platform/kernel/linux-rpi3.git
6 years agoxfs: add scrub cross-referencing helpers for the inode btrees
Darrick J. Wong [Wed, 17 Jan 2018 02:52:12 +0000 (18:52 -0800)]
xfs: add scrub cross-referencing helpers for the inode btrees

Add a couple of functions to the inode btrees that will be used
to cross-reference metadata against the inobt.

Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com>
Reviewed-by: Dave Chinner <dchinner@redhat.com>
6 years agoxfs: add scrub cross-referencing helpers for the free space btrees
Darrick J. Wong [Wed, 17 Jan 2018 02:52:12 +0000 (18:52 -0800)]
xfs: add scrub cross-referencing helpers for the free space btrees

Add a couple of functions to the free space btrees that will be used
to cross-reference metadata against the bnobt/cntbt, and a generic
btree function that provides the real implementation.

Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com>
Reviewed-by: Dave Chinner <dchinner@redhat.com>
6 years agoxfs: cancel tx on xfs_defer_finish() error during xattr set/remove
Brian Foster [Tue, 16 Jan 2018 22:53:28 +0000 (14:53 -0800)]
xfs: cancel tx on xfs_defer_finish() error during xattr set/remove

Chris Dunlop reports a problem where an xattr operation fails,
reports the following error to syslog and hangs during unmount:

 ================================================
 [ BUG: lock held when returning to user space! ]
 ...
 ------------------------------------------------
 <PID> is leaving the kernel with locks still held!
 1 lock held by <PID>:
  #0:  (sb_internal){......}, at: [<ffffffffa07692a3>] xfs_trans_alloc+0xe3/0x130 [xfs]

The failure/shutdown occurs during deferred ops processing which
leads to an error return from xfs_defer_finish() via
xfs_attr_leaf_addname(). While the root cause of the failure is
unknown corruption, the cause of the subsequent BUG above and
unmount hang is failure to cancel the transaction before returning
to userspace.

The transaction is not cancelled because the out_defer_cancel error
handling paths in the xfs_attr_[leaf|node]_[add|remove]name()
functions clear args.trans without releasing the transaction. The
callers therefore lose the reference to the transaction and fail to
cancel it.

Since xfs_attr_[set|remove]() always cancel args.trans when != NULL
and xfs_defer_finish()->...->xfs_trans_roll() should always return
with a valid transaction, update the leaf/node xattr functions to
not reset args.trans in the error path responsible for cancelling
deferred ops.

Reported-by: Chris Dunlop <chris@onthe.net.au>
Signed-off-by: Brian Foster <bfoster@redhat.com>
Reviewed-by: Darrick J. Wong <darrick.wong@oracle.com>
Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com>
6 years agoxfs: account finobt blocks properly in perag reservation
Brian Foster [Fri, 12 Jan 2018 22:07:21 +0000 (14:07 -0800)]
xfs: account finobt blocks properly in perag reservation

XFS started using the perag metadata reservation pool for free inode
btree blocks in commit 76d771b4cbe33 ("xfs: use per-AG reservations
for the finobt"). To handle backwards compatibility, finobt blocks
are accounted against the pool so long as the full reservation is
available at mount time. Otherwise the ->m_inotbt_nores flag is set
and the filesystem falls back to the traditional per-transaction
finobt reservation.

This commit has two problems:

- finobt blocks are always accounted against the metadata
  reservation on allocation, regardless of ->m_inotbt_nores state
- finobt blocks are never returned to the reservation pool on free

The first problem affects reflink+finobt filesystems where the full
finobt reservation is not available at mount time. finobt blocks are
essentially stolen from the reflink reservation, putting refcountbt
management at risk of allocation failure. The second problem is an
unconditional leak of metadata reservation whenever finobt is
enabled.

Update the finobt block allocation callouts to consider
->m_inotbt_nores and account blocks appropriately. Blocks should be
consistently accounted against the metadata pool when
->m_inotbt_nores is false and otherwise tagged as RESV_NONE.

Signed-off-by: Brian Foster <bfoster@redhat.com>
Reviewed-by: Darrick J. Wong <darrick.wong@oracle.com>
Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com>
6 years agoxfs: fix check on struct_version for versions 4 or greater
Colin Ian King [Fri, 12 Jan 2018 16:47:50 +0000 (08:47 -0800)]
xfs: fix check on struct_version for versions 4 or greater

It appears that the check for versions 4 or more is incorrect and is
off-by-one. Fix this.

Detected by CoverityScan, CID#1463775 ("Logically dead code")

Fixes: ac503a4cc9e8 ("xfs: refactor the geometry structure filling function")
Signed-off-by: Colin Ian King <colin.king@canonical.com>
Reviewed-by: Darrick J. Wong <darrick.wong@oracle.com>
Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com>
6 years agoxfs: destroy mutex pag_ici_reclaim_lock before free
Xiongwei Song [Thu, 11 Jan 2018 17:45:51 +0000 (09:45 -0800)]
xfs: destroy mutex pag_ici_reclaim_lock before free

The mutex pag_ici_reclaim_lock of xfs_perag_t structure is initialized in
xfs_initialize_perag. If happen errors in xfs_initialize_perag, or free
resources in xfs_free_perag, wo need to destroy the mutex before free
perag.

Signed-off-by: Xiongwei Song <sxwjean@me.com>
Reviewed-by: Brian Foster <bfoster@redhat.com>
Reviewed-by: Darrick J. Wong <darrick.wong@oracle.com>
Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com>
6 years agoxfs: use %px for data pointers when debugging
Darrick J. Wong [Tue, 9 Jan 2018 20:02:55 +0000 (12:02 -0800)]
xfs: use %px for data pointers when debugging

Starting with commit 57e734423ad ("vsprintf: refactor %pK code out of
pointer"), the behavior of the raw '%p' printk format specifier was
changed to print a 32-bit hash of the pointer value to avoid leaking
kernel pointers into dmesg.  For most situations that's good.

This is /undesirable/ behavior when we're trying to debug XFS, however,
so define a PTR_FMT that prints the actual pointer when we're in debug
mode.

Note that %p for tracepoints still prints the raw pointer, so in the
long run we could consider rewriting some of these messages as
tracepoints.

Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com>
Reviewed-by: Dave Chinner <dchinner@redhat.com>
6 years agoxfs: use %pS printk format for direct instruction addresses
Darrick J. Wong [Tue, 9 Jan 2018 19:46:05 +0000 (11:46 -0800)]
xfs: use %pS printk format for direct instruction addresses

Use the %pS instead of the %pF printk format specifier for printing
symbols from direct addresses. This is needed for the ia64, ppc64 and
parisc64 architectures.

While we're at it, be consistent with the capitalization of the 'S'.

Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com>
Reviewed-by: Dave Chinner <dchinner@redhat.com>
6 years agoxfs: change 0x%p -> %p in print messages
Darrick J. Wong [Tue, 9 Jan 2018 19:43:36 +0000 (11:43 -0800)]
xfs: change 0x%p -> %p in print messages

Since %p prepends "0x" to the outputted string, we can drop the prefix.

Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com>
Reviewed-by: Dave Chinner <dchinner@redhat.com>
6 years agoxfs: clarify units in the failed metadata io message
Darrick J. Wong [Mon, 8 Jan 2018 19:39:18 +0000 (11:39 -0800)]
xfs: clarify units in the failed metadata io message

If a metadata IO error happens, we report the location of the failed IO
request in units of daddrs.  However, the printk message misleads people
into thinking that the units are fs blocks, so fix the reported units.

Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com>
Reviewed-by: Dave Chinner <dchinner@redhat.com>
6 years agoxfs: harden directory integrity checks some more
Darrick J. Wong [Tue, 9 Jan 2018 19:11:42 +0000 (11:11 -0800)]
xfs: harden directory integrity checks some more

If a malicious filesystem image contains a block+ format directory
wherein the directory inode's core.mode is set such that
S_ISDIR(core.mode) == 0, and if there are subdirectories of the
corrupted directory, an attempt to traverse up the directory tree will
crash the kernel in __xfs_dir3_data_check.  Running the online scrub's
parent checks will tend to do this.

The crash occurs because the directory inode's d_ops get set to
xfs_dir[23]_nondir_ops (it's not a directory) but the parent pointer
scrubber's indiscriminate call to xfs_readdir proceeds past the ASSERT
if we have non fatal asserts configured.

Fix the null pointer dereference crash in __xfs_dir3_data_check by
looking for S_ISDIR or wrong d_ops; and teach the parent scrubber
to bail out if it is fed a non-directory "parent".

Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com>
Reviewed-by: Brian Foster <bfoster@redhat.com>
6 years agoxfs: refactor the geometry structure filling function
Darrick J. Wong [Mon, 8 Jan 2018 18:51:27 +0000 (10:51 -0800)]
xfs: refactor the geometry structure filling function

Refactor the geometry structure filling function to use the superblock
to fill the fields.  While we're at it, make the function less indenty
and use some whitespace to make the function easier to read.

Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com>
Reviewed-by: Brian Foster <bfoster@redhat.com>
6 years agoxfs: hoist xfs_fs_geometry to libxfs
Darrick J. Wong [Mon, 8 Jan 2018 18:51:27 +0000 (10:51 -0800)]
xfs: hoist xfs_fs_geometry to libxfs

Move xfs_fs_geometry to libxfs so that we can clean up the fs geometry
reporting in xfsprogs.

Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com>
Reviewed-by: Brian Foster <bfoster@redhat.com>
6 years agoxfs: trace log reservations at mount time
Darrick J. Wong [Mon, 8 Jan 2018 18:51:26 +0000 (10:51 -0800)]
xfs: trace log reservations at mount time

At each mount, emit the transaction reservation type information via
tracepoints.  This makes it easier to compare the log reservation info
calculated by the kernel and xfsprogs so that we can more easily diagnose
minimum log size failures on freshly formatted filesystems.

Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com>
Reviewed-by: Brian Foster <bfoster@redhat.com>
6 years agoxfs: dump the first 128 bytes of any corrupt buffer
Darrick J. Wong [Mon, 8 Jan 2018 18:51:26 +0000 (10:51 -0800)]
xfs: dump the first 128 bytes of any corrupt buffer

Increase the corrupt buffer dump to the first 128 bytes since v5
filesystems have larger block headers than before.

Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com>
Reviewed-by: Dave Chinner <dchinner@redhat.com>
6 years agoxfs: teach error reporting functions to take xfs_failaddr_t
Darrick J. Wong [Mon, 8 Jan 2018 18:51:25 +0000 (10:51 -0800)]
xfs: teach error reporting functions to take xfs_failaddr_t

Convert the two other error reporting functions to take xfs_failaddr_t
when the caller wishes to capture a code pointer instead of the classic
void * pointer.

Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com>
Reviewed-by: Dave Chinner <dchinner@redhat.com>
6 years agoxfs: standardize quota verification function outputs
Darrick J. Wong [Mon, 8 Jan 2018 18:51:25 +0000 (10:51 -0800)]
xfs: standardize quota verification function outputs

Rename xfs_dqcheck to xfs_dquot_verify and make it return an
xfs_failaddr_t like every other structure verifier function.
This enables us to check on-disk quotas in the same way that we check
everything else.  Callers are now responsible for logging errors, as
XFS_QMOPT_DOWARN goes away.

Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com>
Reviewed-by: Dave Chinner <dchinner@redhat.com>
6 years agoxfs: separate dquot repair into a separate function
Darrick J. Wong [Mon, 8 Jan 2018 18:51:24 +0000 (10:51 -0800)]
xfs: separate dquot repair into a separate function

Move the dquot repair code into a separate function and remove
XFS_QMOPT_DQREPAIR in favor of calling the helper directly.  Remove
other dead code because quotacheck is the only caller of DQREPAIR.

Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com>
Reviewed-by: Dave Chinner <dchinner@redhat.com>
6 years agoxfs: create a new buf_ops pointer to verify structure metadata
Darrick J. Wong [Mon, 8 Jan 2018 18:51:08 +0000 (10:51 -0800)]
xfs: create a new buf_ops pointer to verify structure metadata

Expose all metadata structure buffer verifier functions via buf_ops.
These will be used by the online scrub mechanism to look for problems
with buffers that are already sitting around in memory.

Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com>
Reviewed-by: Dave Chinner <dchinner@redhat.com>
6 years agoxfs: fail out of xfs_attr3_leaf_lookup_int if it looks corrupt
Darrick J. Wong [Mon, 8 Jan 2018 18:51:07 +0000 (10:51 -0800)]
xfs: fail out of xfs_attr3_leaf_lookup_int if it looks corrupt

If the xattr leaf block looks corrupt, return -EFSCORRUPTED to userspace
instead of ASSERTing on debug kernels or running off the end of the
buffer on regular kernels.

Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com>
Reviewed-by: Dave Chinner <dchinner@redhat.com>
6 years agoxfs: provide a centralized method for verifying inline fork data
Darrick J. Wong [Mon, 8 Jan 2018 18:51:06 +0000 (10:51 -0800)]
xfs: provide a centralized method for verifying inline fork data

Replace the current haphazard dir2 shortform verifier callsites with a
centralized verifier function that can be called either with the default
verifier functions or with a custom set.  This helps us strengthen
integrity checking while providing us with flexibility for repair tools.

xfs_repair wants this to be able to supply its own verifier functions
when trying to fix possibly corrupt metadata.

Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com>
Reviewed-by: Dave Chinner <dchinner@redhat.com>
6 years agoxfs: refactor short form directory structure verifier function
Darrick J. Wong [Mon, 8 Jan 2018 18:51:06 +0000 (10:51 -0800)]
xfs: refactor short form directory structure verifier function

Change the short form directory structure verifier function to return
the instruction pointer of a failing check or NULL if everything's ok.

Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com>
Reviewed-by: Dave Chinner <dchinner@redhat.com>
6 years agoxfs: create structure verifier function for short form symlinks
Darrick J. Wong [Mon, 8 Jan 2018 18:51:05 +0000 (10:51 -0800)]
xfs: create structure verifier function for short form symlinks

Create a function to check the structure of short form symlink targets.

Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com>
Reviewed-by: Dave Chinner <dchinner@redhat.com>
6 years agoxfs: create structure verifier function for shortform xattrs
Darrick J. Wong [Mon, 8 Jan 2018 18:51:05 +0000 (10:51 -0800)]
xfs: create structure verifier function for shortform xattrs

Create a function to perform structure verification for short form
extended attributes.

Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com>
Reviewed-by: Dave Chinner <dchinner@redhat.com>
6 years agoxfs: move inode fork verifiers to xfs_dinode_verify
Darrick J. Wong [Mon, 8 Jan 2018 18:51:04 +0000 (10:51 -0800)]
xfs: move inode fork verifiers to xfs_dinode_verify

Consolidate the fork size and format verifiers to xfs_dinode_verify so
that we can reject bad inodes earlier and in a single place.

Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com>
Reviewed-by: Dave Chinner <dchinner@redhat.com>
6 years agoxfs: verify dinode header first
Darrick J. Wong [Mon, 8 Jan 2018 18:51:04 +0000 (10:51 -0800)]
xfs: verify dinode header first

Move the v3 inode integrity information (crc, owner, metauuid) before we
look at anything else in the inode so that we don't waste time on a torn
write or a totally garbled block.  This makes xfs_dinode_verify more
consistent with the other verifiers.

Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com>
Reviewed-by: Dave Chinner <dchinner@redhat.com>
6 years agoxfs: refactor verifier callers to print address of failing check
Darrick J. Wong [Mon, 8 Jan 2018 18:51:03 +0000 (10:51 -0800)]
xfs: refactor verifier callers to print address of failing check

Refactor the callers of verifiers to print the instruction address of a
failing check.

Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com>
Reviewed-by: Dave Chinner <dchinner@redhat.com>
6 years agoxfs: have buffer verifier functions report failing address
Darrick J. Wong [Mon, 8 Jan 2018 18:51:03 +0000 (10:51 -0800)]
xfs: have buffer verifier functions report failing address

Modify each function that checks the contents of a metadata buffer to
return the instruction address of the failing test so that we can report
more precise failure errors to the log.

Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com>
Reviewed-by: Dave Chinner <dchinner@redhat.com>
6 years agoxfs: refactor xfs_verifier_error and xfs_buf_ioerror
Darrick J. Wong [Mon, 8 Jan 2018 18:51:02 +0000 (10:51 -0800)]
xfs: refactor xfs_verifier_error and xfs_buf_ioerror

Since all verification errors also mark the buffer as having an error,
we can combine these two calls.  Later we'll add a xfs_failaddr_t
parameter to promote the idea of reporting corruption errors and the
address of the failing check to enable better debugging reports.

Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com>
Reviewed-by: Dave Chinner <dchinner@redhat.com>
6 years agoxfs: remove XFS_WANT_CORRUPTED_RETURN from dir3 data verifiers
Darrick J. Wong [Mon, 8 Jan 2018 18:51:01 +0000 (10:51 -0800)]
xfs: remove XFS_WANT_CORRUPTED_RETURN from dir3 data verifiers

Since __xfs_dir3_data_check verifies on-disk metadata, we can't have it
noisily blowing asserts and hanging the system on corrupt data coming in
off the disk.  Instead, have it return a boolean like all the other
checker functions, and only have it noisily fail if we fail in debug
mode.

Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com>
Reviewed-by: Dave Chinner <dchinner@redhat.com>
6 years agoxfs: refactor short form btree pointer verification
Darrick J. Wong [Mon, 8 Jan 2018 18:51:01 +0000 (10:51 -0800)]
xfs: refactor short form btree pointer verification

Now that we have xfs_verify_agbno, use it to verify short form btree
pointers instead of open-coding them.

Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com>
Reviewed-by: Dave Chinner <dchinner@redhat.com>
6 years agoxfs: refactor long-format btree header verification routines
Darrick J. Wong [Mon, 8 Jan 2018 18:51:00 +0000 (10:51 -0800)]
xfs: refactor long-format btree header verification routines

Create two helper functions to verify the headers of a long format
btree block.  We'll use this later for the realtime rmapbt.

Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com>
Reviewed-by: Dave Chinner <dchinner@redhat.com>
6 years agoxfs: remove XFS_FSB_SANITY_CHECK
Darrick J. Wong [Mon, 8 Jan 2018 18:51:00 +0000 (10:51 -0800)]
xfs: remove XFS_FSB_SANITY_CHECK

We already have a function to verify fsb pointers, so get rid of the
last users of the (less robust) macro.

Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com>
Reviewed-by: Dave Chinner <dchinner@redhat.com>
6 years agoxfs: distinguish between corrupt inode and invalid inum in xfs_scrub_get_inode
Darrick J. Wong [Mon, 8 Jan 2018 18:49:04 +0000 (10:49 -0800)]
xfs: distinguish between corrupt inode and invalid inum in xfs_scrub_get_inode

In xfs_scrub_get_inode, we don't do a good enough job distinguishing
EINVAL returns from xfs_iget w/ IGET_UNTRUSTED -- this can happen if the
passed in inode number is invalid (past eofs, inobt says it isn't an
inode) or if the inum is actually valid but the inode buffer fails
verifier.  In the first case we still want to return ENOENT, but in the
second case we want to capture the corruption error.

Therefore, if xfs_iget returns EINVAL, try the raw imap lookup.  If that
succeeds, we conclude it's a corruption error, otherwise we just bounce
out to userspace.

Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com>
Reviewed-by: Dave Chinner <dchinner@redhat.com>
6 years agoxfs: always grab transaction when scrubbing inode
Darrick J. Wong [Mon, 8 Jan 2018 18:49:03 +0000 (10:49 -0800)]
xfs: always grab transaction when scrubbing inode

Always allocate a transaction for inode scrubbing, even if the _iget
fails.  This is something that is nice to have now for consistency with
the other scrubbers but will become critical when we get to online
repair where we'll actually use the transaction + raw buffer read to fix
the verifier errors.

Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com>
Reviewed-by: Dave Chinner <dchinner@redhat.com>
6 years agoxfs: xfs_scrub_bmap should use for_each_xfs_iext
Darrick J. Wong [Mon, 8 Jan 2018 18:49:03 +0000 (10:49 -0800)]
xfs: xfs_scrub_bmap should use for_each_xfs_iext

Refactor xfs_scrub_bmap to use for_each_xfs_iext now that it exists.

Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com>
Reviewed-by: Dave Chinner <dchinner@redhat.com>
6 years agoxfs: catch a few more error codes when scrubbing secondary sb
Darrick J. Wong [Mon, 8 Jan 2018 18:49:02 +0000 (10:49 -0800)]
xfs: catch a few more error codes when scrubbing secondary sb

The superblock validation routines return a variety of error codes to
reject a mount request.  For scrub we can assume that the mount
succeeded, so if we see these things appear when scrubbing secondary sb
X, we can treat them all like corruption.

Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com>
Reviewed-by: Dave Chinner <dchinner@redhat.com>
6 years agoxfs: ignore agfl read errors when not scrubbing agfl
Darrick J. Wong [Mon, 8 Jan 2018 18:49:02 +0000 (10:49 -0800)]
xfs: ignore agfl read errors when not scrubbing agfl

In xfs_scrub_ag_read_headers, if we're not scrubbing the AGFL but
hit a read error reading the AGFL, we should reset the error code
so that it doesn't propagate up into the caller.

Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com>
Reviewed-by: Dave Chinner <dchinner@redhat.com>
6 years agoiomap: report collisions between directio and buffered writes to userspace
Darrick J. Wong [Mon, 8 Jan 2018 18:41:39 +0000 (10:41 -0800)]
iomap: report collisions between directio and buffered writes to userspace

If two programs simultaneously try to write to the same part of a file
via direct IO and buffered IO, there's a chance that the post-diowrite
pagecache invalidation will fail on the dirty page.  When this happens,
the dio write succeeded, which means that the page cache is no longer
coherent with the disk!

Programs are not supposed to mix IO types and this is a clear case of
data corruption, so store an EIO which will be reflected to userspace
during the next fsync.  Replace the WARN_ON with a ratelimited pr_crit
so that the developers have /some/ kind of breadcrumb to track down the
offending program(s) and file(s) involved.

Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com>
Reviewed-by: Liu Bo <bo.li.liu@oracle.com>
6 years agoxfs: eliminate duplicate icreate tx reservation functions
Brian Foster [Mon, 8 Jan 2018 18:41:38 +0000 (10:41 -0800)]
xfs: eliminate duplicate icreate tx reservation functions

The create transaction reservation calculation has two different
branches of code depending on whether the filesystem is a v5 format
fs or older. Each branch considers the max reservation between the
allocation case (new chunk allocation + record insert) and the
modify case (chunk exists, record modification) of inode allocation.

The modify case is the same for both superblock versions with the
exception of the finobt. The finobt helper checks the feature bit,
however, and so the modify case already shares the same code.

Now that inode chunk allocation has been refactored into a helper
that checks the superblock version to calculate the appropriate
reservation for the create transaction, the only remaining
difference between the create and icreate branches is the call to
the finobt helper. As noted above, the finobt helper is a no-op when
the feature is not enabled. Therefore, these branches are
effectively duplicate and can be condensed.

Remove the xfs_calc_create_*() branch of functions and update the
various callers to use the xfs_calc_icreate_*() variant. The latter
creates the same reservation size for v4 create transactions as the
removed branch. As such, this patch does not result in transaction
reservation changes.

Signed-off-by: Brian Foster <bfoster@redhat.com>
Reviewed-by: Dave Chinner <dchinner@redhat.com>
Reviewed-by: Darrick J. Wong <darrick.wong@oracle.com>
Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com>
6 years agoxfs: refactor inode chunk alloc/free tx reservation
Brian Foster [Mon, 8 Jan 2018 18:41:38 +0000 (10:41 -0800)]
xfs: refactor inode chunk alloc/free tx reservation

The reservation for the various forms of inode allocation is
scattered across several different functions. This includes two
variants of chunk allocation (v5 icreate transactions vs. older
create transactions) and the inode free transaction.

To clean up some of this code and clarify the purpose of specific
allocfree reservations, continue the pattern of defining helper
functions for smaller operational units of broader transactions.
Refactor the reservation into an inode chunk alloc/free helper that
considers the various conditions based on filesystem format.

An inode chunk free involves an extent free and buffer
invalidations. The latter requires reservation for log headers only.
An inode chunk allocation modifies the free space btrees and logs
the chunk on v4 supers. v5 supers initialize the inode chunk using
ordered buffers and so do not log the chunk.

As a side effect of this refactoring, add one more allocfree res to
the ifree transaction. Technically this does not serve a specific
purpose because inode chunks are freed via deferred operations and
thus occur after a transaction roll. tr_ifree has a bit of a history
of tx overruns caused by too many agfl fixups during sustained file
deletion workloads, so add this extra reservation as a form of
padding nonetheless.

Signed-off-by: Brian Foster <bfoster@redhat.com>
Reviewed-by: Dave Chinner <dchinner@redhat.com>
Reviewed-by: Darrick J. Wong <darrick.wong@oracle.com>
Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com>
6 years agoxfs: include an allocfree res for inobt modifications
Brian Foster [Mon, 8 Jan 2018 18:41:37 +0000 (10:41 -0800)]
xfs: include an allocfree res for inobt modifications

Analysis of recent reports of log reservation overruns and code
inspection has uncovered that the reservations associated with inode
operations may not cover the worst case scenarios. In particular,
many cases only include one allocfree res. for a particular
operation even though said operations may also entail AGFL fixups
and inode btree block allocations in addition to the actual inode
chunk allocation. This can easily turn into two or three block
allocations (or frees) per operation.

In theory, the only way to define the worst case reservation is to
include an allocfree res for each individual allocation in a
transaction. Since that is impractical (we can perform multiple agfl
fixups per tx and not every allocation results in a full tree
operation), we need to find a reasonable compromise that addresses
the deficiency in practice without blowing out the size of the
transactions.

Since the inode btrees are not filled by the AGFL, record insertion
and removal can directly result in block allocations and frees
depending on the shape of the tree. These allocations and frees
occur in the same transaction context as the inobt update itself,
but are separate from the allocation/free that might be required for
an inode chunk. Therefore, it makes sense to assume that an [f]inobt
insert/remove can directly result in one or more block allocations
on behalf of the tree.

Refactor the inode transaction reservations to include one allocfree
res. per inode btree modification to cover allocations required by
the tree itself. This separates the reservation required to allocate
the inode chunk from the reservation required for inobt record
insertion/removal. Apply the same logic to the finobt. This results
in killing off the finobt modify condition because we no longer
assume that the broader transaction reservation will cover finobt
block allocations and finobt shape changes can occur in either of
the inobt allocation or modify situations.

Suggested-by: Dave Chinner <david@fromorbit.com>
Signed-off-by: Brian Foster <bfoster@redhat.com>
Reviewed-by: Dave Chinner <dchinner@redhat.com>
Reviewed-by: Darrick J. Wong <darrick.wong@oracle.com>
Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com>
6 years agoxfs: truncate transaction does not modify the inobt
Brian Foster [Mon, 8 Jan 2018 18:41:37 +0000 (10:41 -0800)]
xfs: truncate transaction does not modify the inobt

The truncate transaction does not ever modify the inode btree, but
includes an associated log reservation. Update
xfs_calc_itruncate_reservation() to remove the reservation
associated with inobt updates.

Signed-off-by: Brian Foster <bfoster@redhat.com>
Reviewed-by: Dave Chinner <dchinner@redhat.com>
Reviewed-by: Darrick J. Wong <darrick.wong@oracle.com>
Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com>
6 years agoxfs: fix up agi unlinked list reservations
Brian Foster [Mon, 8 Jan 2018 18:41:36 +0000 (10:41 -0800)]
xfs: fix up agi unlinked list reservations

The current AGI unlinked list addition and removal reservations do
not reflect the worst case log usage. An unlinked list removal can
log up to two on-disk inode clusters but only includes reservation
for one. An unlinked list addition logs the on-disk cluster but
includes reservation for an in-core inode.

Update the AGI unlinked list reservation helpers to calculate the
correct worst case reservation for the associated operations.

Signed-off-by: Brian Foster <bfoster@redhat.com>
Reviewed-by: Dave Chinner <dchinner@redhat.com>
Reviewed-by: Darrick J. Wong <darrick.wong@oracle.com>
Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com>
6 years agoxfs: include inobt buffers in ifree tx log reservation
Brian Foster [Mon, 8 Jan 2018 18:41:36 +0000 (10:41 -0800)]
xfs: include inobt buffers in ifree tx log reservation

The tr_ifree transaction handles inode unlinks and inode chunk
frees. The current transaction calculation does not accurately
reflect worst case changes to the inode btree, however. The inobt
portion of the current transaction reservation only covers
modification of a single inobt buffer (for the particular inode
record). This is a historical artifact from the days before XFS
supported full inode chunk removal.

When support for inode chunk removal was added in commit
254f6311ed1b ("Implement deletion of inode clusters in XFS."), the
additional log reservation required for chunk removal was not added
correctly. The new reservation only considered the header overhead
of associated buffers rather than the full contents of the btrees
and AGF and AGFL buffers affected by the transaction. The
reservation for the free space btrees was subsequently fixed up in
commit 5fe6abb82f76 ("Add space for inode and allocation btrees to
ITRUNCATE log reservation"), but the res. for full inobt joins has
never been added.

Further review of the ifree reservation uncovered a couple more
problems:

- The undocumented +2 blocks are intended for the AGF and AGFL, but
  are also not sized correctly and should be logged as full sectors
  (not FSBs).
- The additional single block header is undocumented and serves no
  apparent purpose.

