| Message ID | 20240501122918.3831734-1-usama.anjum@collabora.com |
|---|---|
| Headers | show |
| Series | selftests: x86: build suite with clang | expand |
On Wed, May 1, 2024 at 5:29 AM Muhammad Usama Anjum <usama.anjum@collabora.com> wrote: > > This series fixes build errors found by clang to allow the x86 suite to > get built with the clang. > > Unfortunately, there is one bug [1] in the clang becuase of which > extended asm isn't handled correctly by it and build fails for > sysret_rip.c. Hence even after this series the build of this test would > fail with clang. Should we disable this test for now when clang is used > until the bug is fixed in clang? Not sure. Any opinions? > > [1] https://github.com/llvm/llvm-project/issues/53728 I've closed https://github.com/llvm/llvm-project/issues/53728 (constant expression folding for relocatable file output (MCObjectStreamer)). I've sent a patch to address https://github.com/llvm/llvm-project/issues/62520 (constant expression folding for relocatable file output when inline assembly is used). (I am subscribed to llvm@lists.linux.dev but rarely read it. I noticed this message accidentally :) ) > Muhammad Usama Anjum (8): > selftests: x86: Remove dependence of headers file > selftests: x86: check_initial_reg_state: remove -no-pie while using > -static > selftests: x86: test_vsyscall: remove unused function > selftests: x86: fsgsbase_restore: fix asm directive from =rm to =r > selftests: x86: syscall_arg_fault_32: remove unused variable > selftests: x86: test_FISTTP: use fisttps instead of ambigous fisttp > selftests: x86: fsgsbase: Remove unused function and variable > selftests: x86: amx: Remove unused functions > > tools/testing/selftests/x86/Makefile | 9 +++++---- > tools/testing/selftests/x86/amx.c | 16 ---------------- > tools/testing/selftests/x86/fsgsbase.c | 6 ------ > tools/testing/selftests/x86/fsgsbase_restore.c | 2 +- > tools/testing/selftests/x86/syscall_arg_fault.c | 1 - > tools/testing/selftests/x86/test_FISTTP.c | 8 ++++---- > tools/testing/selftests/x86/test_vsyscall.c | 5 ----- > 7 files changed, 10 insertions(+), 37 deletions(-) > > -- > 2.39.2 > >
On Mon, May 6, 2024 at 4:26 PM Fangrui Song <maskray@google.com> wrote: > > On Wed, May 1, 2024 at 5:29 AM Muhammad Usama Anjum > <usama.anjum@collabora.com> wrote: > > > > This series fixes build errors found by clang to allow the x86 suite to > > get built with the clang. > > > > Unfortunately, there is one bug [1] in the clang becuase of which > > extended asm isn't handled correctly by it and build fails for > > sysret_rip.c. Hence even after this series the build of this test would > > fail with clang. Should we disable this test for now when clang is used > > until the bug is fixed in clang? Not sure. Any opinions? > > > > [1] https://github.com/llvm/llvm-project/issues/53728 > > I've closed https://github.com/llvm/llvm-project/issues/53728 > (constant expression folding for relocatable file output > (MCObjectStreamer)). > > I've sent a patch to address https://github.com/llvm/llvm-project/issues/62520 > (constant expression folding for relocatable file output when inline > assembly is used). > > (I am subscribed to llvm@lists.linux.dev but rarely read it. I noticed > this message accidentally :) ) On the LLVM/Clang side, I've landed https://github.com/llvm/llvm-project/pull/91082 to make the following .if directive work for clang -c. clang -S still doesn't work (https://discourse.llvm.org/t/rfc-clang-assembly-object-equivalence-for-files-with-inline-assembly/78841/4?u=maskray) but people can probably live with that ``` % cat b.cc asm(R"( .pushsection .text,"ax" .globl _start; _start: ret .if . -_start == 1 ret .endif .popsection )"); % clang -c b.cc # succeeded with this patch % clang -S b.cc # still failed <inline asm>:4:5: error: expected absolute expression 4 | .if . -_start == 1 | ^ 1 error generated. ``` > > Muhammad Usama Anjum (8): > > selftests: x86: Remove dependence of headers file > > selftests: x86: check_initial_reg_state: remove -no-pie while using > > -static > > selftests: x86: test_vsyscall: remove unused function > > selftests: x86: fsgsbase_restore: fix asm directive from =rm to =r > > selftests: x86: syscall_arg_fault_32: remove unused variable > > selftests: x86: test_FISTTP: use fisttps instead of ambigous fisttp > > selftests: x86: fsgsbase: Remove unused function and variable > > selftests: x86: amx: Remove unused functions > > > > tools/testing/selftests/x86/Makefile | 9 +++++---- > > tools/testing/selftests/x86/amx.c | 16 ---------------- > > tools/testing/selftests/x86/fsgsbase.c | 6 ------ > > tools/testing/selftests/x86/fsgsbase_restore.c | 2 +- > > tools/testing/selftests/x86/syscall_arg_fault.c | 1 - > > tools/testing/selftests/x86/test_FISTTP.c | 8 ++++---- > > tools/testing/selftests/x86/test_vsyscall.c | 5 ----- > > 7 files changed, 10 insertions(+), 37 deletions(-) > > > > -- > > 2.39.2 > > > >
On 5/1/24 6:29 AM, Muhammad Usama Anjum wrote: > This series fixes build errors found by clang to allow the x86 suite to > get built with the clang. > > Unfortunately, there is one bug [1] in the clang becuase of which > extended asm isn't handled correctly by it and build fails for > sysret_rip.c. Hence even after this series the build of this test would > fail with clang. Should we disable this test for now when clang is used > until the bug is fixed in clang? Not sure. Any opinions? Its seems like the bug has been fixed in clang. I'll verify it in next release and draft a patch separately to fix that error. I think this series is good to accept then. > > [1] https://github.com/llvm/llvm-project/issues/53728 > > Muhammad Usama Anjum (8): > selftests: x86: Remove dependence of headers file > selftests: x86: check_initial_reg_state: remove -no-pie while using > -static > selftests: x86: test_vsyscall: remove unused function > selftests: x86: fsgsbase_restore: fix asm directive from =rm to =r > selftests: x86: syscall_arg_fault_32: remove unused variable > selftests: x86: test_FISTTP: use fisttps instead of ambigous fisttp > selftests: x86: fsgsbase: Remove unused function and variable > selftests: x86: amx: Remove unused functions > > tools/testing/selftests/x86/Makefile | 9 +++++---- > tools/testing/selftests/x86/amx.c | 16 ---------------- > tools/testing/selftests/x86/fsgsbase.c | 6 ------ > tools/testing/selftests/x86/fsgsbase_restore.c | 2 +- > tools/testing/selftests/x86/syscall_arg_fault.c | 1 - > tools/testing/selftests/x86/test_FISTTP.c | 8 ++++---- > tools/testing/selftests/x86/test_vsyscall.c | 5 ----- > 7 files changed, 10 insertions(+), 37 deletions(-) >
On 5/27/24 23:04, Muhammad Usama Anjum wrote: > Soft reminder > > On 5/1/24 5:29 PM, Muhammad Usama Anjum wrote: >> This series fixes build errors found by clang to allow the x86 suite to >> get built with the clang. >> >> Unfortunately, there is one bug [1] in the clang becuase of which >> extended asm isn't handled correctly by it and build fails for >> sysret_rip.c. Hence even after this series the build of this test would >> fail with clang. Should we disable this test for now when clang is used >> until the bug is fixed in clang? Not sure. Any opinions? >> >> [1] https://github.com/llvm/llvm-project/issues/53728 >> >> Muhammad Usama Anjum (8): >> selftests: x86: Remove dependence of headers file >> selftests: x86: check_initial_reg_state: remove -no-pie while using >> -static >> selftests: x86: test_vsyscall: remove unused function >> selftests: x86: fsgsbase_restore: fix asm