linux.git/fs/compat_ioctl.c, branch v5.5

Merge tag 'xfs-5.5-merge-16' of git://git.kernel.org/pub/scm/fs/xfs/xfs-linux

2019-12-02T22:46:22+00:00

Pull XFS updates from Darrick Wong:
 "For this release, we changed quite a few things.

  Highlights:

   - Fixed some long tail latency problems in the block allocator

   - Removed some long deprecated (and for the past several years no-op)
     mount options and ioctls

   - Strengthened the extended attribute and directory verifiers

   - Audited and fixed all the places where we could return EFSCORRUPTED
     without logging anything

   - Refactored the old SGI space allocation ioctls to make the
     equivalent fallocate calls

   - Fixed a race between fallocate and directio

   - Fixed an integer overflow when files have more than a few
     billion(!) extents

   - Fixed a longstanding bug where quota accounting could be incorrect
     when performing unwritten extent conversion on a freshly mounted fs

   - Fixed various complaints in scrub about soft lockups and
     unresponsiveness to signals

   - De-vtable'd the directory handling code, which should make it
     faster

   - Converted to the new mount api, for better or for worse

   - Cleaned up some memory leaks

  and quite a lot of other smaller fixes and cleanups.

  A more detailed summary:

   - Fill out the build string

   - Prevent inode fork extent count overflows

   - Refactor the allocator to reduce long tail latency

   - Rework incore log locking a little to reduce spinning

   - Break up the xfs_iomap_begin functions into smaller more cohesive
     parts

   - Fix allocation alignment being dropped too early when the
     allocation request is for more blocks than an AG is large

   - Other small cleanups

   - Clean up file buftarg retrieval helpers

   - Hoist the resvsp and unresvsp ioctls to the vfs

   - Remove the undocumented biosize mount option, since it has never
     been mentioned as existing or supported on linux

   - Clean up some of the mount option printing and parsing

   - Enhance attr leaf verifier to check block structure

   - Check dirent and attr names for invalid characters before passing
     them to the vfs

   - Refactor open-coded bmbt walking

   - Fix a few places where we return EIO instead of EFSCORRUPTED after
     failing metadata sanity checks

   - Fix a synchronization problem between fallocate and aio dio
     corrupting the file length

   - Clean up various loose ends in the iomap and bmap code

   - Convert to the new mount api

   - Make sure we always log something when returning EFSCORRUPTED

   - Fix some problems where long running scrub loops could trigger soft
     lockup warnings and/or fail to exit due to fatal signals pending

   - Fix various Coverity complaints

   - Remove most of the function pointers from the directory code to
     reduce indirection penalties

   - Ensure that dquots are attached to the inode when performing
     unwritten extent conversion after io

   - Deuglify incore projid and crtime types

   - Fix another AGI/AGF locking order deadlock when renaming

   - Clean up some quota typedefs

   - Remove the FSSETDM ioctls which haven't done anything in 20 years

   - Fix some memory leaks when mounting the log fails

   - Fix an underflow when updating an xattr leaf freemap

   - Remove some trivial wrappers

   - Report metadata corruption as an error, not a (potentially) fatal
     assertion

   - Clean up the dir/attr buffer mapping code

   - Allow fatal signals to kill scrub during parent pointer checks"

* tag 'xfs-5.5-merge-16' of git://git.kernel.org/pub/scm/fs/xfs/xfs-linux: (198 commits)
  xfs: allow parent directory scans to be interrupted with fatal signals
  xfs: remove the mappedbno argument to xfs_da_get_buf
  xfs: remove the mappedbno argument to xfs_da_read_buf
  xfs: split xfs_da3_node_read
  xfs: remove the mappedbno argument to xfs_dir3_leafn_read
  xfs: remove the mappedbno argument to xfs_dir3_leaf_read
  xfs: remove the mappedbno argument to xfs_attr3_leaf_read
  xfs: remove the mappedbno argument to xfs_da_reada_buf
  xfs: improve the xfs_dabuf_map calling conventions
  xfs: refactor xfs_dabuf_map
  xfs: simplify mappedbno handling in xfs_da_{get,read}_buf
  xfs: report corruption only as a regular error
  xfs: Remove kmem_zone_free() wrapper
  xfs: Remove kmem_zone_destroy() wrapper
  xfs: Remove slab init wrappers
  xfs: fix attr leaf header freemap.size underflow
  xfs: fix some memory leaks in log recovery
  xfs: fix another missing include
  xfs: remove XFS_IOC_FSSETDM and XFS_IOC_FSSETDM_BY_HANDLE
  xfs: remove duplicated include from xfs_dir2_data.c
  ...

fs: add generic UNRESVSP and ZERO_RANGE ioctl handlers

2019-10-28T15:37:55+00:00

These use the same scheme as the pre-existing mapping of the XFS
RESVP ioctls to ->falloc, so just extend it and remove the XFS
implementation.

Signed-off-by: Christoph Hellwig 
[darrick: fix compile error on s390]
Reviewed-by: Darrick J. Wong 
Signed-off-by: Darrick J. Wong

compat_ioctl: move SG_GET_REQUEST_TABLE handling

2019-10-23T15:23:47+00:00

SG_GET_REQUEST_TABLE is now the last ioctl command that needs a conversion
handler. This is only used in a single file, so the implementation should
be there.

I'm trying to simplify it in the process, to get rid of
the compat_alloc_user_space() and extra copy, by adding a
put_compat_request_table() function instead, which copies the data in
the right format to user space.