Update xfs_calc_ifree_reservation() to include a full inobt join in
the reservation calculation. Refactor the undocumented blocks
appropriately and fix up the comments to reflect the current
calculation.

Signed-off-by: Brian Foster <bfoster@redhat.com>
Reviewed-by: Dave Chinner <dchinner@redhat.com>
Reviewed-by: Darrick J. Wong <darrick.wong@oracle.com>
Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com>
6 years agoxfs: print transaction log reservation on overrun
Brian Foster [Mon, 8 Jan 2018 18:41:35 +0000 (10:41 -0800)]
xfs: print transaction log reservation on overrun

The transaction dump code displays the content and reservation
consumption of a particular transaction in the event of an overrun.
It currently displays the reservation associated with the
transaction ticket, but not the original reservation attached to the
transaction.

The latter value reflects the original transaction reservation
calculation before additional reservation overhead is assigned, such
as for the CIL context header and potential split region headers.

Update xlog_print_trans() to also print the original transaction
reservation in the event of overrun. This provides a reference point
to identify how much reservation overhead was added to a particular
ticket by xfs_log_calc_unit_res().

Signed-off-by: Brian Foster <bfoster@redhat.com>
Reviewed-by: Dave Chinner <dchinner@redhat.com>
Reviewed-by: Darrick J. Wong <darrick.wong@oracle.com>
Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com>
6 years agoxfs: scrub inode nsec fields
Darrick J. Wong [Mon, 8 Jan 2018 18:41:35 +0000 (10:41 -0800)]
xfs: scrub inode nsec fields

Check that the nanosecond fields in each timestamp aren't larger
than a billion.

Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com>
Reviewed-by: Dave Chinner <dchinner@redhat.com>
6 years agoxfs: move all scrub input checking to xfs_scrub_validate
Eric Sandeen [Mon, 8 Jan 2018 18:41:34 +0000 (10:41 -0800)]
xfs: move all scrub input checking to xfs_scrub_validate

There were ad-hoc checks for some scrub types but not others;
mark each scrub type with ... it's type, and use that to validate
the allowed and/or required input fields.

Moving these checks out of xfs_scrub_setup_ag_header makes it
a thin wrapper, so unwrap it in the process.

Signed-off-by: Eric Sandeen <sandeen@redhat.com>
[darrick: add xfs_ prefix to enum, check scrub args after checking type]
Reviewed-by: Darrick J. Wong <darrick.wong@oracle.com>
Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com>
6 years agoxfs: factor out scrub input checking
Eric Sandeen [Mon, 8 Jan 2018 18:41:34 +0000 (10:41 -0800)]
xfs: factor out scrub input checking

Do this before adding more core checks.

Signed-off-by: Eric Sandeen <sandeen@redhat.com>
Reviewed-by: Darrick J. Wong <darrick.wong@oracle.com>
Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com>
6 years agoxfs: explicitly initialize meta_scrub_ops array by type
Eric Sandeen [Mon, 8 Jan 2018 18:41:33 +0000 (10:41 -0800)]
xfs: explicitly initialize meta_scrub_ops array by type

An implicit mapping to type by order of initialization seems
error-prone, and doesn't lend itself to cscope-ing.

Also add sanity checks about size of array vs. max types,
and a defensive check that ->scrub exists before using it.

Signed-off-by: Eric Sandeen <sandeen@redhat.com>
Reviewed-by: Darrick J. Wong <darrick.wong@oracle.com>
Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com>
6 years agoxfs: Show realtime device stats on statfs calls if realtime flags set
Richard Wareing [Mon, 8 Jan 2018 18:41:33 +0000 (10:41 -0800)]
xfs: Show realtime device stats on statfs calls if realtime flags set

- Reports realtime device free blocks in statfs calls if (realtime)
  inheritance bit is set on the inode of directory, or realtime flag
  in the case of files.  This is a bit more intuitive, especially for
  use-cases which are using a much larger device for the realtime device.
- Add XFS_IS_REALTIME_MOUNT option to gate based on the existence of a
  realtime device on the mount, similar to the XFS_IS_REALTIME_INODE
  option.

Reviewed-by: Christoph Hellwig <hch@lst.de>
Reviewed-by: Dave Chinner <dchinner@redhat.com>
Signed-off-by: Richard Wareing <rwareing@fb.com>
Reviewed-by: Darrick J. Wong <darrick.wong@oracle.com>
Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com>
6 years agoLinux 4.15-rc7
Linus Torvalds [Sun, 7 Jan 2018 22:22:41 +0000 (14:22 -0800)]
Linux 4.15-rc7

6 years agoMerge branch 'parisc-4.15-3' of git://git.kernel.org/pub/scm/linux/kernel/git/deller...
Linus Torvalds [Sun, 7 Jan 2018 19:42:57 +0000 (11:42 -0800)]
Merge branch 'parisc-4.15-3' of git://git./linux/kernel/git/deller/parisc-linux

Pull parisc fixes from Helge Deller:

 - Many small fixes to show the real physical addresses of devices
   instead of hashed addresses.

 - One important fix to unbreak 32-bit SMP support: We forgot to 16-byte
   align the spinlocks in the assembler code.

 - Qemu support: The host will get a chance to sleep when the parisc
   guest is idle. We use the same mechanism as the power architecture by
   overlaying the "or %r10,%r10,%r10" instruction which is simply a nop
   on real hardware.

* 'parisc-4.15-3' of git://git.kernel.org/pub/scm/linux/kernel/git/deller/parisc-linux:
  parisc: qemu idle sleep support
  parisc: Fix alignment of pa_tlb_lock in assembly on 32-bit SMP kernel
  parisc: Show unhashed EISA EEPROM address
  parisc: Show unhashed HPA of Dino chip
  parisc: Show initial kernel memory layout unhashed
  parisc: Show unhashed hardware inventory

6 years agoMerge tag 'apparmor-pr-2018-01-07' of git://git.kernel.org/pub/scm/linux/kernel/git...
Linus Torvalds [Sun, 7 Jan 2018 19:33:12 +0000 (11:33 -0800)]
Merge tag 'apparmor-pr-2018-01-07' of git://git./linux/kernel/git/jj/linux-apparmor

Pull apparmor fix from John Johansen:
 "This fixes a regression when the kernel feature set is reported as
  supporting mount and policy is pinned to a feature set that does not
  support mount mediation"

* tag 'apparmor-pr-2018-01-07' of git://git.kernel.org/pub/scm/linux/kernel/git/jj/linux-apparmor:
  apparmor: fix regression in mount mediation when feature set is pinned

6 years agoMerge tag 'led_fixes_for_4.15-rc7' of git://git.kernel.org/pub/scm/linux/kernel/git...
Linus Torvalds [Sun, 7 Jan 2018 19:01:59 +0000 (11:01 -0800)]
Merge tag 'led_fixes_for_4.15-rc7' of git://git./linux/kernel/git/j.anaszewski/linux-leds

Pull LED fix from Jacek Anaszewski:
 "The commit 2b83ff96f51d for 4.15-rc6, which was fixing LED brightness
  setting after clearing delay_off broke the behavior on any alteration
  of delay_on{off} properties, due to use of a LED core helper that does
  too much for this particular case"

* tag 'led_fixes_for_4.15-rc7' of git://git.kernel.org/pub/scm/linux/kernel/git/j.anaszewski/linux-leds:
  leds: core: Fix regression caused by commit 2b83ff96f51d

6 years agoMerge tag 'for-linus-20180107' of git://git.infradead.org/linux-mtd
Linus Torvalds [Sun, 7 Jan 2018 19:00:06 +0000 (11:00 -0800)]
Merge tag 'for-linus-20180107' of git://git.infradead.org/linux-mtd

Pull MTD bugfix from Richard Weinberger:
 "A single fix for the pxa3xx NAND driver"

* tag 'for-linus-20180107' of git://git.infradead.org/linux-mtd:
  mtd: nand: pxa3xx: Fix READOOB implementation

6 years agoleds: core: Fix regression caused by commit 2b83ff96f51d
Jacek Anaszewski [Wed, 3 Jan 2018 20:13:45 +0000 (21:13 +0100)]
leds: core: Fix regression caused by commit 2b83ff96f51d

Commit 2b83ff96f51d ("led: core: Fix brightness setting when setting delay_off=0")
replaced del_timer_sync(&led_cdev->blink_timer) with led_stop_software_blink()
in led_blink_set(), which additionally clears LED_BLINK_SW flag as well as
zeroes blink_delay_on and blink_delay_off properties of the struct led_classdev.

