linux-kernel - Re: [PATCH 1/1] tools/build: Let link command read option from file

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [<thread-prev] [day] [month] [year] [list]

Message-ID: <33e72aba-d8ca-4a9c-b761-f83a41f8fab0@windriver.com>
Date: Mon, 11 Aug 2025 09:37:07 +0800
From: Changqing Li <changqing.li@...driver.com>
To: Ian Rogers <irogers@...gle.com>
Cc: namhyung@...nel.org, charlie@...osinc.com, james.clark@...aro.org,
        linux-kernel@...r.kernel.org
Subject: Re: [PATCH 1/1] tools/build: Let link command read option from file


On 7/22/25 23:43, Ian Rogers wrote:
> CAUTION: This email comes from a non Wind River email account!
> Do not click links or open attachments unless you recognize the sender and know the content is safe.
>
> On Mon, Jul 21, 2025 at 11:13 PM <changqing.li@...driver.com> wrote:
>> From: Changqing Li <changqing.li@...driver.com>
>>
>> ld_multi will link multiple objects, when there are many objects, and
>> O=[absolute_path] is set, and the absolute_path is relatively long. It
>> is possile that this line "$(call if_changed,$(host)ld_multi)" will
>> report error:
>> "make[4]: /bin/sh: Argument list too long"
>>
>> So make the ld command read option from file to fix above error. In
>> order to convenient debug, write the file content in dot-target.cmd
>> as comments.
>>
>> Signed-off-by: Changqing Li <changqing.li@...driver.com>
> Hi Changqing,
>
> I believe your change makes sense. I notice that the regular kernel
> build has had to work around this problem too:
> https://web.git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/tree/scripts/Makefile.build#n290
> https://web.git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/tree/scripts/Makefile.build#n463
> but in those workarounds the need for an extra .in file isn't
> necessary. Would such a change be possible here and avoid the need for
> cleaning up an extra file?
>
> Thanks,
> Ian

Hi, Ian

Thanks for your help. I have resend a patch as what you suggested, could 
you help to review it?

subject:  "tools/build: make in-target rule robust against too long 
argument error"


Regards

changqing
>> ---
>>   tools/build/Makefile.build | 9 +++++++--
>>   1 file changed, 7 insertions(+), 2 deletions(-)
>>
>> diff --git a/tools/build/Makefile.build b/tools/build/Makefile.build
>> index 3584ff308607..e57ce8c34685 100644
>> --- a/tools/build/Makefile.build
>> +++ b/tools/build/Makefile.build
>> @@ -70,11 +70,13 @@ quiet_cmd_gen = GEN     $@
>>   # If there's nothing to link, create empty $@ object.
>>   quiet_cmd_ld_multi = LD      $@
>>         cmd_ld_multi = $(if $(strip $(obj-y)),\
>> -                     $(LD) -r -o $@  $(filter $(obj-y),$^),rm -f $@; $(AR) rcs $@)
>> +                     $(LD) -r -o $@ @$@.in,rm -f $@; $(AR) rcs $@)
>>
>>   quiet_cmd_host_ld_multi = HOSTLD  $@
>>         cmd_host_ld_multi = $(if $(strip $(obj-y)),\
>> -                          $(HOSTLD) -r -o $@  $(filter $(obj-y),$^),rm -f $@; $(HOSTAR) rcs $@)
>> +                          $(HOSTLD) -r -o $@ @$@.in,rm -f $@; $(HOSTAR) rcs $@)
>> +
>> +output_ld_multi_dotin = $(if $(quiet),,@printf "# %s:\n# " $@.in >> $(dot-target).cmd;cat $@.in >> $(dot-target).cmd)
>>
>>   ifneq ($(filter $(obj),$(hostprogs)),)
>>     host = host_
>> @@ -145,7 +147,10 @@ $(sort $(subdir-obj-y)): $(subdir-y) ;
>>
>>   $(in-target): $(obj-y) $(test-y) FORCE
>>          $(call rule_mkdir)
>> +       $(file >$@.in,$(filter $(obj-y),$^))
>>          $(call if_changed,$(host)ld_multi)
>> +       $(if $(strip $(any-prereq) $(arg-check)), $(output_ld_multi_dotin))
>> +       @rm $@.in
>>
>>   __build: $(in-target)
>>          @:
>> --
>> 2.34.1
>>