[PATCH]acpi: enlarge NUM_FIXMAP_ACPI_PAGES from 4 to 5

[PATCH]acpi: enlarge NUM_FIXMAP_ACPI_PAGES from 4 to 5

[Zhangbo (Oscar)]

    In __acpi_map_table(), it suppose all ACPI tables may not exceed 4 pages.
    but nowadays ACPI tables, such as SRAT table, are large enough, On my
    BIOS board, the SRAT table achieves the size of 14428B, although it's
    above 3 pages and within 4pages, but in __acpi_map_table(), it maps pages
    to get the table. suppose the start address is near the end of the first page:

              1000B    4096B         4096B          4096B     1140B
               |___|_____________|______________|______________|____|

    although the total page is within 4pages , but it may be in fact across 5
    pages, thus the NUM_FIXMAP_ACPI_PAGES should be at least 5 nowadays. If
    not, xen would wrongly thinks no NUMA configuation could be found as that it
    could not get SRAT table.

diff --git a/xen/include/xen/acpi.h b/xen/include/xen/acpi.h
index 30ec0ee..bd616a1 100644
--- a/xen/include/xen/acpi.h
+++ b/xen/include/xen/acpi.h
@@ -43,7 +43,7 @@
  * Fixmap pages to reserve for ACPI boot-time tables (see asm-x86/fixmap.h or
  * asm-arm/config.h)
  */
-#define NUM_FIXMAP_ACPI_PAGES  4
+#define NUM_FIXMAP_ACPI_PAGES  5

 #define BAD_MADT_ENTRY(entry, end) (                                        \
                 (!(entry)) || (unsigned long)(entry) + sizeof(*(entry)) > (end) ||  \

_______________________________________________
Xen-devel mailing list
Xen-devel@lists.xen.org
https://lists.xen.org/xen-devel

Re: [PATCH]acpi: enlarge NUM_FIXMAP_ACPI_PAGES from 4 to 5

[Jan Beulich]

>>> On 04.05.17 at 04:59, <oscar.zhangbo@huawei.com> wrote:
> --- a/xen/include/xen/acpi.h
> +++ b/xen/include/xen/acpi.h
> @@ -43,7 +43,7 @@
>   * Fixmap pages to reserve for ACPI boot-time tables (see asm-x86/fixmap.h or
>   * asm-arm/config.h)
>   */
> -#define NUM_FIXMAP_ACPI_PAGES  4
> +#define NUM_FIXMAP_ACPI_PAGES  5

Well, this is the kind of fix I don't really like: You make things work for
you without thinking about others. If you found 4 pages aren't
enough, how likely is it that soon someone will find 5 aren't enough
either? IOW, short of eliminating the fixed upper bound altogether,
you should at least add some slack for the foreseeable future. For
this it may also help to estimate the theoretical upper limit of SRAT
(and perhaps other affected tables) for systems currently around
plus, again, some slack. As you may have concluded already it
would therefore also have helped if you had indicated what size a
system you see this relatively large SRAT on.

Jan


_______________________________________________
Xen-devel mailing list
Xen-devel@lists.xen.org
https://lists.xen.org/xen-devel

Re: [PATCH]acpi: enlarge NUM_FIXMAP_ACPI_PAGES from 4 to 5

[Zhangbo (Oscar)]

>> --- a/xen/include/xen/acpi.h
>> +++ b/xen/include/xen/acpi.h
>> @@ -43,7 +43,7 @@
>>   * Fixmap pages to reserve for ACPI boot-time tables (see asm-x86/fixmap.h
>or
>>   * asm-arm/config.h)
>>   */
>> -#define NUM_FIXMAP_ACPI_PAGES  4
>> +#define NUM_FIXMAP_ACPI_PAGES  5
>
>Well, this is the kind of fix I don't really like: You make things work for
>you without thinking about others. If you found 4 pages aren't
>enough, how likely is it that soon someone will find 5 aren't enough
>either? IOW, short of eliminating the fixed upper bound altogether,
>you should at least add some slack for the foreseeable future. For
>this it may also help to estimate the theoretical upper limit of SRAT
>(and perhaps other affected tables) for systems currently around
>plus, again, some slack. As you may have concluded already it
>would therefore also have helped if you had indicated what size a
>system you see this relatively large SRAT on.
>
>Jan

Thanks, Jan, I'm checking the proper number, and patch v2 is on its way.