Cc: linux-scsi@vger.kernel.org
Cc: Doug Gilbert 
Cc: "James E.J. Bottomley" 
Cc: "Martin K. Petersen" 
Signed-off-by: Arnd Bergmann

compat_ioctl: ppp: move simple commands into ppp_generic.c

2019-10-23T15:23:47+00:00

All ppp commands that are not already handled in ppp_compat_ioctl()
are compatible, so they can now handled by calling the native
ppp_ioctl() directly.

Without CONFIG_BLOCK, the generic compat_ioctl table is now empty,
so add a check to avoid a build failure in the looking function for
that configuration.

Cc: netdev@vger.kernel.org
Cc: linux-ppp@vger.kernel.org
Cc: Paul Mackerras 
Cc: "David S. Miller" 
Signed-off-by: Arnd Bergmann

compat_ioctl: handle PPPIOCGIDLE for 64-bit time_t

2019-10-23T15:23:47+00:00

The ppp_idle structure is defined in terms of __kernel_time_t, which is
defined as 'long' on all architectures, and this usage is not affected
by the y2038 problem since it transports a time interval rather than an
absolute time.

However, the ppp user space defines the same structure as time_t, which
may be 64-bit wide on new libc versions even on 32-bit architectures.

It's easy enough to just handle both possible structure layouts on
all architectures, to deal with the possibility that a user space ppp
implementation comes with its own ppp_idle structure definition, as well
as to document the fact that the driver is y2038-safe.

Doing this also avoids the need for a special compat mode translation,
since 32-bit and 64-bit kernels now support the same interfaces.  The old
32-bit structure is also available on native 64-bit architectures now,
but this is harmless.

Cc: netdev@vger.kernel.org
Cc: linux-ppp@vger.kernel.org
Cc: Paul Mackerras 
Cc: "David S. Miller" 
Signed-off-by: Arnd Bergmann

compat_ioctl: move PPPIOCSCOMPRESS to ppp_generic

2019-10-23T15:23:47+00:00

Rather than using a compat_alloc_user_space() buffer, moving
this next to the native handler allows sharing most of
the code, leaving only the user copy portion distinct.

Signed-off-by: Al Viro 
Cc: netdev@vger.kernel.org
Cc: linux-ppp@vger.kernel.org
Cc: Paul Mackerras 
Cc: "David S. Miller" 
Signed-off-by: Arnd Bergmann

compat_ioctl: unify copy-in of ppp filters

2019-10-23T15:23:47+00:00

Now that isdn4linux is gone, the is only one implementation of PPPIOCSPASS
and PPPIOCSACTIVE in ppp_generic.c, so this is where the compat_ioctl
support should be implemented.

The two commands are implemented in very similar ways, so introduce
new helpers to allow sharing between the two and between native and
compat mode.

Signed-off-by: Al Viro 
[arnd: rebased, and added changelog text]
Cc: netdev@vger.kernel.org
Cc: linux-ppp@vger.kernel.org
Cc: Paul Mackerras 
Cc: "David S. Miller" 
Signed-off-by: Arnd Bergmann

compat_ioctl: move SIOCOUTQ out of compat_ioctl.c

2019-10-23T15:23:46+00:00

All users of this call are in socket or tty code, so handling
it there means we can avoid the table entry in fs/compat_ioctl.c.

Reviewed-by: Greg Kroah-Hartman 
Cc: Eric Dumazet 
Cc: netdev@vger.kernel.org
Cc: "David S. Miller" 
Signed-off-by: Arnd Bergmann

compat_ioctl: reimplement SG_IO handling

2019-10-23T15:23:46+00:00

There are two code locations that implement the SG_IO ioctl: the old
sg.c driver, and the generic scsi_ioctl helper that is in turn used by
multiple drivers.

To eradicate the old compat_ioctl conversion handler for the SG_IO
command, I implement a readable pair of put_sg_io_hdr() /get_sg_io_hdr()
helper functions that can be used for both compat and native mode,
and then I call this from both drivers.

For the iovec handling, there is already a compat_import_iovec() function
that can simply be called in place of import_iovec().

To avoid having to pass the compat/native state through multiple
indirections, I mark the SG_IO command itself as compatible in
fs/compat_ioctl.c and use in_compat_syscall() to figure out where
we are called from.

As a side-effect of this, the sg.c driver now also accepts the 32-bit
sg_io_hdr format in compat mode using the read/write interface, not
just ioctl. This should improve compatiblity with old 32-bit binaries,
but it would break if any application intentionally passes the 64-bit
data structure in compat mode here.

Steffen Maier helped debug an issue in an earlier version of this patch.

Cc: Steffen Maier 
Cc: linux-scsi@vger.kernel.org
Cc: Doug Gilbert 
Cc: "James E.J. Bottomley" 
Cc: "Martin K. Petersen" 
Signed-off-by: Arnd Bergmann

compat_ioctl: move WDIOC handling into wdt drivers

2019-10-23T15:23:46+00:00

All watchdog drivers implement the same set of ioctl commands, and
fortunately all of them are compatible between 32-bit and 64-bit
architectures.

Modern drivers always go through drivers/watchdog/wdt.c as an abstraction
layer, but older ones implement their own file_operations on a character
device for this.

Move the handling from fs/compat_ioctl.c into the individual drivers.

Note that most of the legacy drivers will never be used on 64-bit
hardware, because they are for an old 32-bit SoC implementation, but
doing them all at once is safer than trying to guess which ones do
or do not need the compat_ioctl handling.

Reviewed-by: Guenter Roeck 
Signed-off-by: Arnd Bergmann