Do you have any problems with the stuff in
git://git.kernel.org/pub/scm/linux/kernel/git/viro/vfs.git #uaccess.hpsa ?
Background: I would rather get rid of compat_alloc_user_space() -
it's almost always easier to do without it these days. The only reason
it used to be useful had been doing compat ioctl handling far away
from where the native ones are dealt with (i.e. in late unlamented
fs/compat_ioctl.c).
In case of hpsa, it's definitely less headache to just lift the
copyin/copyout of {BIG_,}IOCTL_Command_struct into the callers and
have compat passthru handlers call the same functions. Without
going through "build a native struct on kernel stack, allocate on
user stack, copy there, only to have hpsa_ioctl() copy it back into
another instance on kernel stack" song and dance.
It's v5.7-rc1-based, with 4 commits in it:
hpsa passthrough: lift {BIG_,}IOCTL_Command_struct copy{in,out} into hpsa_ioctl()
hpsa: don't bother with vmalloc for BIG_IOCTL_Command_struct
hpsa: get rid of compat_alloc_user_space()
hpsa_ioctl(): tidy up a bit
drivers/scsi/hpsa.c | 199 ++++++++++++++++++++++++----------------------------
1 file changed, 90 insertions(+), 109 deletions(-)
Could you give a look to what's in that branch and comment on it?
