On Monday 03 July 2017, Karel Zak wrote: > On Mon, Jul 03, 2017 at 04:39:49PM +0200, Karel Zak wrote: > > On Mon, Jul 03, 2017 at 01:52:56PM +0200, Ruediger Meier wrote: > > > From: Ruediger Meier <ruediger.meier@xxxxxxxxxxx> > > > > > > path_vcreate_with_prefix() is going to be removed. > > > > > > Signed-off-by: Ruediger Meier <ruediger.meier@xxxxxxxxxxx> > > > --- > > > lib/path.c | 111 > > > ++++++++++++++++++++++++++++++++++++++++++++++++++++--------- 1 > > > file changed, 96 insertions(+), 15 deletions(-) > > > > > > diff --git a/lib/path.c b/lib/path.c > > > index 79c1e7a..94181fe 100644 > > > --- a/lib/path.c > > > +++ b/lib/path.c > > > @@ -26,18 +26,60 @@ > > > #include <stdio.h> > > > #include <inttypes.h> > > > #include <errno.h> > > > +#include <sys/stat.h> > > > > > > #include "all-io.h" > > > #include "path.h" > > > #include "nls.h" > > > #include "c.h" > > > > > > +static int rootfd = -1; > > > static size_t prefixlen; > > > +static char pathbuf_prefix[PATH_MAX]; > > > static char pathbuf[PATH_MAX]; > > > > We use sysfs.c in shared libs, it means that global variables are > > unacceptable. Yes I understand. This global way was just for first tests without changing much code. To be honest, while I was very happy first about quick results, I am now a bit unsure whether this sysroot feature will be much useful at all. See, even lscpu --sysroot does not work fully transparently. We have to ignore "Architecture". So maybe we would also need a fake-syscall interface for many cases!? Also we are sometimes writing to /sys and probably expect some events. I wonder if it wouldn't be easier to construct chroot or container environments with whatever /sys, /proc, /etc. to test certain things. So I thought we could use my quick-and-dirty approach to implement some more --sysroot tests just to see if we will get something useful at all. cu, Rudi > > It would be better to have "struct path_cxt" and use it within > > "struct sysfs_cxt". For tools where we don't use "struct sysfs_cxt" > > (e.g. lscpu) we can directly use "struct path_cxt". > > > > I guess it should be possible to replace sysfs_cxt->dir_fd with > > "struct path_cxt" where it points to $SYSROOT/sysfs/block/<device>. > And if we want to be real perfectionists than in sysfs.c we should > differentiate between block functions ($SYSROOT/sysfs/block/) and > another targets ($SYSROOT/sys/devices/system/cpu, ...etc). > > Examples: > > path.c only code: > > uint64_t x; > struct path_cxt *path = path_new_context(); > > path_set_system_root(path, argv[1]); > > path_read_s32(path, _PATH_SYS_CPU "/kernel_max", &x); > > > sysfs based code: > > uint64_t x; > struct path_cxt *path = path_new_context(); > > path_set_system_root(path, argv[1]); > > sysfs = sysfs_new_context(path); > sysfs_assign_block_device(sysfs, devno, NULL); > > sysfs_read_u64(sysfs, "queue/minimum_io_size", &x); > > > Do you see more light in the proposal now? :-) -- To unsubscribe from this list: send the line "unsubscribe util-linux" in the body of a message to majordomo@xxxxxxxxxxxxxxx More majordomo info at http://vger.kernel.org/majordomo-info.html
