From mboxrd@z Thu Jan 1 00:00:00 1970 From: panand@redhat.com (Pratyush Anand) Date: Fri, 29 Jul 2016 09:05:14 +0530 Subject: [PATCH v2 4/4] arm64: Add support for binary image files In-Reply-To: <1469729083.16559.26.camel@infradead.org> References: <20160728040125.GA27150@localhost.localdomain> <1469729083.16559.26.camel@infradead.org> Message-ID: <20160729033514.GC27150@localhost.localdomain> To: linux-arm-kernel@lists.infradead.org List-Id: linux-arm-kernel.lists.infradead.org Hi Geoff, On 28/07/2016:11:04:43 AM, Geoff Levand wrote: > Hi Pratyush, > > On Thu, 2016-07-28 at 09:31 +0530, Pratyush Anand wrote: > > On 26/07/2016:08:18:57 PM, Geoff Levand wrote: > > > You removed page_offset calculation in rework. > > I am unable to understand that how arm64_mem.page_offset will be filled up now > > for binary image case. > > > > I think, we still need to keep page_offset user input. > > The current kernel linker scripts are setup to output virtual > address values to the vmlinux elf file. > > To load the elf sections of vmlinux using the provided elf > file loader, elf_exec_load(), we need to convert those elf > header values to physical addresses since we work with > physical addresses in the kernel's kimage structure. > > That conversion is > > p = v - page_offset + phys_offset; > > We get phys_offset from the memory ranges, and we calculate > page_offset from the elf header info. > > As for the binary image file, we load that with > add_segment_phys_virt(), which can put the binary data of the > Image file into memory at a physical address we specify: > > image_base = phys_offset + text_offset. > > phys_offset is again from the memory ranges, and text_offset > from the arm64 image header. > > > > You may take following > > two patches which does this, or you may leave binary image support patch which I > > will send after your patches are merged. > > > > https://github.com/pratyushanand/kexec-tools/commit/5b7e49a75d1d6cd4ac846f50ff10275fd54cb545 > > This one may be useful, for debugging if anything. > > > https://github.com/pratyushanand/kexec-tools/commit/a0ce0ce673755c4061c1f081170a3a75dfa1d1fb > > We do not need, nor want, a PAGE_OFFSET option. See kexec/crashdump-elf.c:FUNC() We have: 223 phdr->p_vaddr = phys_to_virt(elf_info, mstart); Now, if we do not have page_offset then we will not have correct p_vaddr, and then vmcore-dmesg/vmcore-dmesg.c:vaddr_to_offset() fails with No program header covering vaddr 0xfffffc0008c312f0found kexec bug? ~Pratyush