linux-kernel - Re: [PATCH v3 2/5] EDAC/amd64: Remove NUM

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: <98e347a3-459e-41e4-8ad3-5dac086b6e74@amd.com>
Date: Wed, 10 Sep 2025 10:48:22 -0500
From: "Naik, Avadhut" <avadnaik@....com>
To: Yazen Ghannam <yazen.ghannam@....com>, Avadhut Naik <avadhut.naik@....com>
Cc: linux-edac@...r.kernel.org, bp@...en8.de, linux-kernel@...r.kernel.org
Subject: Re: [PATCH v3 2/5] EDAC/amd64: Remove NUM_CONTROLLERS macro



On 9/10/2025 10:05, Yazen Ghannam wrote:
> On Tue, Sep 09, 2025 at 06:53:11PM +0000, Avadhut Naik wrote:
>> Currently, the NUM_CONTROLLERS macro is only used to statically allocate
>> the csels array of struct chip_select in struct amd64_pvt.
>>
>> The size of this array, however, will never exceed the number of UMCs on
>> the SOC. Since, max_mcs variable in struct amd64_pvt already stores the
>> number of UMCs on the SOC, the macro can be removed and the static array
>> can be dynamically allocated instead.
> 
> You should note that max_mcs and the csels array are also used in legacy
> systems with 'DCTs'.
> 
> Those had a max of 2 controllers which we already set in
> per_family_init() as the global default. So the legacy systems are
> covered by this change too.
> 
> Without noting this, it seems like that case may be overlooked.
> 
Will mention this in the commit message!

>>
>> Signed-off-by: Avadhut Naik <avadhut.naik@....com>
>> ---
>> Changes in v3:
>> Patch introduced.
>> ---
>>  drivers/edac/amd64_edac.c | 19 +++++++++++++------
>>  drivers/edac/amd64_edac.h |  5 ++---
>>  2 files changed, 15 insertions(+), 9 deletions(-)
>>
>> diff --git a/drivers/edac/amd64_edac.c b/drivers/edac/amd64_edac.c
>> index 3989794e4f29..0fade110c3fb 100644
>> --- a/drivers/edac/amd64_edac.c
>> +++ b/drivers/edac/amd64_edac.c
>> @@ -4000,30 +4000,34 @@ static int probe_one_instance(unsigned int nid)
>>  	if (ret < 0)
>>  		goto err_enable;
>>  
>> +	pvt->csels = kcalloc(pvt->max_mcs, sizeof(*pvt->csels), GFP_KERNEL);
>> +	if (!pvt->csels)
>> +		goto err_enable;
>> +
> 
> You can move this allocation to the end of per_family_init(). That's
> where we determine 'max_mcs'.
> 
> If you do so, then the 'goto' changes below are not needed.
> 
> Another option is to put it in hw_info_get() like we do for UMCs. But
> that means adding the allocation to three different helper functions
> rather than just the one with per_family_init().
> 

Had considered moving allocation to per_family_init() while sending
this set. But then didn't since I had stated that I would be adding
this allocation in probe_one_instance().
In any case, will move it to per_family_init().

>>  	ret = pvt->ops->hw_info_get(pvt);
>>  	if (ret < 0)
>> -		goto err_enable;
>> +		goto err_csels;
>>  
>>  	ret = 0;
>>  	if (!instance_has_memory(pvt)) {
>>  		amd64_info("Node %d: No DIMMs detected.\n", nid);
>> -		goto err_enable;
>> +		goto err_csels;
>>  	}
>>  
>>  	if (!pvt->ops->ecc_enabled(pvt)) {
>>  		ret = -ENODEV;
>>  
>>  		if (!ecc_enable_override)
>> -			goto err_enable;
>> +			goto err_csels;
>>  
>>  		if (boot_cpu_data.x86 >= 0x17) {
>>  			amd64_warn("Forcing ECC on is not recommended on newer systems. Please enable ECC in BIOS.");
>> -			goto err_enable;
>> +			goto err_csels;
>>  		} else
>>  			amd64_warn("Forcing ECC on!\n");
>>  
>>  		if (!enable_ecc_error_reporting(s, nid, F3))
>> -			goto err_enable;
>> +			goto err_csels;
>>  	}
>>  
>>  	ret = init_one_instance(pvt);
>> @@ -4033,7 +4037,7 @@ static int probe_one_instance(unsigned int nid)
>>  		if (boot_cpu_data.x86 < 0x17)
>>  			restore_ecc_error_reporting(s, nid, F3);
>>  
>> -		goto err_enable;
>> +		goto err_csels;
>>  	}
>>  
>>  	amd64_info("%s detected (node %d).\n", pvt->ctl_name, pvt->mc_node_id);
>> @@ -4043,6 +4047,8 @@ static int probe_one_instance(unsigned int nid)
>>  
>>  	return ret;
>>  
>> +err_csels:
>> +	kfree(pvt->csels);
> 
> This can go in hw_info_put(). We have kfree(pvt->umc) there already.
> 
Okay. Will move it to hw_info_put().

-- 
Thanks,
Avadhut Naik