- 07 Jun, 2023 28 commits
-
-
Josh Poimboeuf authored
Objtool doesn't use DWARF at all, and the DWARF sections' data take up a lot of memory. Skip reading them. Note this only skips the DWARF base sections, not the rela sections. The relas are needed because their symbol references may need to be reindexed if any local symbols get added by elf_create_symbol(). Also note the DWARF data will eventually be read by libelf anyway, when writing the object file. But that's fine, the goal here is to reduce *peak* memory usage, and the previous patch (which freed insn memory) gave some breathing room. So the allocation gets shifted to a later time, resulting in lower peak memory usage. With allyesconfig + CONFIG_DEBUG_INFO: - Before: peak heap memory consumption: 29.93G - After: peak heap memory consumption: 25.47G Link: https://lore.kernel.org/r/52a9698835861dd35f2ec35c49f96d0bb39fb177.1685464332.git.jpoimboe@kernel.orgSigned-off-by:
Josh Poimboeuf <jpoimboe@kernel.org>
-
Josh Poimboeuf authored
Free the decoded instructions as they're no longer needed after this point. This frees up a big chunk of heap, which will come handy when skipping the reading of DWARF section data. Link: https://lore.kernel.org/r/4d4bca1a0f869de020dac80d91f9acbf6df77eab.1685464332.git.jpoimboe@kernel.orgSigned-off-by:
-
Josh Poimboeuf authored
Get the relocation entry info from the underlying rsec->data. With allyesconfig + CONFIG_DEBUG_INFO: - Before: peak heap memory consumption: 35.12G - After: peak heap memory consumption: 29.93G Link: https://lore.kernel.org/r/2be32323de6d8cc73179ee0ff14b71f4e7cefaa0.1685464332.git.jpoimboe@kernel.orgSigned-off-by:
-
Josh Poimboeuf authored
Instead of using hlist for the 'struct elf' hashes, use a custom single-linked list scheme. With allyesconfig + CONFIG_DEBUG_INFO: - Before: peak heap memory consumption: 36.89G - After: peak heap memory consumption: 35.12G Link: https://lore.kernel.org/r/6e8cd305ed22e743c30d6e72cfdc1be20fb94cd4.1685464332.git.jpoimboe@kernel.orgSigned-off-by:
-
Josh Poimboeuf authored
Convert it to a singly-linked list. With allyesconfig + CONFIG_DEBUG_INFO: - Before: peak heap memory consumption: 38.64G - After: peak heap memory consumption: 36.89G Link: https://lore.kernel.org/r/a51f0a6f9bbf2494d5a3a449807307e78a940988.1685464332.git.jpoimboe@kernel.orgSigned-off-by:
-
Josh Poimboeuf authored
Rework the jump table logic slightly so 'jump_table_start' is no longer needed. With allyesconfig + CONFIG_DEBUG_INFO: - Before: peak heap memory consumption: 40.37G - After: peak heap memory consumption: 38.64G Link: https://lore.kernel.org/r/e1602ed8a6171ada3cfac0bd8449892ec82bd188.1685464332.git.jpoimboe@kernel.orgSigned-off-by:
-
Josh Poimboeuf authored
Get the addend from the embedded GElf_Rel[a] struct. With allyesconfig + CONFIG_DEBUG_INFO: - Before: peak heap memory consumption: 42.10G - After: peak heap memory consumption: 40.37G Link: https://lore.kernel.org/r/ad2354f95d9ddd86094e3f7687acfa0750657784.1685464332.git.jpoimboe@kernel.orgSigned-off-by:
-
Josh Poimboeuf authored
Get the type from the embedded GElf_Rel[a] struct. Link: https://lore.kernel.org/r/d1c1f8da31e4f052a2478aea585fcf355cacc53a.1685464332.git.jpoimboe@kernel.orgSigned-off-by:
-
Josh Poimboeuf authored
Get the offset from the embedded GElf_Rel[a] struct. With allyesconfig + CONFIG_DEBUG_INFO: - Before: peak heap memory consumption: 43.83G - After: peak heap memory consumption: 42.10G Link: https://lore.kernel.org/r/2b9ec01178baa346a99522710bf2e82159412e3a.1685464332.git.jpoimboe@kernel.orgSigned-off-by:
-
Josh Poimboeuf authored
Use the array offset to calculate the reloc index. With allyesconfig + CONFIG_DEBUG_INFO: - Before: peak heap memory consumption: 45.56G - After: peak heap memory consumption: 43.83G Link: https://lore.kernel.org/r/7351d2ebad0519027db14a32f6204af84952574a.1685464332.git.jpoimboe@kernel.orgSigned-off-by:
-
Josh Poimboeuf authored
