| Message ID | 20180620143448.44388-2-arnd@arndb.de |
|---|---|
| State | New |
| Headers | show |
| Series | [1/3] btrfs: use monotonic time for transaction handling | expand |
On Wed, Jun 20, 2018 at 04:34:33PM +0200, Arnd Bergmann wrote: > The structure already has 64-bit fields for the timestamps, but > calling get_seconds() may truncate and risk overflow on 32-bit > architectures. > > This changes the dev-replace code to use ktime_get_real_seconds() > instead, which always returns 64-bit timestamps. > > Signed-off-by: Arnd Bergmann <arnd@arndb.de> Thanks but there's a patch already fixng that, sent a few days ago https://patchwork.kernel.org/patch/10473195/ and added to patch queue for the next dev cycle as it does not appear to urgent for 4.18.
On Wed, Jun 20, 2018 at 4:36 PM, David Sterba <dsterba@suse.cz> wrote: > On Wed, Jun 20, 2018 at 04:34:33PM +0200, Arnd Bergmann wrote: >> The structure already has 64-bit fields for the timestamps, but >> calling get_seconds() may truncate and risk overflow on 32-bit >> architectures. >> >> This changes the dev-replace code to use ktime_get_real_seconds() >> instead, which always returns 64-bit timestamps. >> >> Signed-off-by: Arnd Bergmann <arnd@arndb.de> > > Thanks but there's a patch already fixng that, sent a few days ago > > https://patchwork.kernel.org/patch/10473195/ > > and added to patch queue for the next dev cycle as it does not appear to > urgent for 4.18. Ok, sounds good. I had missed that Allen has independently sent out some of the same patches that I created in the last weeks. Allen, do you have more patches pending? I have sent out most of what I did (around 80 patches I think), with just ext4, ceph, nfs and xfs pending at the moment. It seems we also sent identical patches for procfs and I have something pending for ceph that duplicates another patch you did. Arnd
diff --git a/fs/btrfs/dev-replace.c b/fs/btrfs/dev-replace.c index e2ba0419297a..1b30c38d05c9 100644 --- a/fs/btrfs/dev-replace.c +++ b/fs/btrfs/dev-replace.c @@ -465,7 +465,7 @@ int btrfs_dev_replace_start(struct btrfs_fs_info *fs_info, * go to the tgtdev as well (refer to btrfs_map_block()). */ dev_replace->replace_state = BTRFS_IOCTL_DEV_REPLACE_STATE_STARTED; - dev_replace->time_started = get_seconds(); + dev_replace->time_started = ktime_get_real_seconds(); dev_replace->cursor_left = 0; dev_replace->committed_cursor_left = 0; dev_replace->cursor_left_last_write_of_item = 0; @@ -618,7 +618,7 @@ static int btrfs_dev_replace_finishing(struct btrfs_fs_info *fs_info, : BTRFS_IOCTL_DEV_REPLACE_STATE_FINISHED; dev_replace->tgtdev = NULL; dev_replace->srcdev = NULL; - dev_replace->time_stopped = get_seconds(); + dev_replace->time_stopped = ktime_get_real_seconds(); dev_replace->item_needs_writeback = 1; /* replace old device with new one in mapping tree */ @@ -807,7 +807,7 @@ int btrfs_dev_replace_cancel(struct btrfs_fs_info *fs_info) break; } dev_replace->replace_state = BTRFS_IOCTL_DEV_REPLACE_STATE_CANCELED; - dev_replace->time_stopped = get_seconds(); + dev_replace->time_stopped = ktime_get_real_seconds(); dev_replace->item_needs_writeback = 1; btrfs_dev_replace_write_unlock(dev_replace); btrfs_scrub_cancel(fs_info); @@ -848,7 +848,7 @@ void btrfs_dev_replace_suspend_for_unmount(struct btrfs_fs_info *fs_info) case BTRFS_IOCTL_DEV_REPLACE_STATE_STARTED: dev_replace->replace_state = BTRFS_IOCTL_DEV_REPLACE_STATE_SUSPENDED; - dev_replace->time_stopped = get_seconds(); + dev_replace->time_stopped = ktime_get_real_seconds(); dev_replace->item_needs_writeback = 1; btrfs_info(fs_info, "suspending dev_replace for unmount"); break;
The structure already has 64-bit fields for the timestamps, but calling get_seconds() may truncate and risk overflow on 32-bit architectures. This changes the dev-replace code to use ktime_get_real_seconds() instead, which always returns 64-bit timestamps. Signed-off-by: Arnd Bergmann <arnd@arndb.de> --- fs/btrfs/dev-replace.c | 8 ++++---- 1 file changed, 4 insertions(+), 4 deletions(-) -- 2.9.0