Re: [PATCH] Documentation: x86: exception-tables: document CONFIG_BUILDTIME_TABLE_SORT

["H. Peter Anvin" <hpa@zytor.com>]

On 2020-03-23 16:22, Nick Desaulniers wrote:
> Provide more information about __ex_table sorting post link.
> 
> The exception tables and fixup tables use a commonly recurring pattern
> in the kernel of storing the address of labels as date in custom ELF
> sections, then finding these sections, iterating elements within them,
> and possibly revisiting them or modifying the data at these addresses.
> 
> Sorting readonly arrays to minimize runtime penalties is quite clever.
> 
> Signed-off-by: Nick Desaulniers <ndesaulniers@google.com>
> ---
>  Documentation/x86/exception-tables.rst | 9 +++++++++
>  1 file changed, 9 insertions(+)
> 
> diff --git a/Documentation/x86/exception-tables.rst b/Documentation/x86/exception-tables.rst
> index ed6d4b0cf62c..15455b2f7ba8 100644
> --- a/Documentation/x86/exception-tables.rst
> +++ b/Documentation/x86/exception-tables.rst
> @@ -257,6 +257,9 @@ the fault, in our case the actual value is c0199ff5:
>  the original assembly code: > 3:      movl $-14,%eax
>  and linked in vmlinux     : > c0199ff5 <.fixup+10b5> movl   $0xfffffff2,%eax
>  
> +If the fixup was able to handle the exception, control flow may be returned
> +to the instruction after the one that triggered the fault, ie. local label 2b.
> +
>  The assembly code::
>  
>   > .section __ex_table,"a"
> @@ -344,3 +347,9 @@ pointer which points to one of:
>       it as special.
>  
>  More functions can easily be added.
> +
> +CONFIG_BUILDTIME_TABLE_SORT allows the __ex_table section to be sorted post
> +link of the kernel image, via a host utility scripts/sorttable. It will set the
> +symbol main_extable_sort_needed to 0, avoiding sorting the __ex_table section
> +at boot time. With the exception table sorted, at runtime when an exception
> +occurs we can quickly lookup the __ex_table entry via binary search.
> 

It is more than that. It not only saves the boot execution time needed
to sort the table, but it is required for early exception handling to
work -- and in the x86 code, we use the exception handling *extremely*
early (on i386 before paging is even turned on!), long before the kernel
would have had any opportunity to sort it.

So it is not just performance; it is also a correctness issue.

	-hpa