public inbox for linux-acpi@vger.kernel.org
 help / color / mirror / Atom feed
From: Jonathan Cameron <jonathan.cameron@huawei.com>
To: Mauro Carvalho Chehab <mchehab+huawei@kernel.org>
Cc: "Rafael J. Wysocki" <rafael@kernel.org>,
	Ankit Agrawal <ankita@nvidia.com>, Borislav Petkov <bp@alien8.de>,
	Breno Leitao <leitao@debian.org>,
	Hanjun Guo <guohanjun@huawei.com>, Ingo Molnar <mingo@kernel.org>,
	Jason Tian <jason@os.amperecomputing.com>,
	"Len Brown" <lenb@kernel.org>,
	Mauro Carvalho Chehab <mchehab@kernel.org>,
	Shuai Xue <xueshuai@linux.alibaba.com>,
	Smita Koralahalli <Smita.KoralahalliChannabasappa@amd.com>,
	Tony Luck <tony.luck@intel.com>, <linux-efi@vger.kernel.org>,
	<linux-acpi@vger.kernel.org>, <linux-edac@vger.kernel.org>,
	<linux-kernel@vger.kernel.org>
Subject: Re: [PATCH v3 1/2] apei/ghes: ARM processor Error: don't go past allocated memory
Date: Mon, 22 Dec 2025 11:38:51 +0000	[thread overview]
Message-ID: <20251222113851.000048f6@huawei.com> (raw)
In-Reply-To: <e80bc4eba43d0211713fe66958ec0c582d9bfda7.1766140788.git.mchehab+huawei@kernel.org>

On Fri, 19 Dec 2025 11:49:59 +0100
Mauro Carvalho Chehab <mchehab+huawei@kernel.org> wrote:

> If the BIOS generates a very small ARM Processor Error, or
> an incomplete one, the current logic will fail to deferrence
> 
> 	err->section_length
> and
> 	ctx_info->size
> 
> Add checks to avoid that. With such changes, such GHESv2
> records won't cause OOPSes like this:
> 
> [    1.492129] Internal error: Oops: 0000000096000005 [#1]  SMP
> [    1.495449] Modules linked in:
> [    1.495820] CPU: 0 UID: 0 PID: 9 Comm: kworker/0:0 Not tainted 6.18.0-rc1-00017-gabadcc3553dd-dirty #18 PREEMPT
> [    1.496125] Hardware name: QEMU QEMU Virtual Machine, BIOS unknown 02/02/2022
> [    1.496433] Workqueue: kacpi_notify acpi_os_execute_deferred
> [    1.496967] pstate: 814000c5 (Nzcv daIF +PAN -UAO -TCO +DIT -SSBS BTYPE=--)
> [    1.497199] pc : log_arm_hw_error+0x5c/0x200
> [    1.497380] lr : ghes_handle_arm_hw_error+0x94/0x220
> 
> 0xffff8000811c5324 is in log_arm_hw_error (../drivers/ras/ras.c:75).
> 70		err_info = (struct cper_arm_err_info *)(err + 1);
> 71		ctx_info = (struct cper_arm_ctx_info *)(err_info + err->err_info_num);
> 72		ctx_err = (u8 *)ctx_info;
> 73
> 74		for (n = 0; n < err->context_info_num; n++) {
> 75			sz = sizeof(struct cper_arm_ctx_info) + ctx_info->size;
> 76			ctx_info = (struct cper_arm_ctx_info *)((long)ctx_info + sz);
> 77			ctx_len += sz;
> 78		}
> 79
> 
> and similar ones while trying to access section_length on an
> error dump with too small size.
> 
> Signed-off-by: Mauro Carvalho Chehab <mchehab+huawei@kernel.org>
Hi Mauro,

This is fiddly stuff to read in the spec but I think you have a double
counting of the "ARM Processors Error Information Structure" size as
the length in that this time is the length of the structure itself,
not a following body.

Jonathan


> ---
>  drivers/acpi/apei/ghes.c | 33 +++++++++++++++++++++++++++++----
>  drivers/ras/ras.c        |  6 +++++-
>  2 files changed, 34 insertions(+), 5 deletions(-)
> 
> diff --git a/drivers/acpi/apei/ghes.c b/drivers/acpi/apei/ghes.c
> index 0dc767392a6c..9bf4ec84f160 100644
> --- a/drivers/acpi/apei/ghes.c
> +++ b/drivers/acpi/apei/ghes.c
> @@ -552,21 +552,46 @@ static bool ghes_handle_arm_hw_error(struct acpi_hest_generic_data *gdata,
>  {
>  	struct cper_sec_proc_arm *err = acpi_hest_get_payload(gdata);
>  	int flags = sync ? MF_ACTION_REQUIRED : 0;
> +	int length = gdata->error_data_length;
>  	char error_type[120];
>  	bool queued = false;
>  	int sec_sev, i;
>  	char *p;
>  
>  	sec_sev = ghes_severity(gdata->error_severity);
> -	log_arm_hw_error(err, sec_sev);
> +	if (length >= sizeof(*err)) {
> +		log_arm_hw_error(err, sec_sev);
> +	} else {
> +		pr_warn(FW_BUG "arm error length: %d\n", length);
> +		pr_warn(FW_BUG "length is too small\n");
> +		pr_warn(FW_BUG "firmware-generated error record is incorrect\n");
> +		return false;
> +	}
> +
>  	if (sev != GHES_SEV_RECOVERABLE || sec_sev != GHES_SEV_RECOVERABLE)
>  		return false;
>  
>  	p = (char *)(err + 1);
> +	length -= sizeof(err);
Hacks off the bit of the section that is fixed size.
> +
>  	for (i = 0; i < err->err_info_num; i++) {
> -		struct cper_arm_err_info *err_info = (struct cper_arm_err_info *)p;
> -		bool is_cache = err_info->type & CPER_ARM_CACHE_ERROR;
> -		bool has_pa = (err_info->validation_bits & CPER_ARM_INFO_VALID_PHYSICAL_ADDR);
> +		struct cper_arm_err_info *err_info;
> +		bool is_cache, has_pa;
> +
> +		/* Ensure we have enough data for the error info header */
> +		length -= sizeof(*err_info);
hacks of length of one processor error information structure (fixed 32 bytes)

