On Fri, May 15, 2020 at 09:53:46AM +0000, Johannes Thumshirn wrote: > On 15/05/2020 10:50, Chris Hofstaedtler wrote: > > * Johannes Thumshirn <johannes.thumshirn@xxxxxxx> [200515 10:41]: > >> If a user submits a zone management ioctl from user-space, like a zone > >> reset and a file-system (like zonefs or f2fs) is mounted on the zoned > >> block device, the zone will get reset and the file-system's cached value > >> of the zone's write-pointer becomes invalid. > >> > >> Subsequent writes to this zone from the file-system will result in > >> unaligned writes and the drive will error out. > > > > "error out" meaning what exactly? > > The drive will report an Unaligned Write error. > > > > >> Open the block device file in exclusive mode for submitting these ioctls. > >> If a file-system is mounted the kernel will return -EBUSY and we can't > >> continue issuing the ioctl. > > > > Isn't this something the kernel should enforce, then? > > I did a patch for the kernel yesterday [1] enforcing this limitation, but as > Damien said it's SYS_CAP_ADMIN and with great power comes great responsibility. > We're also allowing other raw block device accesses on block devices. > > [1] https://lore.kernel.org/linux-block/BY5PR04MB69006DE86D1050620B5EDAA4E7BD0@xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx/ > > > What's to stop anybody from calling the ioctl from another tool, > > without using O_EXCL? > > Nothing, but still we don't need to make one's life hard by letting blkzone go > havoc with FS internal caching. If another tool does zone resets, it's up to them > to check for mounted block devices. O_EXCL good idea, it nothing unusual that you can do crazy things with devices with mounted filesystem (see for example fdisk(s), wipefs, ...). And sometimes it is valid use-case to do "bad" things and in this case we have --force option in our tools (for example to avoid O_EXCL). Karel -- Karel Zak <kzak@xxxxxxxxxx> http://karelzak.blogspot.com
