linux-kernel - Re: [PATCH v2 08/16] x86/mce/amd: Clean up enable_deferred_error

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]

Message-ID: <d527fccc-6c32-43bc-bfad-477e6f07e33c@amd.com>
Date: Mon, 29 Apr 2024 10:18:59 -0400
From: Yazen Ghannam <yazen.ghannam@....com>
To: Borislav Petkov <bp@...en8.de>
Cc: yazen.ghannam@....com, linux-edac@...r.kernel.org,
 linux-kernel@...r.kernel.org, tony.luck@...el.com, x86@...nel.org,
 Avadhut.Naik@....com, John.Allen@....com
Subject: Re: [PATCH v2 08/16] x86/mce/amd: Clean up
 enable_deferred_error_interrupt()

On 4/29/2024 9:12 AM, Borislav Petkov wrote:
> On Thu, Apr 04, 2024 at 10:13:51AM -0500, Yazen Ghannam wrote:
>> -/* Deferred error settings */
>> +/* MCA Interrupt Configuration register, one per CPU */
> 
> SMCA?
> 
>>  #define MSR_CU_DEF_ERR		0xC0000410
>> -#define MASK_DEF_LVTOFF		0x000000F0
>> -#define MASK_DEF_INT_TYPE	0x00000006
>> -#define DEF_INT_TYPE_APIC	0x2
>> +#define MSR_MCA_INTR_CFG		0xC0000410
> 
> You do see those other MSRs' prefixes, right?
> 
> MSR_AMD64_SMCA_...
> 
> Is this one not part of the SMCA arch?
> 

No, it is part of SUCCOR. The old define is above: MSR_CU_DEF_ERR.

This is how it is listed in the PPR:
MSRC000_0410 [MCA Interrupt Configuration] (Core::X86::Msr::McaIntrCfg)

>> +#define INTR_CFG_DFR_LVT_OFFSET		GENMASK_ULL(7, 4)
>> +#define INTR_CFG_LEGACY_DFR_INTR_TYPE	GENMASK_ULL(2, 1)
>>  #define INTR_TYPE_APIC			0x1
> 
> Ditto for its bit(s) names.
>

Okay.

>> +static u64 get_mca_intr_cfg(void)
>> +{
>> +	u64 mca_intr_cfg;
>> +
>> +	if (!mce_flags.succor)
>> +		return 0;
>> +
>> +	if (rdmsrl_safe(MSR_MCA_INTR_CFG, &mca_intr_cfg))
>> +		return 0;
>> +
>> +	return mca_intr_cfg;
>> +}
> 
> This is an overkill. If we add a function for every MSR we're reading...
> 
> Do this differently: prepare the value you're writing back into the
> INTR_CFG MSR once, save it into mca_intr_cfg and then write it on each
> core at the end of enable_deferred_error_interrupt().
> 
> And make u64 mca_intr_cfg static global to amd.c so that you can refer
> to it from outside of the functions and then you don't have to pass it
> around as a function param.
> 
> Thx.
> 

Good idea. In fact, we can treat this register as read-only, since we will
only handle (SUCCOR && SMCA) systems. The only need to write this register
would be on !SMCA systems.

We need to assume that the register value will be identical for all CPUs. This
is the expectation, but I'll add a comment to highlight this.

Also, we don't need the entire register. We just need the LVT offset fields
which are 4 bits each.

Thanks,
Yazen