Now that all relocs are allocated in an array, the linked list is no longer needed. With allyesconfig + CONFIG_DEBUG_INFO: - Before: peak heap memory consumption: 49.02G - After: peak heap memory consumption: 45.56G Link: https://lore.kernel.org/r/71e7a2c017dbc46bb497857ec97d67214f832d10.1685464332.git.jpoimboe@kernel.orgSigned-off-by:
-
Josh Poimboeuf authored
Similar to read_relocs(), allocate the reloc structs all together in an array rather than allocating them one at a time. Link: https://lore.kernel.org/r/5332d845c5a2d6c2d052075b381bfba8bcb67ed5.1685464332.git.jpoimboe@kernel.orgSigned-off-by:
-
-
Josh Poimboeuf authored
It's not necessary, objtool's about to exit anyway. Link: https://lore.kernel.org/r/74bdb3058b8f029db8d5b3b5175f2a200804196d.1685464332.git.jpoimboe@kernel.orgSigned-off-by:
-
Josh Poimboeuf authored
Keep the GElf_Rela structs synced with their 'struct reloc' counterparts instead of having to go back and "rebuild" them later. Link: https://lore.kernel.org/r/156d8a3e528a11e5c8577cf552890ed1f2b9567b.1685464332.git.jpoimboe@kernel.orgSigned-off-by:
-
Josh Poimboeuf authored
When creating an annotation section, allocate the reloc section data at the beginning. This simplifies the data model a bit and also saves memory due to the removal of malloc() in elf_rebuild_reloc_section(). With allyesconfig + CONFIG_DEBUG_INFO: - Before: peak heap memory consumption: 53.49G - After: peak heap memory consumption: 49.02G Link: https://lore.kernel.org/r/048e908f3ede9b66c15e44672b6dda992b1dae3e.1685464332.git.jpoimboe@kernel.orgSigned-off-by:
-
Josh Poimboeuf authored
Ensure elf->changed always gets set when sec->changed gets set. Link: https://lore.kernel.org/r/9a810a8d2e28af6ba07325362d0eb4703bb09d3a.1685464332.git.jpoimboe@kernel.orgSigned-off-by:
-
Josh Poimboeuf authored
With CONFIG_DEBUG_INFO, DWARF creates a lot of relocations and reloc_hash is woefully undersized, which can affect performance significantly. Fix that. Link: https://lore.kernel.org/r/38ef60dc8043270bf3b9dfd139ae2a30ca3f75cc.1685464332.git.jpoimboe@kernel.orgSigned-off-by:
-
Josh Poimboeuf authored
The GElf_Rel[a] structs have more similarities than differences. It's safe to hard-code the assumptions about their shared fields as they will never change. Consolidate their handling where possible, getting rid of duplicated code. Also, at least for now we only ever create rela sections, so simplify the relocation creation code to be rela-only. Link: https://lore.kernel.org/r/dcabf6df400ca500ea929f1e4284f5e5ec0b27c8.1685464332.git.jpoimboe@kernel.orgSigned-off-by:
-
Josh Poimboeuf authored
- The term "reloc" is overloaded to mean both "an instance of struct reloc" and "a reloc section". Change the latter to "rsec". - For variable names, use "sec" for regular sections and "rsec" for rela sections to prevent them getting mixed up. - For struct reloc variables, use "reloc" instead of "rel" everywhere for consistency. Link: https://lore.kernel.org/r/8b790e403df46f445c21003e7893b8f53b99a6f3.1685464332.git.jpoimboe@kernel.orgSigned-off-by:
-
Josh Poimboeuf authored
Simplify the elf_create_section() interface a bit by removing the flags argument. Most callers don't care about changing the section header flags. If needed, they can be modified afterwards, just like any other section header field. Link: https://lore.kernel.org/r/515235d9cf62637a14bee37bfa9169ef20065471.1685464332.git.jpoimboe@kernel.orgSigned-off-by:
-
Josh Poimboeuf authored
Reorganize elf.h a bit: - Move the prototypes higher up so they can be used by the inline functions. - Move hash-related code to the bottom. - Remove the unused ELF_HASH_BITS macro. No functional changes. Link: https://lore.kernel.org/r/b1490ed85951868219a6ece177a7cd30a6454d66.1685464332.git.jpoimboe@kernel.orgSigned-off-by:
-
Josh Poimboeuf authored
