| Message ID | 20180619155826.4106487-1-arnd@arndb.de (mailing list archive) |
|---|---|
| State | New, archived |
| Headers | show |
On 06-19 17:58, Arnd Bergmann wrote: > The handling of timestamps outside of the 1970..2038 range in the dlm > glue is rather inconsistent: on 32-bit architectures, this has always > wrapped around to negative timestamps in the 1902..1969 range, while on > 64-bit kernels all timestamps are interpreted as positive 34 bit numbers > in the 1970..2514 year range. > > Now that the VFS code handles 64-bit timestamps on all architectures, > we can make the behavior more consistent here, and return the same result > that we had on 64-bit already, making the file system y2038 safe in the > process. Outside of dlmglue, it already uses 64-bit on-disk timestamps > anway, so that part is fine. > > For consistency, I'm changing ocfs2_pack_timespec() to clamp > anything outside of the supported range to the minimum and maximum > values. This avoids a possible ambiguity of values before 1970 > in particular, which used to be interpreted as times at the end of the > 2514 range previously. > > Signed-off-by: Arnd Bergmann <arnd@arndb.de> Will all values written to LVB be the same with or without the patch? I am considering the situation where in a cluster some machines have this patch and some don't. Depending on that, this may require a version change. > --- > fs/ocfs2/dlmglue.c | 26 +++++++++----------------- > 1 file changed, 9 insertions(+), 17 deletions(-) > > diff --git a/fs/ocfs2/dlmglue.c b/fs/ocfs2/dlmglue.c > index 0ff424c6d17c..50610a9ed9f4 100644 > --- a/fs/ocfs2/dlmglue.c > +++ b/fs/ocfs2/dlmglue.c > @@ -2121,10 +2121,10 @@ static void ocfs2_downconvert_on_unlock(struct ocfs2_super *osb, > > /* LVB only has room for 64 bits of time here so we pack it for > * now. */ > -static u64 ocfs2_pack_timespec(struct timespec *spec) > +static u64 ocfs2_pack_timespec(struct timespec64 *spec) > { > u64 res; > - u64 sec = spec->tv_sec; > + u64 sec = clamp_t(time64_t, spec->tv_sec, 0, 0x3ffffffffull); > u32 nsec = spec->tv_nsec; > > res = (sec << OCFS2_SEC_SHIFT) | (nsec & OCFS2_NSEC_MASK); > @@ -2140,7 +2140,6 @@ static void __ocfs2_stuff_meta_lvb(struct inode *inode) > struct ocfs2_inode_info *oi = OCFS2_I(inode); > struct ocfs2_lock_res *lockres = &oi->ip_inode_lockres; > struct ocfs2_meta_lvb *lvb; > - struct timespec ts; > > lvb = ocfs2_dlm_lvb(&lockres->l_lksb); > > @@ -2161,15 +2160,12 @@ static void __ocfs2_stuff_meta_lvb(struct inode *inode) > lvb->lvb_igid = cpu_to_be32(i_gid_read(inode)); > lvb->lvb_imode = cpu_to_be16(inode->i_mode); > lvb->lvb_inlink = cpu_to_be16(inode->i_nlink); > - ts = timespec64_to_timespec(inode->i_atime); > lvb->lvb_iatime_packed = > - cpu_to_be64(ocfs2_pack_timespec(&ts)); > - ts = timespec64_to_timespec(inode->i_ctime); > + cpu_to_be64(ocfs2_pack_timespec(&inode->i_atime)); > lvb->lvb_ictime_packed = > - cpu_to_be64(ocfs2_pack_timespec(&ts)); > - ts = timespec64_to_timespec(inode->i_mtime); > + cpu_to_be64(ocfs2_pack_timespec(&inode->i_ctime)); > lvb->lvb_imtime_packed = > - cpu_to_be64(ocfs2_pack_timespec(&ts)); > + cpu_to_be64(ocfs2_pack_timespec(&inode->i_mtime)); > lvb->lvb_iattr = cpu_to_be32(oi->ip_attr); > lvb->lvb_idynfeatures = cpu_to_be16(oi->ip_dyn_features); > lvb->lvb_igeneration = cpu_to_be32(inode->i_generation); > @@ -2178,7 +2174,7 @@ static void __ocfs2_stuff_meta_lvb(struct inode *inode) > mlog_meta_lvb(0, lockres); > } > > -static void ocfs2_unpack_timespec(struct timespec *spec, > +static void ocfs2_unpack_timespec(struct timespec64 *spec, > u64 packed_time) > { > spec->tv_sec = packed_time >> OCFS2_SEC_SHIFT; > @@ -2187,7 +2183,6 @@ static void ocfs2_unpack_timespec(struct timespec *spec, > > static void ocfs2_refresh_inode_from_lvb(struct inode *inode) > { > - struct timespec ts; > struct ocfs2_inode_info *oi = OCFS2_I(inode); > struct ocfs2_lock_res *lockres = &oi->ip_inode_lockres; > struct ocfs2_meta_lvb *lvb; > @@ -2215,15 +2210,12 @@ static void ocfs2_refresh_inode_from_lvb(struct inode *inode) > i_gid_write(inode, be32_to_cpu(lvb->lvb_igid)); > inode->i_mode = be16_to_cpu(lvb->lvb_imode); > set_nlink(inode, be16_to_cpu(lvb->lvb_inlink)); > - ocfs2_unpack_timespec(&ts, > + ocfs2_unpack_timespec(&inode->i_atime, > be64_to_cpu(lvb->lvb_iatime_packed)); > - inode->i_atime = timespec_to_timespec64(ts); > - ocfs2_unpack_timespec(&ts, > + ocfs2_unpack_timespec(&inode->i_mtime, > be64_to_cpu(lvb->lvb_imtime_packed)); > - inode->i_mtime = timespec_to_timespec64(ts); > - ocfs2_unpack_timespec(&ts, > + ocfs2_unpack_timespec(&inode->i_ctime, > be64_to_cpu(lvb->lvb_ictime_packed)); > - inode->i_ctime = timespec_to_timespec64(ts); > spin_unlock(&oi->ip_lock); > } > > -- > 2.9.0 > > > _______________________________________________ > Ocfs2-devel mailing list > Ocfs2-devel@oss.oracle.com > https://oss.oracle.com/mailman/listinfo/ocfs2-devel
On Tue, Jun 19, 2018 at 7:14 PM, Goldwyn Rodrigues <rgoldwyn@suse.de> wrote: > > > On 06-19 17:58, Arnd Bergmann wrote: >> The handling of timestamps outside of the 1970..2038 range in the dlm >> glue is rather inconsistent: on 32-bit architectures, this has always >> wrapped around to negative timestamps in the 1902..1969 range, while on >> 64-bit kernels all timestamps are interpreted as positive 34 bit numbers >> in the 1970..2514 year range. ... > > Will all values written to LVB be the same with or without the patch? > I am considering the situation where in a cluster some machines have this > patch and some don't. Depending on that, this may require a version > change. There is one part that may change: >> -static u64 ocfs2_pack_timespec(struct timespec *spec) >> +static u64 ocfs2_pack_timespec(struct timespec64 *spec) >> { >> u64 res; >> - u64 sec = spec->tv_sec; >> + u64 sec = clamp_t(time64_t, spec->tv_sec, 0, 0x3ffffffffull); >> u32 nsec = spec->tv_nsec; >> >> res = (sec << OCFS2_SEC_SHIFT) | (nsec & OCFS2_NSEC_MASK); Here, setting a timestamp before 1970 or after 2514 will get wrapped around in unpatched kernels, but will be clamped to the minimum and maximum times after the patch. It is extremely rare for correct code to need timestamps outside of that range, but it is also trivial to trigger that with a manual 'touch' command from user space. If the change is a problem, I can resend the patch without that one line change. Arnd
