| lists.openwall.net | lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC | |
|
Open Source and information security mailing list archives
| ||
|
Message-ID: <CAMj1kXEe37g=1DEQcWFEDCLRihZLX14Kso0dwon71W+5jqyDrA@mail.gmail.com>
Date: Wed, 5 Apr 2023 10:19:50 +0200
From: Ard Biesheuvel <ardb@...nel.org>
To: Huacai Chen <chenhuacai@...nel.org>
Cc: Youling Tang <tangyouling@...ngson.cn>,
Huacai Chen <chenhuacai@...ngson.cn>,
linux-efi@...r.kernel.org, loongarch@...ts.linux.dev,
Xuefeng Li <lixuefeng@...ngson.cn>,
Xuerui Wang <kernel@...0n.name>, linux-kernel@...r.kernel.org,
loongson-kernel@...ts.loongnix.cn
Subject: Re: [PATCH] efi/loongarch: Reintroduce efi_relocate_kernel() to
relocate kernel
On Wed, 5 Apr 2023 at 10:14, Huacai Chen <chenhuacai@...nel.org> wrote:
>
> Hi, Ard,
>
> What is the status of this patch now?
>
I'd still like to understand how zboot is affected. It has
static unsigned long alloc_preferred_address(unsigned long alloc_size)
{
#ifdef EFI_KIMG_PREFERRED_ADDRESS
efi_physical_addr_t efi_addr = EFI_KIMG_PREFERRED_ADDRESS;
if (efi_bs_call(allocate_pages, EFI_ALLOCATE_ADDRESS, EFI_LOADER_DATA,
alloc_size / EFI_PAGE_SIZE, &efi_addr) == EFI_SUCCESS)
return efi_addr;
#endif
return ULONG_MAX;
}
and if I understand you correctly, we never needed this in the first place?
(I understand this is a separate issue, but it would be nice to get
rid of EFI_KIMG_PREFERRED_ADDRESS entirely if we can.)
> On Wed, Mar 1, 2023 at 8:06 PM Huacai Chen <chenhuacai@...nel.org> wrote:
> >
> > Hi, Youling,
> >
> > On Wed, Mar 1, 2023 at 6:08 PM Youling Tang <tangyouling@...ngson.cn> wrote:
> > >
> > > Hi, Huacai
> > >
> > > On 03/01/2023 05:30 PM, Huacai Chen wrote:
> > > > Hi, Youling,
> > > >
> > > > On Wed, Mar 1, 2023 at 5:21 PM Youling Tang <tangyouling@...ngson.cn> wrote:
> > > >>
> > > >> Hi, Huacai
> > > >>
> > > >> On 03/01/2023 04:55 PM, Huacai Chen wrote:
> > > >>> Since Linux-6.3, LoongArch supports PIE kernel now, so let's reintroduce
> > > >>> efi_relocate_kernel() to relocate the core kernel.
> > > >>
> > > >> If we strongly rely on the PIE feature after the modification, we need
> > > >> to modify it as follows,
> > > > There is no strong rely, efi_relocate_kernel() can also handle the
> > > > non-PIE kernel.
> > >
> > > IMHO, if EFI_KIMG_PREFERRED_ADDRESS is no longer defined as
> > > `PHYSADDR(VMLINUX_LOAD_ADDRESS)` (physical link address), we need to
> > > relocate through the RELOCATABLE function, so it is recommended to
> > > select RELOCATABLE if EFI_STUB is enabled
> > We can select RELOCATABLE when we modify EFI_KIMG_PREFERRED_ADDRESS.
> > Before that I prefer to give a chance to build a non-PIE kernel.
> >
> > Huacai
> > >
> > > Youling.
> > > >
> > > > Huacai
> > > >>
> > > >> --- a/arch/loongarch/Kconfig
> > > >> +++ b/arch/loongarch/Kconfig
> > > >> @@ -137,6 +137,7 @@ config LOONGARCH
> > > >> select PCI_MSI_ARCH_FALLBACKS
> > > >> select PCI_QUIRKS
> > > >> select PERF_USE_VMALLOC
> > > >> + select RELOCATABLE
> > > >> select RTC_LIB
> > > >> select SMP
> > > >> select SPARSE_IRQ
> > > >>
> > > >> or:
> > > >>
> > > >> --- a/arch/loongarch/Kconfig
> > > >> +++ b/arch/loongarch/Kconfig
> > > >> @@ -360,6 +360,7 @@ config EFI_STUB
> > > >> default y
> > > >> depends on EFI
> > > >> select EFI_GENERIC_STUB
> > > >> + select RELOCATABLE
> > > >> help
> > > >>
> > > >> Youling.
> > > >>
> > > >>>
> > > >>> Signed-off-by: Huacai Chen <chenhuacai@...ngson.cn>
> > > >>> ---
> > > >>> drivers/firmware/efi/libstub/loongarch-stub.c | 24 ++++++-------------
> > > >>> 1 file changed, 7 insertions(+), 17 deletions(-)
> > > >>>
> > > >>> diff --git a/drivers/firmware/efi/libstub/loongarch-stub.c b/drivers/firmware/efi/libstub/loongarch-stub.c
> > > >>> index eee7ed43cdfb..72c71ae201f0 100644
> > > >>> --- a/drivers/firmware/efi/libstub/loongarch-stub.c
> > > >>> +++ b/drivers/firmware/efi/libstub/loongarch-stub.c
> > > >>> @@ -21,26 +21,16 @@ efi_status_t handle_kernel_image(unsigned long *image_addr,
> > > >>> efi_loaded_image_t *image,
> > > >>> efi_handle_t image_handle)
> > > >>> {
> > > >>> - int nr_pages = round_up(kernel_asize, EFI_ALLOC_ALIGN) / EFI_PAGE_SIZE;
> > > >>> - efi_physical_addr_t kernel_addr = EFI_KIMG_PREFERRED_ADDRESS;
> > > >>> efi_status_t status;
> > > >>> + unsigned long kernel_addr = 0;
> > > >>>
> > > >>> - /*
> > > >>> - * Allocate space for the kernel image at the preferred offset. This is
> > > >>> - * the only location in memory from where we can execute the image, so
> > > >>> - * no point in falling back to another allocation.
> > > >>> - */
> > > >>> - status = efi_bs_call(allocate_pages, EFI_ALLOCATE_ADDRESS,
> > > >>> - EFI_LOADER_DATA, nr_pages, &kernel_addr);
> > > >>> - if (status != EFI_SUCCESS)
> > > >>> - return status;
> > > >>> -
> > > >>> - *image_addr = EFI_KIMG_PREFERRED_ADDRESS;
> > > >>> - *image_size = kernel_asize;
> > > >>> + kernel_addr = (unsigned long)&kernel_offset - kernel_offset;
> > > >>> +
> > > >>> + status = efi_relocate_kernel(&kernel_addr, kernel_fsize, kernel_asize,
> > > >>> + EFI_KIMG_PREFERRED_ADDRESS, efi_get_kimg_min_align(), 0x0);
> > > >>>
> > > >>> - memcpy((void *)EFI_KIMG_PREFERRED_ADDRESS,
> > > >>> - (void *)&kernel_offset - kernel_offset,
> > > >>> - kernel_fsize);
> > > >>> + *image_addr = kernel_addr;
> > > >>> + *image_size = kernel_asize;
> > > >>>
> > > >>> return status;
> > > >>> }
> > > >>>
> > > >>
> > >
> > >
Powered by blists - more mailing lists