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] [day] [month] [year] [list]
Date:   Wed, 29 Jun 2022 16:27:53 +0200
From:   "Rafael J. Wysocki" <rafael@...nel.org>
To:     James Liu <james.liu@....com>
Cc:     "Rafael J. Wysocki" <rafael@...nel.org>,
        Len Brown <lenb@...nel.org>,
        ACPI Devel Maling List <linux-acpi@...r.kernel.org>,
        Linux Kernel Mailing List <linux-kernel@...r.kernel.org>,
        rwright@....com, david.chang@....com, clayc@....com
Subject: Re: [PATCH v2] ACPI: OSL: Fix the memory mapping of an ACPI GAS that
 addresses a data structure

On Sun, May 22, 2022 at 11:44 PM <james.liu@....com> wrote:
>
> From: James Liu <james.liu@....com>
>
>     Modify acpi_os_map_generic_address() & acpi_os_unmap_generic_address()
>     to correctly handle cases that a GAS table (i.e., Table 5.1, ACPI 6.4)
>     is used to address a data structure; in the case, the GAS has the field
>     of "Register Bit Width" equal to 0.
>
>     For example, "Injection Instruction Entry" (Table 18.25, ACPI 6.4) has
>     a RegisterRegion field that is a GAS that points to a data structure
>     SET_ERROR_TYPE_WITH_ADDRESS (Table 18.30), which is required when using
>     EINJ (Error Injection module).
>
>     This fix preserves a fairly sufficient memory space (i.e. page size) to

The hard-coded PAGE_SIZE allocation size is kind of a concern, because
the spec doesn't seem to limit the size of the structure pointed to by
the GAS to a page.

Also it is not clear from the spec that the RegisterRegion field in
Table 18.25 points to a structure.  On the contrary, Section 18.6.2
says that it points to a register.

In any case, I don't think it is OK to use
acpi_os_map_generic_address() to map an address from a GAS pointing to
a data structure.  acpi_os_map_memory() appears to be the right
interface to map such memory regions.

>     store the data structure to prevent EINJ module from loading failure if
>     platform firmware can correctly support Injection Instruction Entry in
>     an EINJ table.
>
> Signed-off-by: James Liu <james.liu@....com>
> ---
>  drivers/acpi/osl.c | 27 +++++++++++++++++++++++----
>  1 file changed, 23 insertions(+), 4 deletions(-)
>
> diff --git a/drivers/acpi/osl.c b/drivers/acpi/osl.c
> index 45c5c0e45..99f987c8c 100644
> --- a/drivers/acpi/osl.c
> +++ b/drivers/acpi/osl.c
> @@ -452,14 +452,20 @@ EXPORT_SYMBOL_GPL(acpi_os_unmap_memory);
>
>  void __iomem *acpi_os_map_generic_address(struct acpi_generic_address *gas)
>  {
> -       u64 addr;
> +       u64 addr = 0;
>
>         if (gas->space_id != ACPI_ADR_SPACE_SYSTEM_MEMORY)
>                 return NULL;
>
> +       /* Handle a case that GAS is used to address an ACPI data structure */
> +       if (!gas->bit_width) {
> +               pr_info("An ACPI data structure at 0x%llx is mapped\n", addr);
> +               return  acpi_os_map_iomem(addr, PAGE_SIZE);
> +       }
> +
>         /* Handle possible alignment issues */
>         memcpy(&addr, &gas->address, sizeof(addr));
> -       if (!addr || !gas->bit_width)
> +       if (!addr)
>                 return NULL;
>
>         return acpi_os_map_iomem(addr, gas->bit_width / 8);
> @@ -468,15 +474,28 @@ EXPORT_SYMBOL(acpi_os_map_generic_address);
>
>  void acpi_os_unmap_generic_address(struct acpi_generic_address *gas)
>  {
> -       u64 addr;
> +       u64 addr = 0;
>         struct acpi_ioremap *map;
>
>         if (gas->space_id != ACPI_ADR_SPACE_SYSTEM_MEMORY)
>                 return;
>
> +       /* Handle a case that GAS is used to address an ACPI data structure */
> +       if (!gas->bit_width) {
> +               pr_info("An ACPI data structure at 0x%llx is unmapped\n", addr);
> +               mutex_lock(&acpi_ioremap_lock);
> +               map = acpi_map_lookup(addr, PAGE_SIZE);
> +               if (!map) {
> +                       mutex_unlock(&acpi_ioremap_lock);
> +                       return;
> +               }
> +               acpi_os_drop_map_ref(map);
> +               mutex_unlock(&acpi_ioremap_lock);
> +       }
> +
>         /* Handle possible alignment issues */
>         memcpy(&addr, &gas->address, sizeof(addr));
> -       if (!addr || !gas->bit_width)
> +       if (!addr)
>                 return;
>
>         mutex_lock(&acpi_ioremap_lock);
> --
> 2.25.1
>

Powered by blists - more mailing lists

Powered by Openwall GNU/*/Linux Powered by OpenVZ