It was <2020-06-01 pon 16:10>, when Russell King - ARM Linux admin wrote:
> On Mon, Jun 01, 2020 at 04:56:40PM +0200, Lukasz Stelmach wrote:
>> It was <2020-06-01 pon 15:46>, when Russell King - ARM Linux admin wrote:
>> > On Mon, Jun 01, 2020 at 04:27:50PM +0200, Łukasz Stelmach wrote:
>> >> Move the definition of malloc pool size of the decompressor to
>> >> a single place. This value will be exposed later for kexec_file loader.
>> >> 
>> >> Signed-off-by: Łukasz Stelmach <l.stelmach@samsung.com>
>> >> ---
>> >>  arch/arm/boot/compressed/Makefile | 2 ++
>> >>  arch/arm/boot/compressed/head.S   | 6 ++++--
>> >>  2 files changed, 6 insertions(+), 2 deletions(-)
>> >> 
>> >> diff --git a/arch/arm/boot/compressed/Makefile b/arch/arm/boot/compressed/Makefile
>> >> index 9c11e7490292..b3594cd1588c 100644
>> >> --- a/arch/arm/boot/compressed/Makefile
>> >> +++ b/arch/arm/boot/compressed/Makefile
>> >> @@ -125,6 +125,8 @@ KBSS_SZ = $(shell echo $$(($$($(NM) $(obj)/../../../../vmlinux | \
>> >>  		sed -n -e 's/^\([^ ]*\) [AB] __bss_start$$/-0x\1/p' \
>> >>  		       -e 's/^\([^ ]*\) [AB] __bss_stop$$/+0x\1/p') )) )
>> >>  LDFLAGS_vmlinux = --defsym _kernel_bss_size=$(KBSS_SZ)
>> >> +# malloc pool size
>> >> +LDFLAGS_vmlinux += --defsym _malloc_size=0x10000
>> >>  # Supply ZRELADDR to the decompressor via a linker symbol.
>> >>  ifneq ($(CONFIG_AUTO_ZRELADDR),y)
>> >>  LDFLAGS_vmlinux += --defsym zreladdr=$(ZRELADDR)
>> >> diff --git a/arch/arm/boot/compressed/head.S b/arch/arm/boot/compressed/head.S
>> >> index e8e1c866e413..dcc1afa60fb9 100644
>> >> --- a/arch/arm/boot/compressed/head.S
>> >> +++ b/arch/arm/boot/compressed/head.S
>> >> @@ -309,7 +309,8 @@ restart:	adr	r0, LC0
>> >>  #ifndef CONFIG_ZBOOT_ROM
>> >>  		/* malloc space is above the relocated stack (64k max) */
>> >>  		add	sp, sp, r0
>> >> -		add	r10, sp, #0x10000
>> >> +		ldr	r10, =_malloc_size
>> >> +		add	r10, r10, sp
>> >
>> > This says "locate _malloc_size in a literal pool somewhere, and load it
>> > using a PC-relative offset".  Are you sure that the literal pool is
>> > sensibly located?
>> >
>> > Would it be better to use a definition for this?
>> 
>> I've followed ZRELADDR way. I think both values should be handled the
>> same way (it makes it easier to comprehend IMHO). Is there any reason
>> not to?  Should I change ZRELADDR for v2 too?
>
> There's a good reason.  ZRELADDR can be a constant that does not fit
> into the ARMs immediate constants (8 bits of significant data plus
> a multiple of a 2-bit shift), whereas the size of the malloc space
> is guaranteed to fit.  So, ZRELADDR has to use a literal load.
> This doesn't.

Indeed this is how TEXT_OFFSET works. Done.

-- 
Łukasz Stelmach
Samsung R&D Institute Poland
Samsung Electronics