. I will add a patch to rename them with "acpi_" prefix

>>  	acpi_scan_init();
>>  	acpi_ec_init();
>>  	acpi_debugfs_init();
>> diff --git a/drivers/acpi/pci_root.c b/drivers/acpi/pci_root.c
>> index b76db99cced3..14e20af44627 100644
>> --- a/drivers/acpi/pci_root.c
>> +++ b/drivers/acpi/pci_root.c
>> @@ -23,7 +23,7 @@
>>  #include <linux/dmi.h>
>>  #include <linux/platform_data/x86/apple.h>
>>  #include <acpi/apei.h>	/* for acpi_hest_init() */
> 
> This #include looks like it may no longer be necessary.
>
>> -
>> +#include <linux/arm_sdei.h> /* for sdei_init() */
> 
> I don't think this #include is necessary.

Thank you for pointing it out. I will delete it in next version.

> 
>>  #include "internal.h"
>>  
>>  #define ACPI_PCI_ROOT_CLASS		"pci_bridge"
>> @@ -943,7 +943,7 @@ struct pci_bus *acpi_pci_root_create(struct acpi_pci_root *root,
>>  
>>  void __init acpi_pci_root_init(void)
>>  {
>> -	acpi_hest_init();
>> +
> 
> Drop the blank line here.

Got it. Will fix it.

> 
>>  	if (acpi_pci_disabled)
>>  		return;
>>  
>> diff --git a/drivers/firmware/arm_sdei.c b/drivers/firmware/arm_sdei.c
>> index a7e762c352f9..1e1a51510e83 100644
>> --- a/drivers/firmware/arm_sdei.c
>> +++ b/drivers/firmware/arm_sdei.c
>> @@ -1059,14 +1059,14 @@ static bool __init sdei_present_acpi(void)
>>  	return true;
>>  }
>>  
>> -static int __init sdei_init(void)
>> +void __init sdei_init(void)
>>  {
>>  	struct platform_device *pdev;
>>  	int ret;
>>  
>>  	ret = platform_driver_register(&sdei_driver);
>>  	if (ret || !sdei_present_acpi())
>> -		return ret;
>> +		return;
>>  
>>  	pdev = platform_device_register_simple(sdei_driver.driver.name,
>>  					       0, NULL, 0);
>> @@ -1076,17 +1076,8 @@ static int __init sdei_init(void)
>>  		pr_info("Failed to register ACPI:SDEI platform device %d\n",
>>  			ret);
>>  	}
>> -
>> -	return ret;
>>  }
>>  
>> -/*
>> - * On an ACPI system SDEI needs to be ready before HEST:GHES tries to register
>> - * its events. ACPI is initialised from a subsys_initcall(), GHES is initialised
>> - * by device_initcall(). We want to be called in the middle.
>> - */
>> -subsys_initcall_sync(sdei_init);
>> -
>>  int sdei_event_handler(struct pt_regs *regs,
>>  		       struct sdei_registered_event *arg)
>>  {
>> diff --git a/include/acpi/apei.h b/include/acpi/apei.h
>> index ece0a8af2bae..7dbd6363fda7 100644
>> --- a/include/acpi/apei.h
>> +++ b/include/acpi/apei.h
>> @@ -27,8 +27,10 @@ extern int hest_disable;
>>  extern int erst_disable;
>>  #ifdef CONFIG_ACPI_APEI_GHES
>>  extern bool ghes_disable;
>> +void __init ghes_init(void);
>>  #else
>>  #define ghes_disable 1
>> +static inline void ghes_init(void) { return; }
> 
> "return" is unnecessary here.
>
>>  #endif
>>  
>>  #ifdef CONFIG_ACPI_APEI
>> diff --git a/include/linux/arm_sdei.h b/include/linux/arm_sdei.h
>> index 0a241c5c911d..9c987188b692 100644
>> --- a/include/linux/arm_sdei.h
>> +++ b/include/linux/arm_sdei.h
>> @@ -46,9 +46,11 @@ int sdei_unregister_ghes(struct ghes *ghes);
>>  /* For use by arch code when CPU hotplug notifiers are not appropriate. */
>>  int sdei_mask_local_cpu(void);
>>  int sdei_unmask_local_cpu(void);
>> +void __init sdei_init(void);
>>  #else
>>  static inline int sdei_mask_local_cpu(void) { return 0; }
>>  static inline int sdei_unmask_local_cpu(void) { return 0; }
>> +static inline void sdei_init(void) { return ; }
> 
> "return" is unnecessary here.

Got it. Will delete "return" in ghes_init(), sdei_init() and acpi_hest_init().

>>  #endif /* CONFIG_ARM_SDE_INTERFACE */
>>  
>>  
>> -- 
>> 2.20.1.12.g72788fdb
>>
>>
>> _______________________________________________
>> linux-arm-kernel mailing list
>> linux-arm-kernel@...ts.infradead.org
>> http://lists.infradead.org/mailman/listinfo/linux-arm-kernel

Powered by blists - more mailing lists