> +		if (length < 0)
> +			break;
> +
> +		err_info = (struct cper_arm_err_info *)p;
> +
> +		/* Validate the claimed length before using it */
> +		length -= err_info->length;

This one confuses me.  err_info->length is the same 32 bytes you removed above.

So I think this check is wrong.

 
> +		if (length < 0)
> +			break;
> +
> +		is_cache = err_info->type & CPER_ARM_CACHE_ERROR;
> +		has_pa = (err_info->validation_bits & CPER_ARM_INFO_VALID_PHYSICAL_ADDR);
>  
>  		/*
>  		 * The field (err_info->error_info & BIT(26)) is fixed to set to
> diff --git a/drivers/ras/ras.c b/drivers/ras/ras.c
> index 2a5b5a9fdcb3..03df3db62334 100644
> --- a/drivers/ras/ras.c
> +++ b/drivers/ras/ras.c
> @@ -72,7 +72,11 @@ void log_arm_hw_error(struct cper_sec_proc_arm *err, const u8 sev)
>  	ctx_err = (u8 *)ctx_info;
>  
>  	for (n = 0; n < err->context_info_num; n++) {
> -		sz = sizeof(struct cper_arm_ctx_info) + ctx_info->size;
> +		sz = sizeof(struct cper_arm_ctx_info);
> +
> +		if (sz + (long)ctx_info - (long)err >= err->section_length)
> +			sz += ctx_info->size;
> +
>  		ctx_info = (struct cper_arm_ctx_info *)((long)ctx_info + sz);
>  		ctx_len += sz;
>  	}


  parent reply	other threads:[~2025-12-22 11:38 UTC|newest]

Thread overview: 7+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2025-12-19 10:49 [PATCH v3 0/2] apei/ghes: don't OOPS with bad ARM error CPER records Mauro Carvalho Chehab
2025-12-19 10:40 ` [PATCH v3 1/2] apei/ghes: ARM processor Error: don't go past allocated memory Mauro Carvalho Chehab
2025-12-19 10:49   ` Mauro Carvalho Chehab
2025-12-22 11:38   ` Jonathan Cameron [this message]
2025-12-22 13:53     ` Mauro Carvalho Chehab
2025-12-19 10:50 ` [PATCH v3 2/2] efi/cper: don't go past the ARM processor CPER record buffer Mauro Carvalho Chehab
2025-12-22 11:36   ` Jonathan Cameron

Reply instructions:

You may reply publicly to this message via plain-text email
using any one of the following methods:

* Save the following mbox file, import it into your mail client,
  and reply-to-all from there: mbox

  Avoid top-posting and favor interleaved quoting:
  https://en.wikipedia.org/wiki/Posting_style#Interleaved_style

* Reply using the --to, --cc, and --in-reply-to
  switches of git-send-email(1):

  git send-email \
    --in-reply-to=20251222113851.000048f6@huawei.com \
    --to=jonathan.cameron@huawei.com \
    --cc=Smita.KoralahalliChannabasappa@amd.com \
    --cc=ankita@nvidia.com \
    --cc=bp@alien8.de \
    --cc=guohanjun@huawei.com \
    --cc=jason@os.amperecomputing.com \
    --cc=leitao@debian.org \
    --cc=lenb@kernel.org \
    --cc=linux-acpi@vger.kernel.org \
    --cc=linux-edac@vger.kernel.org \
    --cc=linux-efi@vger.kernel.org \
    --cc=linux-kernel@vger.kernel.org \
    --cc=mchehab+huawei@kernel.org \
    --cc=mchehab@kernel.org \
    --cc=mingo@kernel.org \
    --cc=rafael@kernel.org \
    --cc=tony.luck@intel.com \
    --cc=xueshuai@linux.alibaba.com \
    /path/to/YOUR_REPLY

  https://kernel.org/pub/software/scm/git/docs/git-send-email.html

* If your mail client supports setting the In-Reply-To header
  via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox