Re: [ptxdist] [PATCH v2] tf-a: new package for ARM trusted firmware A
On Tue, 18 Feb 2020 10:02:57 +0100, Guillermo Rodriguez Garcia wrote: > El mar., 18 feb. 2020 a las 8:36, Michael Olbrich > () escribió: > > > > On Mon, Feb 17, 2020 at 05:33:57PM +0100, Guillermo Rodriguez Garcia wrote: > > > > > El lun., 17 feb. 2020 a las 17:26, Michael Tretter > > > () escribió: > > > > > > > > On Wed, 12 Feb 2020 17:40:33 +0100, Ahmad Fatoum wrote: > > > > > Trusted Firmware-A (TF-A) is a reference implementation of secure > > > > > world > > > > > software for Arm A-Profile architectures (Armv8-A and Armv7-A). > > > > > > > > I successfully built the TF-A BL31 for the ZynqMP using this rule. > > > > > > > > However, I saw that the TF-A build uses the current git commitish for > > > > BUILD_STRING, which will be written into the binary. If I build the > > > > TF-A with this rule, this ends up to be the commitish of the BSP. I'm > > > > not sure if I actually want this, but I don't know what to put there > > > > instead. > > > > > > Since we finally agreed to keep TF_A_EXTRA_ARGS in the final TF-A rule > > > files, you can actually set BUILD_STRING to anything you want. Just > > > include BUILD_STRING=whatever in the TF_A_EXTRA_ARGS parameter. > > > > There should always be a default BUILD_STRING that does not depend on the > > BSP commit-ish. $(TF_A_VERSION) probably. > > OK, then we could add this to the .make file: > > TF_A_MAKE_OPT:= \ > CROSS_COMPILE=$(BOOTLOADER_CROSS_COMPILE) \ > HOSTCC=$(HOSTCC) \ > PLAT=$(PTXCONF_TF_A_PLATFORM) \ > ARCH=$(PTXCONF_TF_A_ARCH_STRING) \ > ARM_ARCH_MAJOR=$(PTXCONF_TF_A_ARM_ARCH_MAJOR) \ > BUILD_STRING=$(PTXCONF_TF_A_VERSION) \ The version string that is printed during the boot will then be something like "v2.2(Release):v2.2". The first version is the hard coded version in the makefile and the second is the git commitish that was used to build the tf-a. Seems OK to me. Michael > $(call remove_quotes,$(PTXCONF_TF_A_EXTRA_ARGS)) \ > all > > BUILD_STRING can still be overriden in TF_A_EXTRA_ARGS if needed. > > Guillermo > > ___ > ptxdist mailing list > ptxdist@pengutronix.de ___ ptxdist mailing list ptxdist@pengutronix.de
[ptxdist] Custom CFLAGS per packet (LTO masking)
Hi, I want to compile my systems with lto and need to exclude those packages which break with lto enabled. unfortunately I was not able to find a mechanism to filter out CFLAGS for certain packages. There are BLACKLIST variables, but they do not seem to do what I want, and I googling got me nowhere. So - is there a mechanism to use custom CFLAGS or filter the global flags? Flo -- Machines can do the work, so people have time to think. ___ ptxdist mailing list ptxdist@pengutronix.de
[ptxdist] [PATCH] unfs3: fix attribute setting for symlinks
