On Fri, Feb 05, 2021 at 01:23:13PM -0800, Andrew Morton wrote: > On Fri, 5 Feb 2021 14:55:43 -0600 Seth Forshee <seth.forshee@xxxxxxxxxxxxx> wrote: > > > On Fri, Feb 05, 2021 at 12:41:57PM -0800, Andrew Morton wrote: > > > On Fri, 5 Feb 2021 14:21:59 -0600 Seth Forshee <seth.forshee@xxxxxxxxxxxxx> wrote: > > > > > > > Currently there seems to be an assumption in tmpfs that 64-bit > > > > architectures also have a 64-bit ino_t. This is not true; s390 at > > > > least has a 32-bit ino_t. With CONFIG_TMPFS_INODE64=y tmpfs > > > > mounts will get 64-bit inode numbers and display "inode64" in the > > > > mount options, but passing the "inode64" mount option will fail. > > > > This leads to the following behavior: > > > > > > > > # mkdir mnt > > > > # mount -t tmpfs nodev mnt > > > > # mount -o remount,rw mnt > > > > mount: /home/ubuntu/mnt: mount point not mounted or bad option. > > > > > > > > As mount sees "inode64" in the mount options and thus passes it > > > > in the options for the remount. > > > > > > > > Ideally CONFIG_TMPFS_INODE64 would depend on sizeof(ino_t) < 8, > > > > but I don't think it's possible to test for this (potentially > > > > CONFIG_ARCH_HAS_64BIT_INO_T or similar could be added, but I'm > > > > not sure whether or not that is wanted). So fix this by simply > > > > refusing to honor the CONFIG_TMPFS_INODE64 setting when > > > > sizeof(ino_t) < 8. > > > > > > How about changing s390 Kconfig so that CONFIG_TMPFS_INODE64 is not enabled? > > > > I did do that for our config. I see the s390 defconfig has it enabled, > > so I will send a patch for that too. But the fact that it can be > > configured that way and that the code behaves badly still seems > > problematic. > > I meant > > --- a/fs/Kconfig~a > +++ a/fs/Kconfig > @@ -203,7 +203,7 @@ config TMPFS_XATTR > > config TMPFS_INODE64 > bool "Use 64-bit ino_t by default in tmpfs" > - depends on TMPFS && 64BIT > + depends on TMPFS && 64BIT && !S390 > default n > help > tmpfs has historically used only inode numbers as wide as an unsigned > _ Ah. s390 does appear to be the only architecture that doesn't use unsigned long for ino_t, so that seems good to me. I can send a patch if you want, but seems like you've already got one ... Seth
