linux-kernel - Re: [PATCH v9 2/5] crypto: loongson - add Loongson RNG driver support

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: <96118a23-3e6c-c9d1-2135-bd7a22091f35@loongson.cn>
Date: Mon, 19 May 2025 16:13:14 +0800
From: Qunqin Zhao <zhaoqunqin@...ngson.cn>
To: Herbert Xu <herbert@...dor.apana.org.au>
Cc: lee@...nel.org, davem@...emloft.net, peterhuewe@....de,
 jarkko@...nel.org, linux-kernel@...r.kernel.org, loongarch@...ts.linux.dev,
 linux-crypto@...r.kernel.org, jgg@...pe.ca, linux-integrity@...r.kernel.org,
 pmenzel@...gen.mpg.de, Yinggang Gu <guyinggang@...ngson.cn>,
 Huacai Chen <chenhuacai@...ngson.cn>
Subject: Re: [PATCH v9 2/5] crypto: loongson - add Loongson RNG driver support


在 2025/5/19 下午1:57, Herbert Xu 写道:
> On Tue, May 06, 2025 at 11:19:44AM +0800, Qunqin Zhao wrote:
>> +static int loongson_rng_init(struct crypto_tfm *tfm)
>> +{
>> +	struct loongson_rng_ctx *ctx = crypto_tfm_ctx(tfm);
>> +	struct loongson_rng *rng;
>> +	int ret = -EBUSY;
>> +
>> +	mutex_lock(&rng_devices.lock);
>> +	list_for_each_entry(rng, &rng_devices.list, list) {
>> +		if (!rng->is_used) {
>> +			rng->is_used = true;
>> +			ctx->rng = rng;
>> +			ret = 0;
>> +			break;
>> +		}
>> +	}
>> +	mutex_unlock(&rng_devices.lock);
>> +
>> +	return ret;
>> +}
> This isn't right.  The number of TFMs in the system is unlimited.
> You should not pair each tfm with an individual hardwre device.
Then the HISI TRNG driver isn't a right demo?
>
> If you want to do load-balancing you could certainly pick a device
> per tfm, but each device must be able to support an unlimited number
> of tfms.

This can also avoid concurrent access to a device, otherwise i need to

add mutex_lock/unlock in generate and seed callback.

Thanks,

Qunqin.

>
> Cheers,