linux-kernel - Re: [PATCH 1/1] coresight: fix issue where coresight component has no claimtags

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: <20251022092958.GQ281971@e132581.arm.com>
Date: Wed, 22 Oct 2025 10:29:58 +0100
From: Leo Yan <leo.yan@....com>
To: Mike Leach <mike.leach@...aro.org>
Cc: coresight@...ts.linaro.org, linux-arm-kernel@...ts.infradead.org,
	linux-kernel@...r.kernel.org, suzuki.poulose@....com,
	james.clark@...aro.org
Subject: Re: [PATCH 1/1] coresight: fix issue where coresight component has
 no claimtags

On Wed, Oct 22, 2025 at 12:45:20AM +0100, Mike Leach wrote:

[...]

> +/*
> + * Reading CLIAMSET returns  a bitfield representing the number of claim tags
> + * implemented from bit 0 to bit nTag-1, valid bits set to 1.
> + *
> + * Claim protocol requires 2 bits so test for highest bit required,
> + * bit 1 -  CORESIGHT_CLAIM_SELF_HOSTED_BIT
> + *
> + * return true if sufficient claim tags implemented for protocol
> + */
> +static bool coresight_claim_tags_implemented_unlocked(struct csdev_access *csa)
> +{
> +	u32 claim_bits_impl = FIELD_GET(CORESIGHT_CLAIM_BITS_MAX_MASK,
> +			 csdev_access_relaxed_read32(csa, CORESIGHT_CLAIMSET));
> +	return ((claim_bits_impl & CORESIGHT_CLAIM_SELF_HOSTED_BIT) != 0);
> +}
> +

The logic looks good to me.

Can we improve a bit with a cached flag to avoid every time checking the
CLAIMSET, we can read it only once at init time instead.

- We can add a new flag ("bool claim_impl" in the struct csdev_access),
  by default the field will be zero.

- The drivers support claim tags call coresight_clear_self_claim_tag()
  in probe (see __catu_probe() as an example), we can call a new
  function coresight_init_claim_tag() to replace it, this function sets
  "claim_impl" properly and clear the tag if supported.

- Afterwards, simply check the "claim_impl" flag.

[...]

> +/*
> + * Coresight specification defines a maximum of 8 claim tag bits.
> + * The precise number is implementation defined, and may be obtained by
> + * reading the CLAIMSET register.
> + */
> +#define CORESIGHT_CLAIM_BITS_MAX_MASK	GENMASK(7, 0)
> +#define CORESIGHT_CLAIM_SELF_HOSTED_BIT	BIT(1)

Now we only care about the self-host bit. Can reuse existed macros ?

  CORESIGHT_CLAIM_MASK
  CORESIGHT_CLAIM_SELF_HOSTED

I don't mind if extend CORESIGHT_CLAIM_MASK to 8 bits and even remove
CORESIGHT_CLAIM_INVALID (as it cannot cover other invalid values). Or we
can defer the extension as needed later.

Thanks,
Leo