Cleansing of the latter ones wasn't required to fix the original issue but
wasn't considered harmful. It nonetheless turned out to be so in case when
pointer to one or both props is passed to led_blink_set() like in the
ledtrig-timer.c. In such cases zeroes are passed later in delay_on and/or
delay_off arguments to led_blink_setup(), which results either in stopping
the software blinking or setting blinking frequency always to 1Hz.

Avoid using led_stop_software_blink() and add a single call required
to clear LED_BLINK_SW flag, which was the only needed modification to
fix the original issue.

Fixes 2b83ff96f51d ("led: core: Fix brightness setting when setting delay_off=0")
Signed-off-by: Jacek Anaszewski <jacek.anaszewski@gmail.com>
6 years agoMerge branch 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/viro/vfs
Linus Torvalds [Sun, 7 Jan 2018 01:13:21 +0000 (17:13 -0800)]
Merge branch 'for-linus' of git://git./linux/kernel/git/viro/vfs

Pull vfs fixes from Al Viro:

 - untangle sys_close() abuses in xt_bpf

 - deal with register_shrinker() failures in sget()

* 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/viro/vfs:
  fix "netfilter: xt_bpf: Fix XT_BPF_MODE_FD_PINNED mode of 'xt_bpf_info_v1'"
  sget(): handle failures of register_shrinker()
  mm,vmscan: Make unregister_shrinker() no-op if register_shrinker() failed.

6 years agoMerge tag 'for-linus' of git://git.kernel.org/pub/scm/virt/kvm/kvm
Linus Torvalds [Sun, 7 Jan 2018 01:05:05 +0000 (17:05 -0800)]
Merge tag 'for-linus' of git://git./virt/kvm/kvm

Pull KVM fixes from Radim Krčmář:
 "s390:
   - Two fixes for potential bitmap overruns in the cmma migration code

  x86:
   - Clear guest provided GPRs to defeat the Project Zero PoC for CVE
     2017-5715"

* tag 'for-linus' of git://git.kernel.org/pub/scm/virt/kvm/kvm:
  kvm: vmx: Scrub hardware GPRs at VM-exit
  KVM: s390: prevent buffer overrun on memory hotplug during migration
  KVM: s390: fix cmma migration for multiple memory slots

6 years agomtd: nand: pxa3xx: Fix READOOB implementation
Boris Brezillon [Mon, 18 Dec 2017 10:32:45 +0000 (11:32 +0100)]
mtd: nand: pxa3xx: Fix READOOB implementation

In the current driver, OOB bytes are accessed in raw mode, and when a
page access is done with NDCR_SPARE_EN set and NDCR_ECC_EN cleared, the
driver must read the whole spare area (64 bytes in case of a 2k page,
16 bytes for a 512 page). The driver was only reading the free OOB
bytes, which was leaving some unread data in the FIFO and was somehow
leading to a timeout.

We could patch the driver to read ->spare_size + ->ecc_size instead of
just ->spare_size when READOOB is requested, but we'd better make
in-band and OOB accesses consistent.
Since the driver is always accessing in-band data in non-raw mode (with
the ECC engine enabled), we should also access OOB data in this mode.
That's particularly useful when using the BCH engine because in this
mode the free OOB bytes are also ECC protected.

Fixes: 43bcfd2bb24a ("mtd: nand: pxa3xx: Add driver-specific ECC BCH support")
Cc: stable@vger.kernel.org
Reported-by: Sean Nyekjær <sean.nyekjaer@prevas.dk>
Tested-by: Willy Tarreau <w@1wt.eu>
Signed-off-by: Boris Brezillon <boris.brezillon@free-electrons.com>
Acked-by: Ezequiel Garcia <ezequiel@vanguardiasur.com.ar>
Tested-by: Sean Nyekjaer <sean.nyekjaer@prevas.dk>
Acked-by: Robert Jarzmik <robert.jarzmik@free.fr>
Signed-off-by: Richard Weinberger <richard@nod.at>
6 years agoMerge tag 'powerpc-4.15-6' of git://git.kernel.org/pub/scm/linux/kernel/git/powerpc...
Linus Torvalds [Sat, 6 Jan 2018 17:48:27 +0000 (09:48 -0800)]
Merge tag 'powerpc-4.15-6' of git://git./linux/kernel/git/powerpc/linux

Pull powerpc fix from Michael Ellerman:
 "Just one fix to correctly return SEGV_ACCERR when we take a SEGV on a
  mapped region. The bug was introduced in the refactoring of the page
  fault handler we did in the previous release.

  Thanks to John Sperbeck"

* tag 'powerpc-4.15-6' of git://git.kernel.org/pub/scm/linux/kernel/git/powerpc/linux:
  powerpc/mm: Fix SEGV on mapped region to return SEGV_ACCERR

6 years agoMerge tag 'kvm-s390-master-4.15-2' of git://git.kernel.org/pub/scm/linux/kernel/git...
Radim Krčmář [Sat, 6 Jan 2018 16:26:37 +0000 (17:26 +0100)]
Merge tag 'kvm-s390-master-4.15-2' of git://git./linux/kernel/git/kvms390/linux

KVM: s390: fixes for cmma migration

Two fixes for potential bitmap overruns in the cmma migration
code.

6 years agoparisc: qemu idle sleep support
Helge Deller [Fri, 5 Jan 2018 20:55:38 +0000 (21:55 +0100)]
parisc: qemu idle sleep support

Add qemu idle sleep support when running under qemu with SeaBIOS PDC
firmware.

Like the power architecture we use the "or" assembler instructions,
which translate to nops on real hardware, to indicate that qemu shall
idle sleep.

Signed-off-by: Helge Deller <deller@gmx.de>
Cc: Richard Henderson <rth@twiddle.net>
CC: stable@vger.kernel.org # v4.9+
6 years agoMerge branch 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/dtor/input
Linus Torvalds [Sat, 6 Jan 2018 00:45:06 +0000 (16:45 -0800)]
Merge branch 'for-linus' of git://git./linux/kernel/git/dtor/input

Pull input fixes from Dmitry Torokhov:
 "Just a few driver fixups, nothing exciting"

* 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/dtor/input:
  Input: xen-kbdfront - do not advertise multi-touch pressure support
  Input: hideep - fix compile error due to missing include file
  Input: elants_i2c - do not clobber interrupt trigger on x86
  Input: joystick/analog - riscv has get_cycles()
  Input: elantech - add new icbody type 15
  Input: ims-pcu - fix typo in the error message

6 years agoMerge tag 'iommu-v4.15-rc7' of git://github.com/awilliam/linux-vfio
Linus Torvalds [Sat, 6 Jan 2018 00:17:16 +0000 (16:17 -0800)]
Merge tag 'iommu-v4.15-rc7' of git://github.com/awilliam/linux-vfio

Pull IOMMU fixes from Alex Williamson:
 "Fixes via Will Deacon for arm-smmu-v3.

