Re: [PATCH] tools: hv: Fix cross-compilation
From: Saurabh Singh Sengar
Date: Thu Dec 12 2024 - 23:01:14 EST
On Thu, Dec 12, 2024 at 11:06:16PM +0000, Wei Liu wrote:
> On Thu, Dec 12, 2024 at 12:28:34AM -0800, Saurabh Sengar wrote:
> > Use the native ARCH only incase it is not set, this will allow
> > the cross complilation where ARCH is explicitly set. Add few
> > info prints as well to know what arch and toolchain is getting
> > used to build it.
> >
> > Additionally, simplify the check for ARCH so that fcopy daemon
> > is build only for x86_64.
> >
> > Fixes: 82b0945ce2c2 ("tools: hv: Add new fcopy application based on uio driver")
> > Reported-by: Adrian Vladu <avladu@xxxxxxxxxxxxxxxxxxxxxx>
> > Closes: https://lore.kernel.org/linux-hyperv/Z1Y9ZkAt9GPjQsGi@liuwe-devbox-debian-v2/
> > Signed-off-by: Saurabh Sengar <ssengar@xxxxxxxxxxxxxxxxxxx>
> > ---
> > tools/hv/Makefile | 14 +++++++++++---
> > 1 file changed, 11 insertions(+), 3 deletions(-)
> >
> > diff --git a/tools/hv/Makefile b/tools/hv/Makefile
> > index 34ffcec264ab..d29e6be6309b 100644
> > --- a/tools/hv/Makefile
> > +++ b/tools/hv/Makefile
> > @@ -2,7 +2,7 @@
> > # Makefile for Hyper-V tools
> > include ../scripts/Makefile.include
> >
> > -ARCH := $(shell uname -m 2>/dev/null)
> > +ARCH ?= $(shell uname -m 2>/dev/null)
> > sbindir ?= /usr/sbin
> > libexecdir ?= /usr/libexec
> > sharedstatedir ?= /var/lib
> > @@ -20,18 +20,26 @@ override CFLAGS += -O2 -Wall -g -D_GNU_SOURCE -I$(OUTPUT)include
> > override CFLAGS += -Wno-address-of-packed-member
> >
> > ALL_TARGETS := hv_kvp_daemon hv_vss_daemon
> > -ifneq ($(ARCH), aarch64)
> > +ifeq ($(ARCH), x86_64)
>
> Technically speaking, you can also build this for x86 (32bit). Whether
> anybody uses it is another question.
My intention is to allow fcopy daemon build only for the arch it has
been tested. IMO its better than restricting only for arm64/aarch64.
I tried with gcc '-m32' switch which I believe is for 32 bit x86 compilation
I see problems with it on other (kvp daemon) daemons too. I think we never
cared about 32 bit.
saurabh@Saurabh:/work/linux-next/tools/hv$ make ARCH=x86 CFLAGS=-m32
make[1]: Entering directory '/work/linux-next/tools/hv'
CC hv_kvp_daemon.o
hv_kvp_daemon.c:25:10: fatal error: sys/poll.h: No such file or directory
25 | #include <sys/poll.h>
| ^~~~~~~~~~~~
compilation terminated.
make[1]: *** [/work/linux-next/tools/build/Makefile.build:106: hv_kvp_daemon.o] Error 1
make[1]: Leaving directory '/work/linux-next/tools/hv'
make: *** [Makefile:37: hv_kvp_daemon-in.o] Error 2
I don't have any strong opinion here, if you want I can allow x86 compilation
for fcopy daemon as well.
Please let me know what is your preference.
>
> > ALL_TARGETS += hv_fcopy_uio_daemon
> > endif
> > ALL_PROGRAMS := $(patsubst %,$(OUTPUT)%,$(ALL_TARGETS))
> >
> > ALL_SCRIPTS := hv_get_dhcp_info.sh hv_get_dns_info.sh hv_set_ifconfig.sh
> >
> > -all: $(ALL_PROGRAMS)
> > +all: info $(ALL_PROGRAMS)
> >
> > export srctree OUTPUT CC LD CFLAGS
> > include $(srctree)/tools/build/Makefile.include
> >
> > +info:
> > + @echo "---------------------"
> > + @echo "Building for:"
> > + @echo "CC $(CC)"
> > + @echo "LD $(LD)"
> > + @echo "ARCH $(ARCH)"
> > + @echo "---------------------"
>
> I don't think this is needed. Anyone who's building the kernel source
> should know what tool chain they are using and architecture they're
> building for.
I am fine removing it in V2.
- Saurabh