directive from =rm to =r >> selftests: x86: syscall_arg_fault_32: remove unused variable >> selftests: x86: test_FISTTP: use fisttps instead of ambigous fisttp >> selftests: x86: fsgsbase: Remove unused function and variable >> selftests: x86: amx: Remove unused functions >> >> tools/testing/selftests/x86/Makefile | 9 +++++---- >> tools/testing/selftests/x86/amx.c | 16 ---------------- >> tools/testing/selftests/x86/fsgsbase.c | 6 ------ >> tools/testing/selftests/x86/fsgsbase_restore.c | 2 +- >> tools/testing/selftests/x86/syscall_arg_fault.c | 1 - >> tools/testing/selftests/x86/test_FISTTP.c | 8 ++++---- >> tools/testing/selftests/x86/test_vsyscall.c | 5 ----- >> 7 files changed, 10 insertions(+), 37 deletions(-) >> > These patches usually go through x86 repo. I need ack from x86 maintainers to take these. I don't see x86 list cc'ed. Please make sure to include everybody on these threads to get quicker response. thanks, -- Shuah
On 6/7/24 13:50, Shuah Khan wrote: > These patches usually go through x86 repo. > > I need ack from x86 maintainers to take these. I don't see x86 list > cc'ed. > > Please make sure to include everybody on these threads to get quicker > response. John Hubbard (cc'd) had a partially overlapping set with these. He and Muhammad were going to work together to get a set they could both get behind. I _think_ John picked up one of Muhammad's, but we haven't heard back from Muhammad last I checked.
On 6/7/24 2:17 PM, Dave Hansen wrote: > On 6/7/24 13:50, Shuah Khan wrote: >> These patches usually go through x86 repo. >> >> I need ack from x86 maintainers to take these. I don't see x86 list >> cc'ed. >> >> Please make sure to include everybody on these threads to get quicker >> response. > > John Hubbard (cc'd) had a partially overlapping set with these. He and > Muhammad were going to work together to get a set they could both get > behind. I _think_ John picked up one of Muhammad's, but we haven't > heard back from Muhammad last I checked. Yes I did that. My latest series has 1 fix from Muhammad and 6 more from me. Muhammad hasn't formally ack'd nor responded to the latest series [1] yet, although he was OK doing it this way [2]. [1] https://lore.kernel.org/20240531193838.108454-1-jhubbard@nvidia.com [2] https://lore.kernel.org/046d2d1d-3583-426b-b745-59f3696fb418@collabora.com thanks,
This series fixes build errors found by clang to allow the x86 suite to get built with the clang. Unfortunately, there is one bug [1] in the clang becuase of which extended asm isn't handled correctly by it and build fails for sysret_rip.c. Hence even after this series the build of this test would fail with clang. Should we disable this test for now when clang is used until the bug is fixed in clang? Not sure. Any opinions? [1] https://github.com/llvm/llvm-project/issues/53728 Muhammad Usama Anjum (8): selftests: x86: Remove dependence of headers file selftests: x86: check_initial_reg_state: remove -no-pie while using -static selftests: x86: test_vsyscall: remove unused function selftests: x86: fsgsbase_restore: fix asm directive from =rm to =r selftests: x86: syscall_arg_fault_32: remove unused variable selftests: x86: test_FISTTP: use fisttps instead of ambigous fisttp selftests: x86: fsgsbase: Remove unused function and variable selftests: x86: amx: Remove unused functions tools/testing/selftests/x86/Makefile | 9 +++++---- tools/testing/selftests/x86/amx.c | 16 ---------------- tools/testing/selftests/x86/fsgsbase.c | 6 ------ tools/testing/selftests/x86/fsgsbase_restore.c | 2 +- tools/testing/selftests/x86/syscall_arg_fault.c | 1 - tools/testing/selftests/x86/test_FISTTP.c | 8 ++++---- tools/testing/selftests/x86/test_vsyscall.c | 5 ----- 7 files changed, 10 insertions(+), 37 deletions(-)