unreachable instruction in mm/workingset.o

[ms178]

Makefile.txt
config.txt
arch_x86_Makefile.txt

With a vanilla clang version 9.0.0 (363704) and Kernel 5.2-rc5 (Suse), I get the following objtool error:

mm/workingset.o: warning: objtool: count_shadow_nodes()+0x12f: unreachable instruction

I'll file others shortly which were seen with the same config and Makefile. As I didn't finish the compilation, I might post more of them if I try a second time. But before spamming the list with more of these, I wanted to hear from you if these are welcome or if there is something wrong with my setup.

CPU is an Intel i5-2410m (Sandy Bridge, x86-64). Please notice, that I have used my custom compiler and linker flags.

[ms178]

By the way, I could reproduce these errors without using Polly (and I missed adding -fopenmp after -polly-parallel in the Makefile uploaded here).

[MaskRay]

With clang trunk (9.0.0 r364163), https://github.com/ClangBuiltLinux/linux/files/3310077/config.txt, and the CFLAGS changes (no polly options):

diff --git i/Makefile w/Makefile
index 9514dac2660a..ebae1e47e547 100644
--- i/Makefile
+++ w/Makefile
@@ -460,0 +461 @@ KBUILD_CFLAGS   := -Wall -Wundef -Werror=strict-prototypes -Wno-trigraphs \
+                  -O3 -march=native -mtune=native \
diff --git i/arch/x86/Makefile w/arch/x86/Makefile
index 56e748a7679f..6d19308267b8 100644
--- i/arch/x86/Makefile
+++ w/arch/x86/Makefile
@@ -128 +128 @@ else
-        cflags-$(CONFIG_GENERIC_CPU) += $(call cc-option,-mtune=generic)
+        cflags-$(CONFIG_GENERIC_CPU) += $(call cc-option,-O3 -march=native -mtune=native)

% make -j"$(nproc)" HOSTCC=clang HOSTLD=ld.lld CC=clang LD=ld.lld O=out modules
...
/bin/sh: line 1: 151049 Segmentation fault      ./tools/objtool/objtool orc generate --module --no-fp --uaccess
drivers/hwmon/pmbus/adm1275.o

[dileks]

@ms178

Can you retest this with Linux v5.3-rc1+ which includes "objtool fixes and enhancements v2" from [2]?

More details see thread "x86 - clang / objtool status" [1].
(It is still worked on... I see only 4 objtool warnings.)

Can you check your open CBL bugs and comment (or close?).

[1] https://lore.kernel.org/all/[email protected]/
[2] https://git.kernel.org/pub/scm/linux/kernel/git/jpoimboe/linux.git/log/?h=objtool-many-fixes-v2

[nickdesaulniers]

So @ms178 I'd verify that -O3 builds cleanly first; there's probably bugs there. Once we can build that cleanly, should we look into -march flags.

[ms178]

Just a sidenote: As I am on vacation for the next couple of weeks, I only have access to a different machine with a Westmere CPU as of right now and cannot use the exact same Kernel config. Once I get access to a newer Kernel RC and Clang build, I'll re-test it.

[ms178]

With -O3, clang version 9.0.0 (365992) and Linux-Next-Next from 24 th July 2019, I get 37 objtool warnings in total:

arch/x86/kvm/svm.o: warning: objtool: is_erratum_383()+0xb1: unreachable instruction
mm/workingset.o: warning: objtool: count_shadow_nodes()+0x12f: unreachable instruction
fs/hfsplus/xattr.o: warning: objtool: hfsplus_create_attributes_file()+0x38f: unreachable instruction
kernel/module.o: warning: objtool: verify_exported_symbols()+0x147: unreachable instruction
arch/x86/pci/fixup.o: warning: objtool: pci_fixup_i450nx()+0xed: unreachable instruction
sound/pci/hda/hda_generic.o: warning: objtool: parse_output_paths()+0x15f: unreachable instruction
sound/soc/codecs/rt5663.o: warning: objtool: rt5663_v2_jack_detect()+0x2ce: unreachable instruction
sound/soc/intel/haswell/sst-haswell-pcm.o: warning: objtool: hsw_pcm_prepare()+0x1c9: unreachable instruction
net/bridge/netfilter/ebtables.o: warning: objtool: compat_calc_entry()+0x1bb: unreachable instruction
drivers/block/skd_main.o: warning: objtool: skd_pci_probe()+0x393: unreachable instruction
drivers/gpu/drm/amd/amdgpu/atom.o: warning: objtool: atom_op_move() falls through to next function atom_op_and()
drivers/hwmon/pmbus/adm1275.o: warning: objtool: adm1275_probe()+0x744: unreachable instruction
drivers/hwmon/ad7414.o: warning: objtool: ad7414_update_device()+0xf2: unreachable instruction
drivers/hwmon/adm9240.o: warning: objtool: adm9240_update_device()+0x236: unreachable instruction
drivers/hwmon/w83l785ts.o: warning: objtool: w83l785ts_read_value()+0xa1: unreachable instruction
drivers/gpu/drm/amd/amdgpu/../display/dc/calcs/dce_calcs.o: warning: objtool: print_bw_calcs_data()+0x284b: unreachable instruction
drivers/gpu/drm/ttm/ttm_bo.o: warning: objtool: ttm_bo_device_release()+0x207: unreachable instruction
drivers/input/mouse/elan_i2c_core.o: warning: objtool: elan_write_fw_block()+0x8d: unreachable instruction
drivers/input/tablet/gtco.o: warning: objtool: gtco_probe()+0x3d0: unreachable instruction
drivers/misc/mei/bus-fixup.o: warning: objtool: mei_fwver()+0x1ce: unreachable instruction
drivers/misc/mei/hw-txe.o: warning: objtool: mei_txe_fw_status()+0xce: unreachable instruction
drivers/input/touchscreen/elants_i2c.o: warning: objtool: elants_i2c_query_test_version()+0x16e: unreachable instruction
drivers/nfc/pn544/i2c.o: warning: objtool: pn544_hci_i2c_platform_init()+0x140: unreachable instruction
drivers/input/touchscreen/usbtouchscreen.o: warning: objtool: mtouch_init()+0x256: unreachable instruction
drivers/mtd/ssfdc.o: warning: objtool: get_logical_address()+0x161: unreachable instruction
drivers/ntb/hw/mscc/ntb_hw_switchtec.o: warning: objtool: crosslink_enum_partition()+0x107: unreachable instruction
net/netfilter/nf_conntrack_proto_gre.o: warning: objtool: nf_ct_gre_keymap_destroy()+0xe7: unreachable instruction
drivers/platform/x86/intel_turbo_max_3.o: warning: objtool: get_oc_core_priority()+0xf1: unreachable instruction
drivers/staging/rts5208/rtsx_card.o: warning: objtool: card_rw()+0x1cc: unreachable instruction
drivers/staging/rts5208/sd.o: warning: objtool: sd_tuning_rx()+0x167: unreachable instruction
drivers/thermal/intel/intel_soc_dts_iosf.o: warning: objtool: intel_soc_dts_iosf_interrupt_handler()+0xe6: unreachable instruction
drivers/usb/atm/usbatm.o: warning: objtool: usbatm_usb_probe()+0x41c: unreachable instruction
drivers/usb/storage/sierra_ms.o: warning: objtool: sierra_ms_init()+0x136: unreachable instruction
drivers/usb/serial/f81534.o: warning: objtool: f81534_calc_num_ports()+0x49e: unreachable instruction
drivers/usb/serial/keyspan.o: warning: objtool: keyspan_open()+0x153: unreachable instruction
drivers/scsi/qla2xxx/qla_target.o: warning: objtool: qlt_lport_dump()+0x116: unreachable instruction
drivers/scsi/ufs/ufshcd.o: warning: objtool: ufshcd_dme_set_attr()+0xfc: unreachable instruction

I've added the config, the Makefile and the complete build log.
config_Westmere.txt
kernellog_Westmere.txt
Makefile_Westmere.txt
objtool_warnings_Westmere.txt

System: Intel Xeon X5675, 24 GB RAM, Asus P6T WS Pro, AMD RX 580

[ms178]

@nickdesaulniers @dileks Here are the results with -O3, linux-next-next-20190726 and clang version 10.0.0 (366783). There is some good progress to report, we came down from 37 to 2 warnings with the attached config on my Xeon X5675 system (mentioned above).

drivers/gpu/drm/amd/amdgpu/atom.o: warning: objtool: atom_op_move() falls through to next function atom_op_and()
drivers/hwmon/pmbus/adm1275.o: warning: objtool: adm1275_probe()+0x744: unreachable instruction

This build also doesn't succeed with several errors in drivers/gpu/drm/amd/amdgpu/amdgpu.ko at the end but this is a known issue mentioned on the LKML already.

buildlog_Westmere_2.txt

config_Westmere_2.txt

Makefile_Westmere_2.txt

arch_x86_Makefile.txt

[nickdesaulniers]

@ms178, thanks for the report :

1. is the issue in mm/workingset.o still reproducible? If no, please close this issue and file one for each individual issue you observe w/ steps to reproduce, if there's not already a tracking issue.
2. Are the above observed at -O2 or only at -O3?

[ms178]

@nickdesaulniers

1. The mm/workingset.o issue is also not reproducible anymore (according to the buildlog), I'll close this issue then and will open two issues for the two observed. I noticed that you tag fixed issues to categorize if it was fixed in the Kernel or LLVM/Clang (etc.) but as I don't know which respective fix was responsible for fixing these, could you add the tags if necessary?

2. I just tested -O3 with that specific config but could also check with -O2 tomorrow and could add more fancy flags after that.