On 06/25, 何云蕾(Yunlei he) wrote: > > On 2023/6/24 3:07, Jaegeuk Kim wrote: > > On 06/20, Chao Yu wrote: > > > On 2023/6/20 10:42, 何云蕾(Yunlei he) wrote: > > > > On 2023/6/20 8:33, Chao Yu wrote: > > > > > On 2023/6/13 16:52, Yunlei He wrote: > > > > > > File set both cold and hot advise bit is confusion, so > > > > > > return EINVAL to avoid this case. > > > > > > > > > > > > Signed-off-by: Yunlei He<heyun...@oppo.com> > > > > > > --- > > > > > > fs/f2fs/xattr.c | 3 +++ > > > > > > 1 file changed, 3 insertions(+) > > > > > > > > > > > > diff --git a/fs/f2fs/xattr.c b/fs/f2fs/xattr.c > > > > > > index 213805d3592c..917f3ac9f1a1 100644 > > > > > > --- a/fs/f2fs/xattr.c > > > > > > +++ b/fs/f2fs/xattr.c > > > > > > @@ -127,6 +127,9 @@ static int f2fs_xattr_advise_set(const struct > > > > > > xattr_handler *handler, > > > > > > return -EINVAL; > > > > > > > > > > > > new_advise = new_advise & FADVISE_MODIFIABLE_BITS; > > > > > > + if ((new_advise & FADVISE_COLD_BIT) && (new_advise & > > > > > > FADVISE_HOT_BIT)) > > > > > > + return -EINVAL; > > Why not this to allow setting one bit only? > > > > @@ -123,7 +123,8 @@ static int f2fs_xattr_advise_set(const struct > > xattr_handler *handler, > > return -EINVAL; > > > > new_advise = *(char *)value; > > - if (new_advise & ~FADVISE_MODIFIABLE_BITS) > > + if (new_advise & ~FADVISE_MODIFIABLE_BITS || > > + new_advise == FADVISE_MODIFIABLE_BITS) > > return -EINVAL; > > Hi,Jaegeuk, > > If new modifiable advise bit is added in the future, maybe multi-bits > is allowed? >
Looks like making a single bit assumption would be better in general at this moment. > Thanks > > > > > > > > Yunlei, > > > > > > > > > > What about the below case: > > > > > > > > > > 1. f2fs_xattr_advise_set(FADVISE_COLD_BIT) > > > > > 2. f2fs_xattr_advise_set(FADVISE_HOT_BIT) > > > > Hi, Chao, > > > > > > > > I test this case work well with this patch, case below will > > > > return -EINVAL: > > > > > > > > f2fs_xattr_advise_set(FADVISE_COLD_BIT | FADVISE_HOT_BIT) > > > Correct, I missed to check below statement. > > > > > > new_advise |= old_advise & ~FADVISE_MODIFIABLE_BITS; > > > > > > Anyway, the patch looks good to me. > > > > > > Reviewed-by: Chao Yu<c...@kernel.org> > > > > > > Thanks, _______________________________________________ Linux-f2fs-devel mailing list Linux-f2fs-devel@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/linux-f2fs-devel
