| Message ID | 20250225-nolibc-mips-n32-v2-0-664b47d87fa0@weissschuh.net (mailing list archive) |
|---|---|
| Headers | show |
| Series | tools/nolibc: MIPS: entrypoint cleanups and N32/N64 ABIs | expand |
Hi Thomas! On Tue, Feb 25, 2025 at 06:02:34PM +0100, Thomas Weißschuh wrote: > Introduce support for the N32 and N64 ABIs. As preparation, the > entrypoint is first simplified significantly. Thanks to Maciej for all > the valuable information. > > Signed-off-by: Thomas Weißschuh <linux@weissschuh.net> > --- > Changes in v2: > - Clean up entrypoint first > - Annotate #endifs > - Link to v1: https://lore.kernel.org/r/20250212-nolibc-mips-n32-v1-1-6892e58d1321@weissschuh.net OK I tested this series on my glinet (MIPS 24Kc, XARCH=mips32be) and it worked fine, confirming that the stack alignments were not needed and that the cleanup is quite welcome! Tested-by: Willy Tarreau <w@1wt.eu> Willy
On Sat, 1 Mar 2025, Willy Tarreau wrote: > > Introduce support for the N32 and N64 ABIs. As preparation, the > > entrypoint is first simplified significantly. Thanks to Maciej for all > > the valuable information. > > > > Signed-off-by: Thomas Weißschuh <linux@weissschuh.net> > > --- > > Changes in v2: > > - Clean up entrypoint first > > - Annotate #endifs > > - Link to v1: https://lore.kernel.org/r/20250212-nolibc-mips-n32-v1-1-6892e58d1321@weissschuh.net > > OK I tested this series on my glinet (MIPS 24Kc, XARCH=mips32be) and > it worked fine, confirming that the stack alignments were not needed > and that the cleanup is quite welcome! I do hope it can wait two weeks until I'm back from my holiday. I mean to double-check the code visually and verify it with my R3000 and R4000 hardware (the latter for n64/n32 too), both of which are less forgiving when it comes to instruction scheduling (I can check with a 74Kf too). Maciej
On 2025-03-01 15:47:52+0000, Maciej W. Rozycki wrote: > On Sat, 1 Mar 2025, Willy Tarreau wrote: > > > > Introduce support for the N32 and N64 ABIs. As preparation, the > > > entrypoint is first simplified significantly. Thanks to Maciej for all > > > the valuable information. > > > > > > Signed-off-by: Thomas Weißschuh <linux@weissschuh.net> > > > --- > > > Changes in v2: > > > - Clean up entrypoint first > > > - Annotate #endifs > > > - Link to v1: https://lore.kernel.org/r/20250212-nolibc-mips-n32-v1-1-6892e58d1321@weissschuh.net > > > > OK I tested this series on my glinet (MIPS 24Kc, XARCH=mips32be) and > > it worked fine, confirming that the stack alignments were not needed > > and that the cleanup is quite welcome! > > I do hope it can wait two weeks until I'm back from my holiday. Absolutely. > I mean > to double-check the code visually and verify it with my R3000 and R4000 > hardware (the latter for n64/n32 too), both of which are less forgiving > when it comes to instruction scheduling (I can check with a 74Kf too). Your testing and feedback is very valuable, I'm happy to wait for it. Thanks, Thomas
On Sat, Mar 01, 2025 at 03:47:52PM +0000, Maciej W. Rozycki wrote: > On Sat, 1 Mar 2025, Willy Tarreau wrote: > > > > Introduce support for the N32 and N64 ABIs. As preparation, the > > > entrypoint is first simplified significantly. Thanks to Maciej for all > > > the valuable information. > > > > > > Signed-off-by: Thomas Weißschuh <linux@weissschuh.net> > > > --- > > > Changes in v2: > > > - Clean up entrypoint first > > > - Annotate #endifs > > > - Link to v1: https://lore.kernel.org/r/20250212-nolibc-mips-n32-v1-1-6892e58d1321@weissschuh.net > > > > OK I tested this series on my glinet (MIPS 24Kc, XARCH=mips32be) and > > it worked fine, confirming that the stack alignments were not needed > > and that the cleanup is quite welcome! > > I do hope it can wait two weeks until I'm back from my holiday. I mean > to double-check the code visually and verify it with my R3000 and R4000 > hardware (the latter for n64/n32 too), both of which are less forgiving > when it comes to instruction scheduling (I can check with a 74Kf too). Oh that would be great, thank you Maciej! Willy
On 2025-02-25 18:02:34 [+0100], Thomas Weißschuh wrote: > Introduce support for the N32 and N64 ABIs. As preparation, the > entrypoint is first simplified significantly. Thanks to Maciej for all > the valuable information. > > Signed-off-by: Thomas Weißschuh <linux@weissschuh.net> Tested-by: Sebastian Andrzej Siewior <sebastian@breakpoint.cc> [ICT Loongson-3] mips32le works as-is. For mips64le I had to s/-march=mips64r6/-march=mips64r2 to match the ABI. Which makes me wonder: Why do do we need to pass -march here and can't rely on toolchain defaults? Sebastian