   - Fix duplicate Stream ID handling in arm-smmu-v3

   - Fix arm-smmu-v3 page table ops double free"

* tag 'iommu-v4.15-rc7' of git://github.com/awilliam/linux-vfio:
  iommu/arm-smmu-v3: Cope with duplicated Stream IDs
  iommu/arm-smmu-v3: Don't free page table ops twice

6 years agoMerge tag 'arc-4.15-rc7' of git://git.kernel.org/pub/scm/linux/kernel/git/vgupta/arc
Linus Torvalds [Sat, 6 Jan 2018 00:06:35 +0000 (16:06 -0800)]
Merge tag 'arc-4.15-rc7' of git://git./linux/kernel/git/vgupta/arc

Pull ARC fixes from Vineet Gupta:

 - platform updates for setting up clock correctly

 - fixes to accomodate newer gcc (__builtin_trap, removed inline asm
   modifier)

 - other fixes

* tag 'arc-4.15-rc7' of git://git.kernel.org/pub/scm/linux/kernel/git/vgupta/arc:
  ARC: handle gcc generated __builtin_trap for older compiler
  ARC: handle gcc generated __builtin_trap()
  ARC: uaccess: dont use "l" gcc inline asm constraint modifier
  ARC: [plat-axs103] refactor the quad core DT quirk code
  ARC: [plat-axs103]: Set initial core pll output frequency
  ARC: [plat-hsdk]: Get rid of core pll frequency set in platform code
  ARC: [plat-hsdk]: Set initial core pll output frequency
  ARC: [plat-hsdk] Switch DisplayLink driver from fbdev to DRM
  arc: do not use __print_symbol()
  ARC: Fix detection of dual-issue enabled

6 years agoapparmor: fix regression in mount mediation when feature set is pinned
John Johansen [Thu, 7 Dec 2017 08:28:27 +0000 (00:28 -0800)]
apparmor: fix regression in mount mediation when feature set is pinned

When the mount code was refactored for Labels it was not correctly
updated to check whether policy supported mediation of the mount
class.  This causes a regression when the kernel feature set is
reported as supporting mount and policy is pinned to a feature set
that does not support mount mediation.

BugLink: https://bugs.debian.org/cgi-bin/bugreport.cgi?bug=882697#41
Fixes: 2ea3ffb7782a ("apparmor: add mount mediation")
Reported-by: Fabian Grünbichler <f.gruenbichler@proxmox.com>
Cc: Stable <stable@vger.kernel.org>
Signed-off-by: John Johansen <john.johansen@canonical.com>
6 years agoMerge tag 'for-4.15-rc7-tag' of git://git.kernel.org/pub/scm/linux/kernel/git/kdave...
Linus Torvalds [Fri, 5 Jan 2018 21:02:46 +0000 (13:02 -0800)]
Merge tag 'for-4.15-rc7-tag' of git://git./linux/kernel/git/kdave/linux

Pull btrfs fixes from David Sterba:
 "We have two more fixes for 4.15, both aimed for stable.

  The leak fix is obvious, the second patch fixes a bug revealed by the
  refcount API, when it behaves differently than previous atomic_t and
  reports refs going from 0 to 1 in one case"

* tag 'for-4.15-rc7-tag' of git://git.kernel.org/pub/scm/linux/kernel/git/kdave/linux:
  btrfs: fix refcount_t usage when deleting btrfs_delayed_nodes
  btrfs: Fix flush bio leak

6 years agoMerge tag 'xfs-4.15-fixes-10' of git://git.kernel.org/pub/scm/fs/xfs/xfs-linux
Linus Torvalds [Fri, 5 Jan 2018 20:59:32 +0000 (12:59 -0800)]
Merge tag 'xfs-4.15-fixes-10' of git://git./fs/xfs/xfs-linux

Pull XFS fixes from Darrick Wong:
 "I have just a few fixes for bugs and resource cleanup problems this
  week:

   - Fix resource cleanup of failed quota initialization

   - Fix integer overflow problems wrt s_maxbytes"

* tag 'xfs-4.15-fixes-10' of git://git.kernel.org/pub/scm/fs/xfs/xfs-linux:
  xfs: fix s_maxbytes overflow problems
  xfs: quota: check result of register_shrinker()
  xfs: quota: fix missed destroy of qi_tree_lock

6 years agoMerge tag 'mfd-fixes-4.15-1' of git://git.kernel.org/pub/scm/linux/kernel/git/lee/mfd
Linus Torvalds [Fri, 5 Jan 2018 20:56:20 +0000 (12:56 -0800)]
Merge tag 'mfd-fixes-4.15-1' of git://git./linux/kernel/git/lee/mfd

Pull MFD fix from Lee Jones:
 "Late bugfix to plug a leak in rtsx_pcr"

* tag 'mfd-fixes-4.15-1' of git://git.kernel.org/pub/scm/linux/kernel/git/lee/mfd:
  mfd: rtsx: Release IRQ during shutdown

6 years agoMerge branch 'x86-pti-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git...
Linus Torvalds [Fri, 5 Jan 2018 20:23:57 +0000 (12:23 -0800)]
Merge branch 'x86-pti-for-linus' of git://git./linux/kernel/git/tip/tip

Pull  more x86 pti fixes from Thomas Gleixner:
 "Another small stash of fixes for fallout from the PTI work:

   - Fix the modules vs. KASAN breakage which was caused by making
     MODULES_END depend of the fixmap size. That was done when the cpu
     entry area moved into the fixmap, but now that we have a separate
     map space for that this is causing more issues than it solves.

   - Use the proper cache flush methods for the debugstore buffers as
     they are mapped/unmapped during runtime and not statically mapped
     at boot time like the rest of the cpu entry area.

   - Make the map layout of the cpu_entry_area consistent for 4 and 5
     level paging and fix the KASLR vaddr_end wreckage.

   - Use PER_CPU_EXPORT for per cpu variable and while at it unbreak
     nvidia gfx drivers by dropping the GPL export. The subject line of
     the commit tells it the other way around, but I noticed that too
     late.

   - Fix the ASM alternative macros so they can be used in the middle of
     an inline asm block.

   - Rename the BUG_CPU_INSECURE flag to BUG_CPU_MELTDOWN so the attack
     vector is properly identified. The Spectre mitigations will come
     with their own bug bits later"

* 'x86-pti-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip:
  x86/pti: Rename BUG_CPU_INSECURE to BUG_CPU_MELTDOWN
  x86/alternatives: Add missing '\n' at end of ALTERNATIVE inline asm
  x86/tlb: Drop the _GPL from the cpu_tlbstate export
  x86/events/intel/ds: Use the proper cache flush method for mapping ds buffers
  x86/kaslr: Fix the vaddr_end mess
  x86/mm: Map cpu_entry_area at the same place on 4/5 level
  x86/mm: Set MODULES_END to 0xffffffffff000000

6 years agoMerge branch 'efi-urgent-for-linus' of git://git.kernel.org/pub/scm/linux/kernel...
Linus Torvalds [Fri, 5 Jan 2018 20:20:35 +0000 (12:20 -0800)]
Merge branch 'efi-urgent-for-linus' of git://git./linux/kernel/git/tip/tip

Pull EFI updates from Thomas Gleixner:

 - A fix for a add_efi_memmap parameter regression which ensures that
   the parameter is parsed before it is used.