VMware high-bandwidth hypercalls take the RBP register as input. This breaks basic frame pointer convention, as RBP should never be clobbered. So frame pointer unwinding is broken for the instructions surrounding the hypercalls. Fortunately this doesn't break live patching with CONFIG_FRAME_POINTER, as it only unwinds from blocking tasks, and stack traces from preempted tasks are already marked unreliable anyway. However, for live patching with ORC, this could actually be a theoretical problem if vmw_port_hb_{in,out}() were still compiled with a frame pointer due to having an aligned stack. In practice that hasn't seemed to be an issue since the objtool warnings have only been seen with CONFIG_FRAME_POINTER. Add unwind hint annotations to tell the ORC unwinder to mark stack traces as unreliable. Fixes the following warnings: vmlinux.o: warning: objtool: vmw_port_hb_in+0x1df: return with modified stack frame vmlinux.o: warning: objtool: vmw_port_hb_out+0x1dd: return with modified stack frame Fixes: 89da76fd ("drm/vmwgfx: Add VMWare host messaging capability") Reported-by:kernel test robot <lkp@intel.com> Link: https://lore.kernel.org/oe-kbuild-all/202305160135.97q0Elax-lkp@intel.com/ Link: https://lore.kernel.org/r/4c795f2d87bc0391cf6543bcb224fa540b55ce4b.1685981486.git.jpoimboe@kernel.orgSigned-off-by:
-
Josh Poimboeuf authored
If the code specified UNWIND_HINT_UNDEFINED, skip the "undefined stack state" warning due to a stack operation. Just ignore the stack op and continue to propagate the undefined state. Link: https://lore.kernel.org/r/820c5b433f17c84e8761fb7465a8d319d706b1cf.1685981486.git.jpoimboe@kernel.orgSigned-off-by:
-
Josh Poimboeuf authored
There's no need for both thunk functions to jump to the same shared thunk restore code which lives outside the thunk function boundaries. It disrupts i-cache locality and confuses objtool. Keep it simple by keeping each thunk's restore code self-contained within the function. Fixes a bunch of false positive "missing __noreturn" warnings like: vmlinux.o: warning: objtool: do_arch_prctl_common+0xf4: preempt_schedule_thunk() is missing a __noreturn annotation Fixes: fedb724c ("objtool: Detect missing __noreturn annotations") Reported-by:
-
Josh Poimboeuf authored
Commit 396e0b8e ("x86/orc: Make it callthunk aware") attempted to deal with the fact that function prefix code didn't have ORC coverage. However, it didn't work as advertised. Use of the "null" ORC entry just caused affected unwinds to end early. The root cause has now been fixed with commit 5743654f ("objtool: Generate ORC data for __pfx code"). Revert most of commit 396e0b8e ("x86/orc: Make it callthunk aware"). The is_callthunk() function remains as it's now used by other code. Link: https://lore.kernel.org/r/a05b916ef941da872cbece1ab3593eceabd05a79.1684245404.git.jpoimboe@kernel.orgSigned-off-by:
-
Peter Zijlstra authored
For certain configs objtool will complain like: vmlinux.o: warning: objtool: lkdtm_UNSET_SMEP+0x1c3: relocation to !ENDBR: native_write_cr4+0x41 What happens is that GCC optimizes the loop: insn = (unsigned char *)native_write_cr4; for (i = 0; i < MOV_CR4_DEPTH; i++) to read something like: for (insn = (unsigned char *)native_write_cr4; insn < (unsigned char *)native_write_cr4 + MOV_CR4_DEPTH; insn++) Which then obviously generates the text reference native_write_cr4+041. Since none of this is a fast path, simply confuse GCC enough to inhibit this optimization. Reported-by:Peter Zijlstra (Intel) <peterz@infradead.org> Acked-by:
Kees Cook <keescook@chromium.org> Link: https://lore.kernel.org/r/Y3JdgbXRV0MNZ+9h@hirez.programming.kicks-ass.netSigned-off-by:
-
Lu Hongfei authored
There are several places where warnings variables are not needed, remove them and directly return 0. Signed-off-by:
Lu Hongfei <luhongfei@vivo.com> Link: https://lore.kernel.org/r/20230530075649.21661-1-luhongfei@vivo.comSigned-off-by:
-
- 02 Jun, 2023 1 commit
-
-
Josh Poimboeuf authored
Fixes a bunch of warnings like: drivers/input/tests/input_test.o: warning: objtool: input_test_init+0x1cb: stack state mismatch: cfa1=4+64 cfa2=4+56 lib/kunit/kunit-test.o: warning: objtool: kunit_log_newline_test+0xfb: return with modified stack frame ... Fixes: 26075518 ("kunit: Move kunit_abort() call out of kunit_do_failed_assertion()") Reported-by:
Stephen Rothwell <sfr@canb.auug.org.au> Signed-off-by:
-
- 16 May, 2023 11 commits
-
-
Josh Poimboeuf authored
This makes it a little cleaner and easier to maintain. Suggested-by:
Miroslav Benes <mbenes@suse.cz> Link: https://lore.kernel.org/r/cecacf07a69a244c74474c18b7652627de67a528.1681853186.git.jpoimboe@kernel.orgSigned-off-by:
-
Josh Poimboeuf authored
NORETURN is redundant with __noreturn, just use the latter. Reviewed-by:
-
Josh Poimboeuf authored
lbug_with_loc() no longer exists, and resume_play_dead() is static (objtool only checks globals and weaks). Reviewed-by:
-
Josh Poimboeuf authored
This is a hack, but it works for now. Problem is, exc_double_fault() may or may not return, depending on whether CONFIG_X86_ESPFIX64 is set. But objtool has no visibility to the kernel config. "Fix" it by silencing the exc_double_fault() __noreturn warning. This removes the following warning: vmlinux.o: warning: objtool: xenpv_exc_double_fault+0xd: exc_double_fault() is missing a __noreturn annotation Reviewed-by:
-
Josh Poimboeuf authored
Most "unreachable instruction" warnings these days seem to actually be the result of a missing __noreturn annotation. Add an explicit check for that. Reviewed-by:
-
Josh Poimboeuf authored
Include backtrace in verbose mode. This makes it easy to gather all the information needed for diagnosing objtool warnings. Reviewed-by:
-
Josh Poimboeuf authored
When a warning is associated with a function, add an option to disassemble that function. This makes it easier for reporters to submit the information needed to diagnose objtool warnings. Reviewed-by:
-
Josh Poimboeuf authored
Unreachable instruction warnings are limited to once per object file. That no longer makes sense for vmlinux validation, which might have more unreachable instructions lurking in other places. Change it to once per function. Note this affects some other (much rarer) non-fatal warnings as well. In general I think one-warning-per-function makes sense, as related warnings can accumulate quickly and we want to eventually get back to failing the build with -Werror anyway. Reviewed-by:
-
Josh Poimboeuf authored
Sometimes the one-line ORC unwinder warnings aren't very helpful. Add a new 'unwind_debug' cmdline option which will dump the full stack contents of the current task when an error condition is encountered. Reviewed-by:
-
ndesaulniers@google.com authored
prevent_tail_call_optimization() was added in commit a9a3ed1e ("x86: Fix early boot crash on gcc-10, third try") to work around stack canaries getting inserted into functions that would initialize the stack canary in the first place. Now that we have no_stack_protector function attribute (gcc-11+, clang-7+) and use it on start_kernel(), remove the call to prevent_tail_call_optimization() such that we may one day remove it outright. Reviewed-by:
Nathan Chancellor <nathan@kernel.org> Acked-by:
Nick Desaulniers <ndesaulniers@google.com> Link: https://lore.kernel.org/r/20230412-no_stackp-v2-2-116f9fe4bbe7@google.comSigned-off-by:
-
ndesaulniers@google.com authored
Back during the discussion of commit a9a3ed1e ("x86: Fix early boot crash on gcc-10, third try") we discussed the need for a function attribute to control the omission of stack protectors on a per-function basis; at the time Clang had support for no_stack_protector but GCC did not. This was fixed in gcc-11. Now that the function attribute is available, let's start using it. Callers of boot_init_stack_canary need to use this function attribute unless they're compiled with -fno-stack-protector, otherwise the canary stored in the stack slot of the caller will differ upon the call to boot_init_stack_canary. This will lead to a call to __stack_chk_fail() then panic. Link: https://gcc.gnu.org/bugzilla/show_bug.cgi?id=94722 Link: https://lore.kernel.org/all/20200316130414.GC12561@hirez.programming.kicks-ass.net/Tested-by:
Miguel Ojeda <ojeda@kernel.org> Acked-by:
-
