OSDN Git Service

fs/binfmt_elf: Fix AT_PHDR for unusual ELF files
authorAkira Kawata <akirakawata1@gmail.com>
Thu, 27 Jan 2022 12:40:16 +0000 (21:40 +0900)
committerKees Cook <keescook@chromium.org>
Wed, 2 Mar 2022 00:16:27 +0000 (16:16 -0800)
commit0da1d5002745cdc721bc018b582a8a9704d56c42
tree7b0584635f927251d3306f8092548eb4b7e30bf2
parentd65bc29be0ae4ca2368df25dc6f6247aefb57f07
fs/binfmt_elf: Fix AT_PHDR for unusual ELF files

BugLink: https://bugzilla.kernel.org/show_bug.cgi?id=197921
As pointed out in the discussion of buglink, we cannot calculate AT_PHDR
as the sum of load_addr and exec->e_phoff.

: The AT_PHDR of ELF auxiliary vectors should point to the memory address
: of program header. But binfmt_elf.c calculates this address as follows:
:
: NEW_AUX_ENT(AT_PHDR, load_addr + exec->e_phoff);
:
: which is wrong since e_phoff is the file offset of program header and
: load_addr is the memory base address from PT_LOAD entry.
:
: The ld.so uses AT_PHDR as the memory address of program header. In normal
: case, since the e_phoff is usually 64 and in the first PT_LOAD region, it
: is the correct program header address.
:
: But if the address of program header isn't equal to the first PT_LOAD
: address + e_phoff (e.g.  Put the program header in other non-consecutive
: PT_LOAD region), ld.so will try to read program header from wrong address
: then crash or use incorrect program header.

This is because exec->e_phoff
is the offset of PHDRs in the file and the address of PHDRs in the
memory may differ from it. This patch fixes the bug by calculating the
address of program headers from PT_LOADs directly.

Signed-off-by: Akira Kawata <akirakawata1@gmail.com>
Reported-by: kernel test robot <lkp@intel.com>
Acked-by: Kees Cook <keescook@chromium.org>
Signed-off-by: Kees Cook <keescook@chromium.org>
Link: https://lore.kernel.org/r/20220127124014.338760-2-akirakawata1@gmail.com
fs/binfmt_elf.c