Message ID | 48287dd12fcef0440cc49ff86343867d6cb8429e.1533656185.git.jan.kundrat@cesnet.cz |
---|---|
State | Changes Requested |
Headers | show |
Series | mmc-utils: Rely on our own _FORTIFY_SOURCE | expand |
On úterý 7. srpna 2018 17:32:51 CEST, Jan Kundrát wrote: > Upstream always forced _FORTIFY_SOURCE=2. At first, I tried changing > their flags to undefine that macro first, but that did not work well. > This patch ensures that the package always respects our toolchain > settings, even in cases where we are not fortifying the build. > > This fixes a build failure where the mmc-utils package failed to build > with, e.g., BR2_FORTIFY_SOURCE_1: > > <command-line>:0:0: error: "_FORTIFY_SOURCE" redefined [-Werror] > <command-line>:0:0: note: this is the location of the previous definition > > Signed-off-by: Jan Kundrát <jan.kundrat@cesnet.cz> > --- > ...on-the-build-env-for-_FORTIFY_SOURCE.patch | 27 +++++++++++++++++++ > 1 file changed, 27 insertions(+) > create mode 100644 > package/mmc-utils/0001-Rely-on-the-build-env-for-_FORTIFY_SOURCE.patch > > diff --git > a/package/mmc-utils/0001-Rely-on-the-build-env-for-_FORTIFY_SOURCE.patch > b/package/mmc-utils/0001-Rely-on-the-build-env-for-_FORTIFY_SOURCE.patch > new file mode 100644 > index 0000000000..b099efa6c4 > --- /dev/null > +++ b/package/mmc-utils/0001-Rely-on-the-build-env-for-_FORTIFY_SOURCE.patch > @@ -0,0 +1,27 @@ > +From 0c893e6f272351572548264bf423208a7b76bb16 Mon Sep 17 00:00:00 2001 > +From: =?UTF-8?q?Jan=20Kundr=C3=A1t?= <jan.kundrat@cesnet.cz> > +Date: Tue, 7 Aug 2018 17:29:35 +0200 > +Subject: [PATCH] Rely on the build env for _FORTIFY_SOURCE > +MIME-Version: 1.0 > +Content-Type: text/plain; charset=UTF-8 > +Content-Transfer-Encoding: 8bit > + > +Signed-off-by: Jan Kundrát <jan.kundrat@cesnet.cz> > +--- > + Makefile | 2 +- > + 1 file changed, 1 insertion(+), 1 deletion(-) > + > +diff --git a/Makefile b/Makefile > +index aa27ff2..a3890b2 100644 > +--- a/Makefile > ++++ b/Makefile > +@@ -1,5 +1,5 @@ > + CC ?= gcc > +-AM_CFLAGS = -D_FILE_OFFSET_BITS=64 -D_FORTIFY_SOURCE=2 > ++AM_CFLAGS = -D_FILE_OFFSET_BITS=64 > + CFLAGS ?= -g -O2 > + objects = \ > + mmc.o \ > +-- > +2.17.1 > + This patch might not be needed when Matthew Weber's series ([PATCH 0/6] Hardening Flag Bugfix/Enhancement) lands. My recommendation is to hold it for a while. With kind regards, Jan
Hello, On Tue, 7 Aug 2018 17:32:51 +0200, Jan Kundrát wrote: > Upstream always forced _FORTIFY_SOURCE=2. At first, I tried changing > their flags to undefine that macro first, but that did not work well. > This patch ensures that the package always respects our toolchain > settings, even in cases where we are not fortifying the build. > > This fixes a build failure where the mmc-utils package failed to build > with, e.g., BR2_FORTIFY_SOURCE_1: > > <command-line>:0:0: error: "_FORTIFY_SOURCE" redefined [-Werror] > <command-line>:0:0: note: this is the location of the previous definition > > Signed-off-by: Jan Kundrát <jan.kundrat@cesnet.cz> Is this fixing a build issue reported by the autobuilders ? Or only with hardening options ? Your commit log is not very verbose in details about the configuration/condition under which the build failure happens. Thanks! Thomas
All, On Wed, Aug 8, 2018 at 8:17 AM Thomas Petazzoni <thomas.petazzoni@bootlin.com> wrote: > > Hello, > > On Tue, 7 Aug 2018 17:32:51 +0200, Jan Kundrát wrote: > > Upstream always forced _FORTIFY_SOURCE=2. At first, I tried changing > > their flags to undefine that macro first, but that did not work well. > > This patch ensures that the package always respects our toolchain > > settings, even in cases where we are not fortifying the build. > > > > This fixes a build failure where the mmc-utils package failed to build > > with, e.g., BR2_FORTIFY_SOURCE_1: > > > > <command-line>:0:0: error: "_FORTIFY_SOURCE" redefined [-Werror] > > <command-line>:0:0: note: this is the location of the previous definition > > > > Signed-off-by: Jan Kundrát <jan.kundrat@cesnet.cz> > > Is this fixing a build issue reported by the autobuilders ? Or only > with hardening options ? Your commit log is not very verbose in details > about the configuration/condition under which the build failure happens. (I haven't directly talked with Jan on this) It looks very similar to other pkgs him and I looked at where flags were messy in the pkg and now exposed when a build is done using hardening options, not a autobuilder. This isn't related to my hardening fixes patchset (ie those patches won't fix this issue). Instead, I'd refactor this patch to move the AM_CFLAGS "-D_FORTIFY_SOURCE=2" into the CFLAGS ?= . That should be an upstream-able change and preserve their intent. Matt