On 2025-03-26 21:54:34+0100, Sebastian Andrzej Siewior wrote: > On 2025-02-25 18:02:34 [+0100], Thomas Weißschuh wrote: > > Introduce support for the N32 and N64 ABIs. As preparation, the > > entrypoint is first simplified significantly. Thanks to Maciej for all > > the valuable information. > > > > Signed-off-by: Thomas Weißschuh <linux@weissschuh.net> > > Tested-by: Sebastian Andrzej Siewior <sebastian@breakpoint.cc> [ICT Loongson-3] Thanks! > mips32le works as-is. > For mips64le I had to s/-march=mips64r6/-march=mips64r2 to match the > ABI. Which makes me wonder: Why do do we need to pass -march here and > can't rely on toolchain defaults? The goal here is to have an as-wide-as-possible test matrix for nolibc-test, which will mostly be running on QEMU anyways. Also we need to run the correct QEMU user variant; by fixing the architecture this is easy to do. Thomas
On 2025-03-26 22:51:54 [+0100], Thomas Weißschuh wrote: > > mips32le works as-is. > > For mips64le I had to s/-march=mips64r6/-march=mips64r2 to match the > > ABI. Which makes me wonder: Why do do we need to pass -march here and > > can't rely on toolchain defaults? > > The goal here is to have an as-wide-as-possible test matrix for > nolibc-test, which will mostly be running on QEMU anyways. > Also we need to run the correct QEMU user variant; by fixing the > architecture this is easy to do. I would prefer to make distro users as in real hardware first class citizen and not QEMU users. If you run qemu you can specify the ABI anyway. > Thomas Sebastian
On 2025-03-26 23:04:30+0100, Sebastian Andrzej Siewior wrote: > On 2025-03-26 22:51:54 [+0100], Thomas Weißschuh wrote: > > > mips32le works as-is. > > > For mips64le I had to s/-march=mips64r6/-march=mips64r2 to match the > > > ABI. Which makes me wonder: Why do do we need to pass -march here and > > > can't rely on toolchain defaults? > > > > The goal here is to have an as-wide-as-possible test matrix for > > nolibc-test, which will mostly be running on QEMU anyways. > > Also we need to run the correct QEMU user variant; by fixing the > > architecture this is easy to do. > > I would prefer to make distro users as in real hardware first class > citizen and not QEMU users. If you run qemu you can specify the ABI > anyway. Does the following work for you when running kust "make nolibc-test"? diff --git a/tools/testing/selftests/nolibc/Makefile b/tools/testing/selftests/nolibc/Makefile index 58bcbbd029bc..27d5ceb20858 100644 --- a/tools/testing/selftests/nolibc/Makefile +++ b/tools/testing/selftests/nolibc/Makefile @@ -174,10 +174,13 @@ CFLAGS_s390x = -m64 CFLAGS_s390 = -m31 CFLAGS_mips32le = -EL -mabi=32 -fPIC CFLAGS_mips32be = -EB -mabi=32 +ifeq ($(origin XARCH),command line) +CFLAGS_ARCH = $(CFLAGS_$(XARCH)) +endif CFLAGS_STACKPROTECTOR ?= $(call cc-option,-mstack-protector-guard=global $(call cc-option,-fstack-protector-all)) CFLAGS ?= -Os -fno-ident -fno-asynchronous-unwind-tables -std=c89 -W -Wall -Wextra \ $(call cc-option,-fno-stack-protector) $(call cc-option,-Wmissing-prototypes) \ - $(CFLAGS_$(XARCH)) $(CFLAGS_STACKPROTECTOR) $(CFLAGS_EXTRA) + $(CFLAGS_ARCH) $(CFLAGS_STACKPROTECTOR) $(CFLAGS_EXTRA) LDFLAGS := LIBGCC := -lgcc @@ -232,7 +235,7 @@ all: run sysroot: sysroot/$(ARCH)/include -sysroot/$(ARCH)/include: | defconfig +sysroot/$(ARCH)/include: $(Q)rm -rf sysroot/$(ARCH) sysroot/sysroot $(QUIET_MKDIR)mkdir -p sysroot $(Q)$(MAKE) -C $(srctree) outputmakefile
On 2025-03-29 10:51:47 [+0100], Thomas Weißschuh wrote: > > Does the following work for you when running kust "make nolibc-test"? I had to manually apply this on top of b4/nolibc-mips-n32. The resulting make produced the native 64bit binary. Sebastian
Introduce support for the N32 and N64 ABIs. As preparation, the entrypoint is first simplified significantly. Thanks to Maciej for all the valuable information. Signed-off-by: Thomas Weißschuh <linux@weissschuh.net> --- Changes in v2: - Clean up entrypoint first - Annotate #endifs - Link to v1: https://lore.kernel.org/r/20250212-nolibc-mips-n32-v1-1-6892e58d1321@weissschuh.net --- Thomas Weißschuh (4): tools/nolibc: MIPS: drop $gp setup tools/nolibc: MIPS: drop manual stack pointer alignment tools/nolibc: MIPS: drop noreorder option tools/nolibc: MIPS: add support for N64 and N32 ABIs tools/include/nolibc/arch-mips.h | 117 +++++++++++++++++++++------- tools/testing/selftests/nolibc/Makefile | 28 ++++++- tools/testing/selftests/nolibc/run-tests.sh | 2 +- 3 files changed, 118 insertions(+), 29 deletions(-) --- base-commit: 9c812b01f13d37410ea103e00bc47e5e0f6d2bad change-id: 20231105-nolibc-mips-n32-234901bd910d Best regards,