On Thursday 16 January 2020 11:23:07 Christoph Hellwig wrote: > On Thu, Jan 16, 2020 at 11:19:47AM +0100, Pali Rohár wrote: > > However, implementations should only record the value 00h for this > > field when: > > > > 1. Local date and time are actually the same as UTC, in which case > > the value of the OffsetValid field shall be 1 > > > > 2. Local date and time are not known, in which case the value of the > > OffsetValid field shall be 1 and implementations shall consider > > UTC to be local date and time > > Given time zones in Linux are per session I think our situation is > somewhat similar to 2. Seems that 2. is really similar. Ok, storing by default in UTC make sense, but still I'm thinking if there should be (or not) a mount option which override default UTC timezone. > > > Here I would just convert to UTC, which is what we store in the > > > in-memory struct inode anyway. > > > > Ok. If inode timestamp is always in UTC, we should do same thing also > > for exFAT. > > > Hm... both UTC and sys_tz have positives and negatives. And I'm not > > sure which option is better. > > The one big argument for always UTC is simplicity. Always using UTC > kills some arcane an unusual (for Linux file systems) code, and given > how exfat implementations deal with the time zone on reading should > always interoperate fine with other implementations. Now I think that using UTC by default is the better option as sys_tz. Simplicity and "no surprise" (container may use UTC, but kernel has sys_tz in not in UTC) seems like a good arguments. -- Pali Rohár pali.rohar@xxxxxxxxx