_______________________________________________
Xen-devel mailing list
Xen-devel@lists.xen.org
https://lists.xen.org/xen-devel

Re: [PATCH]acpi: enlarge NUM_FIXMAP_ACPI_PAGES from 4 to 5

[Zhangbo (Oscar)]

Ping

>    In __acpi_map_table(), it suppose all ACPI tables may not exceed 4 pages.
>    but nowadays ACPI tables, such as SRAT table, are large enough, On my
>    BIOS board, the SRAT table achieves the size of 14428B, although it's
>    above 3 pages and within 4pages, but in __acpi_map_table(), it maps
>pages
>    to get the table. suppose the start address is near the end of the first page:
>
>          1000B    4096B         4096B          4096B   1140B
>
>         |___|_____________|______________|______________|____|
>
>    although the total page is within 4pages , but it may be in fact across 5
>    pages, thus the NUM_FIXMAP_ACPI_PAGES should be at least 5 nowadays.
>If
>    not, xen would wrongly thinks no NUMA configuation could be found as that
>it
>    could not get SRAT table.
>
>diff --git a/xen/include/xen/acpi.h b/xen/include/xen/acpi.h
>index 30ec0ee..bd616a1 100644
>--- a/xen/include/xen/acpi.h
>+++ b/xen/include/xen/acpi.h
>@@ -43,7 +43,7 @@
>  * Fixmap pages to reserve for ACPI boot-time tables (see asm-x86/fixmap.h or
>  * asm-arm/config.h)
>  */
>-#define NUM_FIXMAP_ACPI_PAGES  4
>+#define NUM_FIXMAP_ACPI_PAGES  5
>
> #define BAD_MADT_ENTRY(entry, end)
>(                                        \
>                 (!(entry)) || (unsigned long)(entry) + sizeof(*(entry)) > (end)
>||  \

_______________________________________________
Xen-devel mailing list
Xen-devel@lists.xen.org
https://lists.xen.org/xen-devel

Re: [PATCH]acpi: enlarge NUM_FIXMAP_ACPI_PAGES from 4 to 5

[Julien Grall]

Hello,

On 05/10/2017 04:29 AM, Zhangbo (Oscar) wrote:
> Ping

Most of the maintainers have nowadays filters to flag e-mail with them 
CCed and look less often to rest. If you want to draw attention to your 
patch, you should CC relevant maintainers (you can find them in 
MAINTAINERS).

>
>>    In __acpi_map_table(), it suppose all ACPI tables may not exceed 4 pages.

The spaces at the beginning of every lines in the commit message are not 
necessary.

>>    but nowadays ACPI tables, such as SRAT table, are large enough, On my
>>    BIOS board, the SRAT table achieves the size of 14428B, although it's
>>    above 3 pages and within 4pages, but in __acpi_map_table(), it maps
>> pages
>>    to get the table. suppose the start address is near the end of the first page:
>>
>>          1000B    4096B         4096B          4096B   1140B
>>
>>         |___|_____________|______________|______________|____|
>>
>>    although the total page is within 4pages , but it may be in fact across 5

s/4pages/4 pages/

>>    pages, thus the NUM_FIXMAP_ACPI_PAGES should be at least 5 nowadays.
>> If
>>    not, xen would wrongly thinks no NUMA configuation could be found as that

s/configuation/configuration/

>> it
>>    could not get SRAT table.

Your signed-off-by is missing here.

>>
>> diff --git a/xen/include/xen/acpi.h b/xen/include/xen/acpi.h
>> index 30ec0ee..bd616a1 100644
>> --- a/xen/include/xen/acpi.h
>> +++ b/xen/include/xen/acpi.h
>> @@ -43,7 +43,7 @@
>>  * Fixmap pages to reserve for ACPI boot-time tables (see asm-x86/fixmap.h or
>>  * asm-arm/config.h)
>>  */
>> -#define NUM_FIXMAP_ACPI_PAGES  4
>> +#define NUM_FIXMAP_ACPI_PAGES  5
>>
>> #define BAD_MADT_ENTRY(entry, end)
>> (                                        \
>>                 (!(entry)) || (unsigned long)(entry) + sizeof(*(entry)) > (end)
>> ||  \
>
> _______________________________________________
> Xen-devel mailing list
> Xen-devel@lists.xen.org
> https://lists.xen.org/xen-devel
>

Cheers,

-- 
Julien Grall

_______________________________________________
Xen-devel mailing list
Xen-devel@lists.xen.org
https://lists.xen.org/xen-devel