[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Message-ID: <Zw6b3V5Mk2tIGmy5@lpieralisi>
Date: Tue, 15 Oct 2024 18:44:13 +0200
From: Lorenzo Pieralisi <lpieralisi@...nel.org>
To: Zheng Zengkai <zhengzengkai@...wei.com>
Cc: guohanjun@...wei.com, sudeep.holla@....com, mark.rutland@....com,
maz@...nel.org, rafael@...nel.org, lenb@...nel.org,
daniel.lezcano@...aro.org, tglx@...utronix.de,
linux-acpi@...r.kernel.org, linux-arm-kernel@...ts.infradead.org,
linux-kernel@...r.kernel.org
Subject: Re: [PATCH v3] ACPI: GTDT: Tighten the check for the array of
platform timer structures
On Tue, Oct 15, 2024 at 11:26:02PM +0800, Zheng Zengkai wrote:
> As suggested by Marc and Lorenzo, first we need to check whether the
I would just describe the change, the tags and Link: are there to
describe this patch history.
> platform_timer entry pointer is within gtdt bounds (< gtdt_end) before
> de-referencing what it points at to detect the length of the platform
> timer struct and then check that the length of current platform_timer
> struct is also valid, i.e. the length is not zero and within gtdt_end.
> Now next_platform_timer() only checks against gtdt_end for the entry of
> subsequent platform timer without checking the length of it and will
> not report error if the check failed and the existing check in function
> acpi_gtdt_init() is also not enough.
>
> Modify the for_each_platform_timer() iterator and use it combined with
> a dedicated check function platform_timer_valid() to do the check
> against table length (gtdt_end) for each element of platform timer
> array in function acpi_gtdt_init(), making sure that both their entry
> and length actually fit in the table.
>
> Suggested-by: Lorenzo Pieralisi <lpieralisi@...nel.org>
> Co-developed-by: Marc Zyngier <maz@...nel.org>
> Signed-off-by: Marc Zyngier <maz@...nel.org>
> Signed-off-by: Zheng Zengkai <zhengzengkai@...wei.com>
> ---
> Changes in v3:
> - based on Marc's patch and reuse the for_each_platform_timer() loop
>
> Changes in v2:
> - Check against gtdt_end for both entry and len of each array element
> Link to v2: https://lore.kernel.org/linux-arm-kernel/20241012085343.6594-1-zhengzengkai@huawei.com/
>
> Link to v1: https://lore.kernel.org/all/20241010144703.113728-1-zhengzengkai@huawei.com/
> ---
> drivers/acpi/arm64/gtdt.c | 32 +++++++++++++++++++++-----------
> 1 file changed, 21 insertions(+), 11 deletions(-)
>
> diff --git a/drivers/acpi/arm64/gtdt.c b/drivers/acpi/arm64/gtdt.c
> index c0e77c1c8e09..3583c99afb0d 100644
> --- a/drivers/acpi/arm64/gtdt.c
> +++ b/drivers/acpi/arm64/gtdt.c
> @@ -36,19 +36,25 @@ struct acpi_gtdt_descriptor {
>
> static struct acpi_gtdt_descriptor acpi_gtdt_desc __initdata;
>
> -static inline __init void *next_platform_timer(void *platform_timer)
> +static __init bool platform_timer_valid(void *platform_timer)
> {
> struct acpi_gtdt_header *gh = platform_timer;
>
> - platform_timer += gh->length;
> - if (platform_timer < acpi_gtdt_desc.gtdt_end)
> - return platform_timer;
> + return (platform_timer >= (void *)(acpi_gtdt_desc.gtdt + 1) &&
> + platform_timer < acpi_gtdt_desc.gtdt_end &&
> + gh->length != 0 &&
> + platform_timer + gh->length <= acpi_gtdt_desc.gtdt_end);
> +}
> +
> +static __init void *next_platform_timer(void *platform_timer)
> +{
> + struct acpi_gtdt_header *gh = platform_timer;
>
> - return NULL;
> + return platform_timer + gh->length;
> }
>
> -#define for_each_platform_timer(_g) \
> - for (_g = acpi_gtdt_desc.platform_timer; _g; \
> +#define for_each_platform_timer(_g, first_entry) \
> + for (_g = first_entry; platform_timer_valid(_g); \
> _g = next_platform_timer(_g))
>
> static inline bool is_timer_block(void *platform_timer)
> @@ -155,8 +161,9 @@ bool __init acpi_gtdt_c3stop(int type)
> int __init acpi_gtdt_init(struct acpi_table_header *table,
> int *platform_timer_count)
> {
> - void *platform_timer;
> + void *platform_timer, *tmp;
It makes more sense - thank you and Marc.
Nit: you don't really need another pointer (ie tmp) but you may keep
it if that makes the code clearer - all you need to do is using
platform_timer as an iterator and initialize
acpi_gtdt_desc.platform_timer = (void *)gtdt + gtdt->platform_timer_offset;
if all checks passed (you are using tmp just because after the loop
platform_timer can't be used to initialize acpi_gtdt_desc.platform_timer).
Reviewed-by: Lorenzo Pieralisi <lpieralisi@...nel.org>
(now let's see if this survives GTDTs out there :))
> struct acpi_table_gtdt *gtdt;
> + int cnt = 0;
>
> gtdt = container_of(table, struct acpi_table_gtdt, header);
> acpi_gtdt_desc.gtdt = gtdt;
> @@ -177,7 +184,10 @@ int __init acpi_gtdt_init(struct acpi_table_header *table,
> }
>
> platform_timer = (void *)gtdt + gtdt->platform_timer_offset;
> - if (platform_timer < (void *)table + sizeof(struct acpi_table_gtdt)) {
> + for_each_platform_timer(tmp, platform_timer)
> + cnt++;
> +
> + if (cnt != gtdt->platform_timer_count) {
> pr_err(FW_BUG "invalid timer data.\n");
> return -EINVAL;
> }
> @@ -305,7 +315,7 @@ int __init acpi_arch_timer_mem_init(struct arch_timer_mem *timer_mem,
> void *platform_timer;
>
> *timer_count = 0;
> - for_each_platform_timer(platform_timer) {
> + for_each_platform_timer(platform_timer, acpi_gtdt_desc.platform_timer) {
> if (is_timer_block(platform_timer)) {
> ret = gtdt_parse_timer_block(platform_timer, timer_mem);
> if (ret)
> @@ -398,7 +408,7 @@ static int __init gtdt_sbsa_gwdt_init(void)
> if (ret || !timer_count)
> goto out_put_gtdt;
>
> - for_each_platform_timer(platform_timer) {
> + for_each_platform_timer(platform_timer, acpi_gtdt_desc.platform_timer) {
> if (is_non_secure_watchdog(platform_timer)) {
> ret = gtdt_import_sbsa_gwdt(platform_timer, gwdt_count);
> if (ret)
> --
> 2.20.1
>
Powered by blists - more mailing lists