From mboxrd@z Thu Jan 1 00:00:00 1970 Received: from eggs.gnu.org ([2001:4830:134:3::10]:41329) by lists.gnu.org with esmtp (Exim 4.71) (envelope-from ) id 1degcx-0005FA-B9 for qemu-devel@nongnu.org; Mon, 07 Aug 2017 07:55:24 -0400 Received: from Debian-exim by eggs.gnu.org with spam-scanned (Exim 4.71) (envelope-from ) id 1degcs-00010k-DB for qemu-devel@nongnu.org; Mon, 07 Aug 2017 07:55:19 -0400 Received: from mx1.redhat.com ([209.132.183.28]:55238) by eggs.gnu.org with esmtps (TLS1.0:DHE_RSA_AES_256_CBC_SHA1:32) (Exim 4.71) (envelope-from ) id 1degcs-0000zP-3T for qemu-devel@nongnu.org; Mon, 07 Aug 2017 07:55:14 -0400 References: <150210580404.1343.7325713896658799315.stgit@bahia.lan> From: Paolo Bonzini Message-ID: <77c63a1b-92f9-8fc5-dbf5-855c28adc03d@redhat.com> Date: Mon, 7 Aug 2017 13:55:09 +0200 MIME-Version: 1.0 In-Reply-To: <150210580404.1343.7325713896658799315.stgit@bahia.lan> Content-Type: text/plain; charset=utf-8 Content-Language: en-US Content-Transfer-Encoding: quoted-printable Subject: Re: [Qemu-devel] [PATCH v2] kvm: workaround build break on gcc-7.1.1 / fedora26 List-Id: List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , To: Greg Kurz , qemu-devel@nongnu.org Cc: Cornelia Huck , Eric Blake , =?UTF-8?Q?Philippe_Mathieu-Daud=c3=a9?= On 07/08/2017 13:36, Greg Kurz wrote: > Building QEMU on fedora26 with the latest gcc package fails: >=20 > CC ppc64-softmmu/target/ppc/kvm.o > In file included from include/sysemu/hw_accel.h:16:0, > from target/ppc/kvm.c:31: > target/ppc/kvm.c: In function =E2=80=98kvmppc_booke_watchdog_enable=E2=80= =99: > include/sysemu/kvm.h:449:35: error: =E2=80=98args_tmp[i]=E2=80=99 may b= e used uninitialized > in this function [-Werror=3Dmaybe-uninitialized] > cap.args[i] =3D args_tmp[i]; = \ > ^ > target/ppc/kvm.c: In function =E2=80=98kvmppc_set_papr=E2=80=99: > include/sysemu/kvm.h:449:35: error: =E2=80=98args_tmp[i]=E2=80=99 may b= e used uninitialized > in this function [-Werror=3Dmaybe-uninitialized] > cc1: all warnings being treated as errors >=20 > $ rpm -q gcc > gcc-7.1.1-3.fc26.ppc64le >=20 > The compiler should obviously optimize this code away when no extra > agument is passed to kvm_vm_enable_cap() and kvm_vcpu_enable_cap(), > but it doesn't. This bug should be fixed one day in gcc, but we can > also change our code pattern so that we don't hit the issue anymore. > We workaround this, by using memcpy() instead of open-coding the copy. Nice way to do it, thanks. I'll queue it for 2.10. Paolo > Signed-off-by: Greg Kurz > --- > v2: - use memcpy() > --- > include/sysemu/kvm.h | 14 ++++---------- > 1 file changed, 4 insertions(+), 10 deletions(-) >=20 > diff --git a/include/sysemu/kvm.h b/include/sysemu/kvm.h > index 91fc07ee9afe..3a458f50e9f4 100644 > --- a/include/sysemu/kvm.h > +++ b/include/sysemu/kvm.h > @@ -428,11 +428,8 @@ int kvm_vm_check_extension(KVMState *s, unsigned i= nt extension); > .flags =3D cap_flags, = \ > }; \ > uint64_t args_tmp[] =3D { __VA_ARGS__ }; = \ > - int i; \ > - for (i =3D 0; i < (int)ARRAY_SIZE(args_tmp) && = \ > - i < ARRAY_SIZE(cap.args); i++) { \ > - cap.args[i] =3D args_tmp[i]; = \ > - } \ > + size_t n =3D MIN(ARRAY_SIZE(args_tmp), ARRAY_SIZE(cap.args)); = \ > + memcpy(cap.args, args_tmp, n * sizeof(cap.args[0])); \ > kvm_vm_ioctl(s, KVM_ENABLE_CAP, &cap); \ > }) > =20 > @@ -443,11 +440,8 @@ int kvm_vm_check_extension(KVMState *s, unsigned i= nt extension); > .flags =3D cap_flags, = \ > }; \ > uint64_t args_tmp[] =3D { __VA_ARGS__ }; = \ > - int i; \ > - for (i =3D 0; i < (int)ARRAY_SIZE(args_tmp) && = \ > - i < ARRAY_SIZE(cap.args); i++) { \ > - cap.args[i] =3D args_tmp[i]; = \ > - } \ > + size_t n =3D MIN(ARRAY_SIZE(args_tmp), ARRAY_SIZE(cap.args)); = \ > + memcpy(cap.args, args_tmp, n * sizeof(cap.args[0])); \ > kvm_vcpu_ioctl(cpu, KVM_ENABLE_CAP, &cap); \ > }) > =20 >=20