On 06-19 21:11, Arnd Bergmann wrote: > On Tue, Jun 19, 2018 at 7:14 PM, Goldwyn Rodrigues <rgoldwyn@suse.de> wrote: > > > > > > On 06-19 17:58, Arnd Bergmann wrote: > >> The handling of timestamps outside of the 1970..2038 range in the dlm > >> glue is rather inconsistent: on 32-bit architectures, this has always > >> wrapped around to negative timestamps in the 1902..1969 range, while on > >> 64-bit kernels all timestamps are interpreted as positive 34 bit numbers > >> in the 1970..2514 year range. > ... > > > > Will all values written to LVB be the same with or without the patch? > > I am considering the situation where in a cluster some machines have this > > patch and some don't. Depending on that, this may require a version > > change. > > There is one part that may change: > > >> -static u64 ocfs2_pack_timespec(struct timespec *spec) > >> +static u64 ocfs2_pack_timespec(struct timespec64 *spec) > >> { > >> u64 res; > >> - u64 sec = spec->tv_sec; > >> + u64 sec = clamp_t(time64_t, spec->tv_sec, 0, 0x3ffffffffull); > >> u32 nsec = spec->tv_nsec; > >> > >> res = (sec << OCFS2_SEC_SHIFT) | (nsec & OCFS2_NSEC_MASK); > > Here, setting a timestamp before 1970 or after 2514 will get wrapped > around in unpatched kernels, but will be clamped to the minimum > and maximum times after the patch. > > It is extremely rare for correct code to need timestamps outside of that > range, but it is also trivial to trigger that with a manual 'touch' command > from user space. > > If the change is a problem, I can resend the patch without that one > line change. > I think you should keep the change, but incrment OCFS2_LVB_VERSION.
On Tue, Jun 19, 2018 at 11:52 PM, Goldwyn Rodrigues <rgoldwyn@suse.de> wrote: > On 06-19 21:11, Arnd Bergmann wrote: >> On Tue, Jun 19, 2018 at 7:14 PM, Goldwyn Rodrigues <rgoldwyn@suse.de> wrote: >> > On 06-19 17:58, Arnd Bergmann wrote: >> Here, setting a timestamp before 1970 or after 2514 will get wrapped >> around in unpatched kernels, but will be clamped to the minimum >> and maximum times after the patch. >> >> It is extremely rare for correct code to need timestamps outside of that >> range, but it is also trivial to trigger that with a manual 'touch' command >> from user space. >> >> If the change is a problem, I can resend the patch without that one >> line change. >> > > I think you should keep the change, but incrment OCFS2_LVB_VERSION. Won't that cause additional incompatibilities? I don't know how this macro gets used, but normally we don't use version numbers in kernel interfaces if that prevents us from using old user space code with newer kernels. Arnd
diff --git a/fs/ocfs2/dlmglue.c b/fs/ocfs2/dlmglue.c index 0ff424c6d17c..50610a9ed9f4 100644 --- a/fs/ocfs2/dlmglue.c +++ b/fs/ocfs2/dlmglue.c @@ -2121,10 +2121,10 @@ static void ocfs2_downconvert_on_unlock(struct ocfs2_super *osb, /* LVB only has room for 64 bits of time here so we pack it for * now. */ -static u64 ocfs2_pack_timespec(struct timespec *spec) +static u64 ocfs2_pack_timespec(struct timespec64 *spec) { u64 res; - u64 sec = spec->tv_sec; + u64 sec = clamp_t(time64_t, spec->tv_sec, 0, 0x3ffffffffull); u32 nsec = spec->tv_nsec; res = (sec << OCFS2_SEC_SHIFT) | (nsec & OCFS2_NSEC_MASK); @@ -2140,7 +2140,6 @@ static void __ocfs2_stuff_meta_lvb(struct inode *inode) struct ocfs2_inode_info *oi = OCFS2_I(inode); struct ocfs2_lock_res *lockres = &oi->ip_inode_lockres; struct ocfs2_meta_lvb *lvb; - struct timespec ts; lvb = ocfs2_dlm_lvb(&lockres->l_lksb); @@ -2161,15 +2160,12 @@ static void __ocfs2_stuff_meta_lvb(struct inode *inode) lvb->lvb_igid = cpu_to_be32(i_gid_read(inode)); lvb->lvb_imode = cpu_to_be16(inode->i_mode); lvb->lvb_inlink = cpu_to_be16(inode->i_nlink); - ts = timespec64_to_timespec(inode->i_atime); lvb->lvb_iatime_packed = - cpu_to_be64(ocfs2_pack_timespec(&ts)); - ts = timespec64_to_timespec(inode->i_ctime); + cpu_to_be64(ocfs2_pack_timespec(&inode->i_atime)); lvb->lvb_ictime_packed = - cpu_to_be64(ocfs2_pack_timespec(&ts)); - ts = timespec64_to_timespec(inode->i_mtime); + cpu_to_be64(ocfs2_pack_timespec(&inode->i_ctime)); lvb->lvb_imtime_packed = - cpu_to_be64(ocfs2_pack_timespec(&ts)); + cpu_to_be64(ocfs2_pack_timespec(&inode->i_mtime)); lvb->lvb_iattr = cpu_to_be32(oi->ip_attr); lvb->lvb_idynfeatures = cpu_to_be16(oi->ip_dyn_features); lvb->lvb_igeneration = cpu_to_be32(inode->i_generation); @@ -2178,7 +2174,7 @@ static void __ocfs2_stuff_meta_lvb(struct inode *inode) mlog_meta_lvb(0, lockres); } -static void ocfs2_unpack_timespec(struct timespec *spec, +static void ocfs2_unpack_timespec(struct timespec64 *spec, u64 packed_time) { spec->tv_sec = packed_time >> OCFS2_SEC_SHIFT; @@ -2187,7 +2183,6 @@ static void ocfs2_unpack_timespec(struct timespec *spec, static void ocfs2_refresh_inode_from_lvb(struct inode *inode) { - struct timespec ts; struct ocfs2_inode_info *oi = OCFS2_I(inode); struct ocfs2_lock_res *lockres = &oi->ip_inode_lockres; struct ocfs2_meta_lvb *lvb; @@ -2215,15 +2210,12 @@ static void ocfs2_refresh_inode_from_lvb(struct inode *inode) i_gid_write(inode, be32_to_cpu(lvb->lvb_igid)); inode->i_mode = be16_to_cpu(lvb->lvb_imode); set_nlink(inode, be16_to_cpu(lvb->lvb_inlink)); - ocfs2_unpack_timespec(&ts, + ocfs2_unpack_timespec(&inode->i_atime, be64_to_cpu(lvb->lvb_iatime_packed)); - inode->i_atime = timespec_to_timespec64(ts); - ocfs2_unpack_timespec(&ts, + ocfs2_unpack_timespec(&inode->i_mtime, be64_to_cpu(lvb->lvb_imtime_packed)); - inode->i_mtime = timespec_to_timespec64(ts); - ocfs2_unpack_timespec(&ts, + ocfs2_unpack_timespec(&inode->i_ctime, be64_to_cpu(lvb->lvb_ictime_packed)); - inode->i_ctime = timespec_to_timespec64(ts); spin_unlock(&oi->ip_lock); }
The handling of timestamps outside of the 1970..2038 range in the dlm glue is rather inconsistent: on 32-bit architectures, this has always wrapped around to negative timestamps in the 1902..1969 range, while on 64-bit kernels all timestamps are interpreted as positive 34 bit numbers in the 1970..2514 year range. Now that the VFS code handles 64-bit timestamps on all architectures, we can make the behavior more consistent here, and return the same result that we had on 64-bit already, making the file system y2038 safe in the process. Outside of dlmglue, it already uses 64-bit on-disk timestamps anway, so that part is fine. For consistency, I'm changing ocfs2_pack_timespec() to clamp anything outside of the supported range to the minimum and maximum values. This avoids a possible ambiguity of values before 1970 in particular, which used to be interpreted as times at the end of the 2514 range previously. Signed-off-by: Arnd Bergmann <arnd@arndb.de> --- fs/ocfs2/dlmglue.c | 26 +++++++++----------------- 1 file changed, 9 insertions(+), 17 deletions(-)