unfs makes use of utime() which is unsuitable to set last access and modification times for symlinks. So make use of the O_PATH flag to open() and use the futimens() function. As a side effect this is less racy and improves resolution from seconds to nanoseconds. Note this will break builds of unfs on Windows. The futimens() function is specified in POSIX.1-2008, so this shouldn't be a problem for other platforms. Signed-off-by: Uwe Kleine-König --- Hello, note I didn't try to contact upstream. Best regards Uwe ...5-attr-use-futimens-instead-of-utime.patch | 102 ++ ...-O_PATH-to-set-attributes-of-symlink.patch | 81 ++ patches/unfs3-0.9.22/series | 4 +- 3 files changed, 186 insertions(+), 1 deletion(-) create mode 100644 patches/unfs3-0.9.22/0005-attr-use-futimens-instead-of-utime.patch create mode 100644 patches/unfs3-0.9.22/0006-attr-make-use-of-O_PATH-to-set-attributes-of-symlink.patch diff --git a/patches/unfs3-0.9.22/0005-attr-use-futimens-instead-of-utime.patch b/patches/unfs3-0.9.22/0005-attr-use-futimens-instead-of-utime.patch new file mode 100644 index ..401142fb1b16 --- /dev/null +++ b/patches/unfs3-0.9.22/0005-attr-use-futimens-instead-of-utime.patch @@ -0,0 +1,102 @@ +From: =?UTF-8?q?Uwe=20Kleine-K=C3=B6nig?= +Date: Tue, 18 Feb 2020 15:20:12 +0100 +Subject: [PATCH] attr: use futimens() instead of utime() + +The former has two relevant advantages: It works on file descriptors +instead of a path name and it has nano second resolution (instead of only +seconds). + +Note, I don't know about Windows which is probably broken by this commit +as I assume it doesn't support futimens(). +--- + attr.c | 53 - + 1 file changed, 48 insertions(+), 5 deletions(-) + +diff --git a/attr.c b/attr.c +index 2653ed0a16bb..08c9d24cb446 100644 +--- a/attr.c b/attr.c +@@ -265,12 +265,51 @@ post_op_attr get_post_cached(struct svc_req * req) + return get_post_buf(st_cache, req); + } + ++static nfsstat3 set_time(int fd, sattr3 new) ++{ ++if (new.atime.set_it != DONT_CHANGE || new.mtime.set_it != DONT_CHANGE) { ++ ++ /* atime in t[0], mtime in t[1] */ ++ struct timespec t[2]; ++ int res; ++ ++ /* compute atime to set */ ++ if (new.atime.set_it == SET_TO_SERVER_TIME) { ++ t[0].tv_sec = UTIME_NOW; ++ t[0].tv_nsec = UTIME_NOW; ++ } else if (new.atime.set_it == SET_TO_CLIENT_TIME) { ++ t[0].tv_sec = new.atime.set_atime_u.atime.seconds; ++ t[0].tv_nsec = new.atime.set_atime_u.atime.nseconds; ++ } else { /* DONT_CHANGE */ ++ t[0].tv_sec = UTIME_OMIT; ++ t[0].tv_nsec = UTIME_OMIT; ++ } ++ ++ /* compute mtime to set */ ++ if (new.mtime.set_it == SET_TO_SERVER_TIME) { ++ t[1].tv_sec = UTIME_NOW; ++ t[1].tv_nsec = UTIME_NOW; ++ } else if (new.mtime.set_it == SET_TO_CLIENT_TIME) { ++ t[1].tv_sec = new.mtime.set_mtime_u.mtime.seconds; ++ t[1].tv_nsec = new.mtime.set_mtime_u.mtime.nseconds; ++ } else { /* DONT_CHANGE */ ++ t[1].tv_sec = UTIME_OMIT; ++ t[1].tv_nsec = UTIME_OMIT; ++ } ++ ++ res = futimens(fd, ); ++ if (res == -1) ++ return setattr_err(); ++} ++return NFS3_OK; ++} ++ + /* + * setting of time, races with local filesystem + * + * there is no futimes() function in POSIX or Linux + */ +-static nfsstat3 set_time(const char *path, backend_statstruct buf, sattr3 new) ++static nfsstat3 set_time_path(const char *path, backend_statstruct buf, sattr3 new) + { + time_t new_atime, new_mtime; + struct utimbuf utim; +@@ -358,7 +397,7 @@ static nfsstat3 set_attr_unsafe(const char *path, nfs_fh3 nfh, sattr3 new) + return setattr_err(); + } + +-return set_time(path, buf, new); ++return set_time_path(path, buf, new); + } + + /* +@@ -454,14 +493,18 @@ nfsstat3 set_attr(const char *path, nfs_fh3 nfh, sattr3 new) + } + } + ++/* finally, set times */ ++res = set_time(fd, new); ++if (res != NFS3_OK) { ++ backend_close(fd); ++ return res; ++} ++ + res = backend_close(fd); + if (res == -1) { + /* error on close probably means attributes didn't make it */ + return NFS3ERR_IO; + } +- +-/* finally, set times */ +-return set_time(path, buf, new); + } + + /* diff --git a/patches/unfs3-0.9.22/0006-attr-make-use-of-O_PATH-to-set-attributes-of-symlink.patch b/patches/unfs3-0.9.22/0006-attr-make-use-of-O_PATH-to-set-attributes-of-symlink.patch new file mode 100644 index ..798927a520e6 --- /dev/null +++ b/patches/unfs3-0.9.22/0006-attr-make-use-of-O_PATH-to-set-attributes-of-symlink.patch @@ -0,0 +1,81 @@ +From: =?UTF-8?q?Uwe=20Kleine-K=C3=B6nig?= +Date: Tue, 18 Feb 2020 15:23:52 +0100 +Subject: [PATCH] attr: make use of O_PATH to set attributes of symlinks and + devices + +This fixes setting
[ptxdist] [PATCH] libinput: version bump 1.15.0 -> 1.15.1
Signed-off-by: Philipp Zabel --- rules/libinput.make | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-) diff --git a/rules/libinput.make b/rules/libinput.make index 6d1659f7174d..c0163fbfa465 100644 --- a/rules/libinput.make +++ b/rules/libinput.make @@ -14,8 +14,8 @@ PACKAGES-$(PTXCONF_LIBINPUT) += libinput # # Paths and names # -LIBINPUT_VERSION := 1.15.0 -LIBINPUT_MD5 := d5fe153c4b1d69ec35dc846588cd47ec +LIBINPUT_VERSION := 1.15.1 +LIBINPUT_MD5 := 93850756b6c07aafb49d5226e88fdc81 LIBINPUT := libinput-$(LIBINPUT_VERSION) LIBINPUT_SUFFIX:= tar.xz LIBINPUT_URL := http://www.freedesktop.org/software/libinput/$(LIBINPUT).$(LIBINPUT_SUFFIX) -- 2.20.1 ___ ptxdist mailing list ptxdist@pengutronix.de
Re: [ptxdist] [PATCH v2] tf-a: new package for ARM trusted firmware A
El mar., 18 feb. 2020 a las 8:36, Michael Olbrich () escribió: > > On Mon, Feb 17, 2020 at 05:33:57PM +0100, Guillermo Rodriguez Garcia wrote: > > El lun., 17 feb. 2020 a las 17:26, Michael Tretter > > () escribió: > > > > > > On Wed, 12 Feb 2020 17:40:33 +0100, Ahmad Fatoum wrote: > > > > Trusted Firmware-A (TF-A) is a reference implementation of secure world > > > > software for Arm A-Profile architectures (Armv8-A and Armv7-A). > > > > > > I successfully built the TF-A BL31 for the ZynqMP using this rule. > > > > > > However, I saw that the TF-A build uses the current git commitish for > > > BUILD_STRING, which will be written into the binary. If I build the > > > TF-A with this rule, this ends up to be the commitish of the BSP. I'm > > > not sure if I actually want this, but I don't know what to put there > > > instead. > > > > Since we finally agreed to keep TF_A_EXTRA_ARGS in the final TF-A rule > > files, you can actually set BUILD_STRING to anything you want. Just > > include BUILD_STRING=whatever in the TF_A_EXTRA_ARGS parameter. > > There should always be a default BUILD_STRING that does not depend on the > BSP commit-ish. $(TF_A_VERSION) probably. OK, then we could add this to the .make file: TF_A_MAKE_OPT:= \ CROSS_COMPILE=$(BOOTLOADER_CROSS_COMPILE) \ HOSTCC=$(HOSTCC) \ PLAT=$(PTXCONF_TF_A_PLATFORM) \ ARCH=$(PTXCONF_TF_A_ARCH_STRING) \ ARM_ARCH_MAJOR=$(PTXCONF_TF_A_ARM_ARCH_MAJOR) \ BUILD_STRING=$(PTXCONF_TF_A_VERSION) \ $(call remove_quotes,$(PTXCONF_TF_A_EXTRA_ARGS)) \ all BUILD_STRING can still be overriden in TF_A_EXTRA_ARGS if needed. Guillermo ___ ptxdist mailing list ptxdist@pengutronix.de