Re: [Xen-devel] [PATCH 2/2] efi/boot: Don't free ebmalloc area at all

["Jan Beulich" <JBeulich@xxxxxxxx>]

>>> On 01.03.17 at 10:00, <daniel.kiper@xxxxxxxxxx> wrote:
> On Wed, Mar 01, 2017 at 01:02:55AM -0700, Jan Beulich wrote:
>> >>> On 28.02.17 at 20:09, <andrew.cooper3@xxxxxxxxxx> wrote:
>> > On 28/02/17 19:01, Daniel Kiper wrote:
>> >> Looking at this I do not have a better idea for fix off the top of my 
>> >> head.
>>
>> Well, one better solution was already suggested: Put it in .init.data and
> 
> This will increase Xen binary size. So, I prefer to leave it in BSS if 
> possible.

I don't think that's very important a criteria.

>> suppress the freeing of the used part of it. The other option is to put
>> a section along the lines of Linux'es .brk _after_ .bss (whether ahead
>> of or after _end would need to be seen), replacing the uses of
>> __2M_rwdata_end by a runtime calculated value (and freeing space
>> past the used portion accordingly, without shattering large pages if
>> used).
> 
> OK, make sense for me. However, from a bootloader point of view it should be
> still part of BSS. Otherwise, the bootloader can put the end of Xen image too
> close to reserved/no ram/etc. region. This does not seem important right now
> but after applying the rest of my patch series it can make a pain.

I'm afraid I don't understand: It needs to be part of the ELF
load segments or PE32+ loaded sections, but whether that's
part of .bss or an adjacent @nobits section at the source level
shouldn't matter at all.

Jan


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