Hi Heinrich,
On Wed, 16 Oct 2024 at 19:57, Heinrich Schuchardt <xypron.g...@gmx.de> wrote:
>
> On 17.09.24 11:02, Ilias Apalodimas wrote:
> > The EFI spec defines special-purpose memory in ยง7.2. That memory
> > serves as a hint to the OS to avoid allocating this memory for core
> > OS data or code that can not be relocated. So let's ignore it when
> > allocating from conventional memory.
> >
> > Signed-off-by: Ilias Apalodimas <ilias.apalodi...@linaro.org>
> > ---
> > lib/efi_loader/efi_memory.c | 6 +++++-
> > 1 file changed, 5 insertions(+), 1 deletion(-)
> >
> > diff --git a/lib/efi_loader/efi_memory.c b/lib/efi_loader/efi_memory.c
> > index c6f1dd09456e..74732e37f8aa 100644
> > --- a/lib/efi_loader/efi_memory.c
> > +++ b/lib/efi_loader/efi_memory.c
> > @@ -422,7 +422,8 @@ static efi_status_t efi_check_allocated(u64 addr, bool
> > must_be_allocated)
> >
> > if (addr >= start && addr < end) {
> > if (must_be_allocated ^
> > - (item->desc.type == EFI_CONVENTIONAL_MEMORY))
> > + (item->desc.type == EFI_CONVENTIONAL_MEMORY) &&
> > + !(item->desc.attribute & EFI_MEMORY_SP))
> > return EFI_SUCCESS;
> > else
> > return EFI_NOT_FOUND;
> > @@ -460,6 +461,9 @@ static uint64_t efi_find_free_memory(uint64_t len,
> > uint64_t max_addr)
>
> Now that Sushosh's LMB patches are merged this function does not exist
> anymore.
>
> I think we should next move the EFI memory map to LMB to avoid double
> book keeping.
>
Yea, I sent that patch before LMB as a fix, but I agreed to wait till
the LMB patches get merged. Now that they are in we'll rework it.
Thanks
/Ilias
> Best regards
>
> Heinrich
>
> > if (desc->type != EFI_CONVENTIONAL_MEMORY)
> > continue;
> >
> > + if (desc->attribute & EFI_MEMORY_SP)
> > + continue;
> > +
> > /* Out of bounds for max_addr */
> > if ((ret + len) > max_addr)
> > continue;
>