On pátek 10. srpna 2018 4:56:39 CEST, Matthew Weber wrote: > This isn't related to my hardening fixes patchset (ie those patches > won't fix this issue). Instead, I'd refactor this patch to move the > AM_CFLAGS "-D_FORTIFY_SOURCE=2" into the CFLAGS ?= . That should be > an upstream-able change and preserve their intent. Hi Matthew, doing just this change, as you suggested: -AM_CFLAGS = -D_FILE_OFFSET_BITS=64 -D_FORTIFY_SOURCE=2 -CFLAGS ?= -g -O2 +AM_CFLAGS = -D_FILE_OFFSET_BITS=64 +CFLAGS ?= -g -O2 -D_FORTIFY_SOURCE=2 ...stil results in an error when building this manually out-of-box, outside of Buildroot, on a Gentoo Linux with hardened GCC which enforces _FORTIFY_SOURCE via compiler spec files. The good news is that it is now possible to override the CFLAGS via an env var when invoking make. I guess I'll send this upstream, then. It still doesn't work out-of-box everyhere, but it is no longer a Buildroot bug (and it will work in Buildroot). With kind regards, Jan
Jan, On Thu, Oct 25, 2018 at 1:03 PM Jan Kundrát <jan.kundrat@cesnet.cz> wrote: > > On pátek 10. srpna 2018 4:56:39 CEST, Matthew Weber wrote: > > This isn't related to my hardening fixes patchset (ie those patches > > won't fix this issue). Instead, I'd refactor this patch to move the > > AM_CFLAGS "-D_FORTIFY_SOURCE=2" into the CFLAGS ?= . That should be > > an upstream-able change and preserve their intent. > > Hi Matthew, > doing just this change, as you suggested: > > -AM_CFLAGS = -D_FILE_OFFSET_BITS=64 -D_FORTIFY_SOURCE=2 > -CFLAGS ?= -g -O2 > +AM_CFLAGS = -D_FILE_OFFSET_BITS=64 > +CFLAGS ?= -g -O2 -D_FORTIFY_SOURCE=2 > > ...stil results in an error when building this manually out-of-box, outside > of Buildroot, on a Gentoo Linux with hardened GCC which enforces > _FORTIFY_SOURCE via compiler spec files. The good news is that it is now > possible to override the CFLAGS via an env var when invoking make. I guess > I'll send this upstream, then. > > It still doesn't work out-of-box everyhere, but it is no longer a Buildroot > bug (and it will work in Buildroot). Good to note. Wasn't sure if you saw the merge of the updated hardening feature set. It is now handled in the toolchain wrapper and more packages (without changes directly in the package to allow flags to be set) should build with the hardening flags. The wrapper appends them to the compiler args transparently and doesn't depend on ?=. Matt
diff --git a/package/mmc-utils/0001-Rely-on-the-build-env-for-_FORTIFY_SOURCE.patch b/package/mmc-utils/0001-Rely-on-the-build-env-for-_FORTIFY_SOURCE.patch new file mode 100644 index 0000000000..b099efa6c4 --- /dev/null +++ b/package/mmc-utils/0001-Rely-on-the-build-env-for-_FORTIFY_SOURCE.patch @@ -0,0 +1,27 @@ +From 0c893e6f272351572548264bf423208a7b76bb16 Mon Sep 17 00:00:00 2001 +From: =?UTF-8?q?Jan=20Kundr=C3=A1t?= <jan.kundrat@cesnet.cz> +Date: Tue, 7 Aug 2018 17:29:35 +0200 +Subject: [PATCH] Rely on the build env for _FORTIFY_SOURCE +MIME-Version: 1.0 +Content-Type: text/plain; charset=UTF-8 +Content-Transfer-Encoding: 8bit + +Signed-off-by: Jan Kundrát <jan.kundrat@cesnet.cz> +--- + Makefile | 2 +- + 1 file changed, 1 insertion(+), 1 deletion(-) + +diff --git a/Makefile b/Makefile +index aa27ff2..a3890b2 100644 +--- a/Makefile ++++ b/Makefile +@@ -1,5 +1,5 @@ + CC ?= gcc +-AM_CFLAGS = -D_FILE_OFFSET_BITS=64 -D_FORTIFY_SOURCE=2 ++AM_CFLAGS = -D_FILE_OFFSET_BITS=64 + CFLAGS ?= -g -O2 + objects = \ + mmc.o \ +-- +2.17.1 +
Upstream always forced _FORTIFY_SOURCE=2. At first, I tried changing their flags to undefine that macro first, but that did not work well. This patch ensures that the package always respects our toolchain settings, even in cases where we are not fortifying the build. This fixes a build failure where the mmc-utils package failed to build with, e.g., BR2_FORTIFY_SOURCE_1: <command-line>:0:0: error: "_FORTIFY_SOURCE" redefined [-Werror] <command-line>:0:0: note: this is the location of the previous definition Signed-off-by: Jan Kundrát <jan.kundrat@cesnet.cz> --- ...on-the-build-env-for-_FORTIFY_SOURCE.patch | 27 +++++++++++++++++++ 1 file changed, 27 insertions(+) create mode 100644 package/mmc-utils/0001-Rely-on-the-build-env-for-_FORTIFY_SOURCE.patch