On 03/04/2013 07:11 AM, Simon Horman wrote:
> [ Cc: linuxppc-dev at lists.ozlabs.org ]
>
> On Sun, Mar 03, 2013 at 01:06:00PM +0530, Suzuki K. Poulose wrote:
>> From: Suzuki K. Poulose <suzuki at in.ibm.com>
>>
>> uImage probe fills the entry point (ep) based on the load_addr
>> from the uImage headers. If we change the load_addr, we should
>> accordingly update the entry point.
>>
>> For ELF, calculate the offset of e_entry from the virtual start
>> address and add it to the physical start address to find the
>> physical address of kernel entry.
>>
>> i.e,
>> pa (e_entry) = pa(phdr[0].p_vaddr) + (e_entry - phdr[0].p_vaddr)
>> = kernel_addr + (e_entry - phdr[0].p_vaddr)
>
> Would it be possible for someone to provide a review of this change?
To make it bit more clear :
The entry point of the kernel is usually at 0 offset from the first
PT_LOAD section. The current code makes this assumption and uses the
pa(phdr[0].p_vaddr) as the kernel entry.
But this *may* not be true always, in such a case the kexec would fail.
While I fixed the uImage case, I thought it would be better to handle
the same case in ELF.
Btw, this calculation is not specific to ppc32.
Thanks
Suzuki
>
>>
>> Signed-off-by: Suzuki K. Poulose <suzuki at in.ibm.com>
>> Cc: Sebastian Andrzej Siewior <bigeasy at linutronix.de>
>> Cc: Matthew McClintock <msm at freescale.com>
>> ---
>> kexec/arch/ppc/kexec-elf-ppc.c | 12 ++++++++----
>> kexec/arch/ppc/kexec-uImage-ppc.c | 6 +++++-
>> 2 files changed, 13 insertions(+), 5 deletions(-)
>>
>> diff --git a/kexec/arch/ppc/kexec-elf-ppc.c b/kexec/arch/ppc/kexec-elf-ppc.c
>> index 8e408cc..5f63a64 100644
>> --- a/kexec/arch/ppc/kexec-elf-ppc.c
>> +++ b/kexec/arch/ppc/kexec-elf-ppc.c
>> @@ -397,10 +397,14 @@ int elf_ppc_load(int argc, char **argv, const char *buf, off_t len,
>> die("Error device tree not loadded to address it was expecting to be loaded too!\n");
>> }
>>
>> - /* set various variables for the purgatory ehdr.e_entry is a
>> - * virtual address, we can use kernel_addr which
>> - * should be the physical start address of the kernel */
>> - addr = kernel_addr;
>> + /*
>> + * set various variables for the purgatory.
>> + * ehdr.e_entry is a virtual address. we know physical start
>> + * address of the kernel (kernel_addr). Find the offset of
>> + * e_entry from the virtual start address(e_phdr[0].p_vaddr)
>> + * and calculate the actual physical address of the 'kernel entry'.
>> + */
>> + addr = kernel_addr + (ehdr.e_entry - ehdr.e_phdr[0].p_vaddr);
>> elf_rel_set_symbol(&info->rhdr, "kernel", &addr, sizeof(addr));
>>
>> addr = dtb_addr;
>> diff --git a/kexec/arch/ppc/kexec-uImage-ppc.c b/kexec/arch/ppc/kexec-uImage-ppc.c
>> index e0bc7bb..900cd16 100644
>> --- a/kexec/arch/ppc/kexec-uImage-ppc.c
>> +++ b/kexec/arch/ppc/kexec-uImage-ppc.c
>> @@ -159,15 +159,19 @@ static int ppc_load_bare_bits(int argc, char **argv, const char *buf,
>>
>> /*
>> * If the provided load_addr cannot be allocated, find a new
>> - * area.
>> + * area. Rebase the entry point based on the new load_addr.
>> */
>> if (!valid_memory_range(info, load_addr, load_addr + (len + _1MiB))) {
>> + int ep_offset = ep - load_addr;
>> +
>> load_addr = locate_hole(info, len + _1MiB, 0, 0, max_addr, 1);
>> if (load_addr == ULONG_MAX) {
>> printf("Can't allocate memory for kernel of len %ld\n",
>> len + _1MiB);
>> return -1;
>> }
>> +
>> + ep = load_addr + ep_offset;
>> }
>>
>> add_segment(info, buf, len, load_addr, len + _1MiB);
>>
>>
>> _______________________________________________
>> kexec mailing list
>> kexec at lists.infradead.org
>> http://lists.infradead.org/mailman/listinfo/kexec
>>
> _______________________________________________
> Linuxppc-dev mailing list
> Linuxppc-dev at lists.ozlabs.org
> https://lists.ozlabs.org/listinfo/linuxppc-dev
>
