This fixes two problems in the handling of negative times:
• rem is signed, but the rem * c->sb.nsec_per_time_unit operation
produced a bogus unsigned result, because s32 * u32 = u32.
• The timespec was not normalized (it could contain more than a
billion nanoseconds).
For example, { .tv_sec = -14245441, .tv_nsec = 750000000 }, after
being round tripped through timespec_to_bch2_time and then
bch2_time_to_timespec would come back as
{ .tv_sec = -14245440, .tv_nsec = 4044967296 } (more than 4 billion
nanoseconds).
Cc: stable@xxxxxxxxxxxxxxx
Fixes: 595c1e9bab7f ("bcachefs: Fix time handling")
Closes: https://github.com/koverstreet/bcachefs/issues/743
Co-developed-by: Erin Shepherd <erin.shepherd@xxxxxx>
Signed-off-by: Erin Shepherd <erin.shepherd@xxxxxx>
Co-developed-by: Ryan Lahfa <ryan@xxxxxxxxx>
Signed-off-by: Ryan Lahfa <ryan@xxxxxxxxx>
Signed-off-by: Alyssa Ross <hi@xxxxxxxxx>
---
I've submitted an RFC to fstests to add a regression test for this:
https://lore.kernel.org/fstests/20240907154527.604864-2-hi@xxxxxxxxx/
fs/bcachefs/bcachefs.h | 7 +++++--
1 file changed, 5 insertions(+), 2 deletions(-)
diff --git a/fs/bcachefs/bcachefs.h b/fs/bcachefs/bcachefs.h
index 0c7086e00d18..81c4d935cca8 100644
--- a/fs/bcachefs/bcachefs.h
+++ b/fs/bcachefs/bcachefs.h
@@ -1195,12 +1195,15 @@ static inline bool btree_id_cached(const struct bch_fs *c, enum btree_id btree)
static inline struct timespec64 bch2_time_to_timespec(const struct bch_fs *c, s64 time)
{
struct timespec64 t;
+ s64 sec;
s32 rem;
time += c->sb.time_base_lo;
- t.tv_sec = div_s64_rem(time, c->sb.time_units_per_sec, &rem);
- t.tv_nsec = rem * c->sb.nsec_per_time_unit;
+ sec = div_s64_rem(time, c->sb.time_units_per_sec, &rem);
+
+ set_normalized_timespec64(&t, sec, rem * (s64)c->sb.nsec_per_time_unit);
+
return t;
}
base-commit: 53f6619554fb1edf8d7599b560d44dbea085c730
--
2.45.2