 - Reinstate the virtual capsule mapping as the cached copy turned out
   to break Quark and other things

 - Remove Matt Fleming as EFI co-maintainer. He stepped back a few days
   ago. Thanks Matt for all your great work!

* 'efi-urgent-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip:
  MAINTAINERS: Remove Matt Fleming as EFI co-maintainer
  efi/capsule-loader: Reinstate virtual capsule mapping
  x86/efi: Fix kernel param add_efi_memmap regression

6 years agoMerge branch 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/s390/linux
Linus Torvalds [Fri, 5 Jan 2018 20:17:33 +0000 (12:17 -0800)]
Merge branch 'for-linus' of git://git./linux/kernel/git/s390/linux

Pull s390 fixes from Martin Schwidefsky:
 "Four bug fixes"

* 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/s390/linux:
  s390/dasd: fix wrongly assigned configuration data
  s390: fix preemption race in disable_sacf_uaccess
  s390/sclp: disable FORTIFY_SOURCE for early sclp code
  s390/pci: handle insufficient resources during dma tlb flush

6 years agoMerge tag 'for-linus-4.15-rc7-tag' of git://git.kernel.org/pub/scm/linux/kernel/git...
Linus Torvalds [Fri, 5 Jan 2018 20:15:12 +0000 (12:15 -0800)]
Merge tag 'for-linus-4.15-rc7-tag' of git://git./linux/kernel/git/xen/tip

Pull xen fix from Juergen Gross:
 "One minor fix adjusting the kmalloc flags in the new pvcalls driver
  added in rc1"

* tag 'for-linus-4.15-rc7-tag' of git://git.kernel.org/pub/scm/linux/kernel/git/xen/tip:
  xen/pvcalls: use GFP_ATOMIC under spin lock

6 years agoMerge branch 'linus' of git://git.kernel.org/pub/scm/linux/kernel/git/herbert/crypto-2.6
Linus Torvalds [Fri, 5 Jan 2018 20:10:06 +0000 (12:10 -0800)]
Merge branch 'linus' of git://git./linux/kernel/git/herbert/crypto-2.6

Pull crypto fixes from Herbert Xu:
 "This fixes the following issues:

   - racy use of ctx->rcvused in af_alg

   - algif_aead crash in chacha20poly1305

   - freeing bogus pointer in pcrypt

   - build error on MIPS in mpi

   - memory leak in inside-secure

   - memory overwrite in inside-secure

   - NULL pointer dereference in inside-secure

   - state corruption in inside-secure

   - build error without CRYPTO_GF128MUL in chelsio

   - use after free in n2"

* 'linus' of git://git.kernel.org/pub/scm/linux/kernel/git/herbert/crypto-2.6:
  crypto: inside-secure - do not use areq->result for partial results
  crypto: inside-secure - fix request allocations in invalidation path
  crypto: inside-secure - free requests even if their handling failed
  crypto: inside-secure - per request invalidation
  lib/mpi: Fix umul_ppmm() for MIPS64r6
  crypto: pcrypt - fix freeing pcrypt instances
  crypto: n2 - cure use after free
  crypto: af_alg - Fix race around ctx->rcvused by making it atomic_t
  crypto: chacha20poly1305 - validate the digest size
  crypto: chelsio - select CRYPTO_GF128MUL

6 years agoMerge branch 'akpm' (patches from Andrew)
Linus Torvalds [Fri, 5 Jan 2018 19:26:09 +0000 (11:26 -0800)]
Merge branch 'akpm' (patches from Andrew)

Merge misc fixes from Andrew Morton:
 "9 fixes"

* emailed patches from Andrew Morton <akpm@linux-foundation.org>:
  mailmap: update Mark Yao's email address
  userfaultfd: clear the vma->vm_userfaultfd_ctx if UFFD_EVENT_FORK fails
  mm/sparse.c: wrong allocation for mem_section
  mm/zsmalloc.c: include fs.h
  mm/debug.c: provide useful debugging information for VM_BUG
  kernel/exit.c: export abort() to modules
  mm/mprotect: add a cond_resched() inside change_pmd_range()
  kernel/acct.c: fix the acct->needcheck check in check_free_space()
  mm: check pfn_valid first in zero_resv_unavail

6 years agofix "netfilter: xt_bpf: Fix XT_BPF_MODE_FD_PINNED mode of 'xt_bpf_info_v1'"
Al Viro [Sun, 3 Dec 2017 01:20:38 +0000 (20:20 -0500)]
fix "netfilter: xt_bpf: Fix XT_BPF_MODE_FD_PINNED mode of 'xt_bpf_info_v1'"

Descriptor table is a shared object; it's not a place where you can
stick temporary references to files, especially when we don't need
an opened file at all.

Cc: stable@vger.kernel.org # v4.14
Fixes: 98589a0998b8 ("netfilter: xt_bpf: Fix XT_BPF_MODE_FD_PINNED mode of 'xt_bpf_info_v1'")
Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
6 years agokvm: vmx: Scrub hardware GPRs at VM-exit
Jim Mattson [Wed, 3 Jan 2018 22:31:38 +0000 (14:31 -0800)]
kvm: vmx: Scrub hardware GPRs at VM-exit

Guest GPR values are live in the hardware GPRs at VM-exit.  Do not
leave any guest values in hardware GPRs after the guest GPR values are
saved to the vcpu_vmx structure.

This is a partial mitigation for CVE 2017-5715 and CVE 2017-5753.
Specifically, it defeats the Project Zero PoC for CVE 2017-5715.

Suggested-by: Eric Northup <digitaleric@google.com>
Signed-off-by: Jim Mattson <jmattson@google.com>
Reviewed-by: Eric Northup <digitaleric@google.com>
Reviewed-by: Benjamin Serebrin <serebrin@google.com>
Reviewed-by: Andrew Honig <ahonig@google.com>
[Paolo: Add AMD bits, Signed-off-by: Tom Lendacky <thomas.lendacky@amd.com>]
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
6 years agox86/pti: Rename BUG_CPU_INSECURE to BUG_CPU_MELTDOWN
Thomas Gleixner [Fri, 5 Jan 2018 14:27:34 +0000 (15:27 +0100)]
x86/pti: Rename BUG_CPU_INSECURE to BUG_CPU_MELTDOWN

Use the name associated with the particular attack which needs page table
isolation for mitigation.

Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Acked-by: David Woodhouse <dwmw@amazon.co.uk>
Cc: Alan Cox <gnomes@lxorguk.ukuu.org.uk>
Cc: Jiri Koshina <jikos@kernel.org>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Tim Chen <tim.c.chen@linux.intel.com>
Cc: Andi Lutomirski <luto@amacapital.net>
Cc: Andi Kleen <ak@linux.intel.com>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Paul Turner <pjt@google.com>
Cc: Tom Lendacky <thomas.lendacky@amd.com>
Cc: Greg KH <gregkh@linux-foundation.org>
Cc: Dave Hansen <dave.hansen@intel.com>
Cc: Kees Cook <keescook@google.com>
Cc: stable@vger.kernel.org
Link: https://lkml.kernel.org/r/alpine.DEB.2.20.1801051525300.1724@nanos
6 years agox86/alternatives: Add missing '\n' at end of ALTERNATIVE inline asm
David Woodhouse [Thu, 4 Jan 2018 14:37:05 +0000 (14:37 +0000)]
x86/alternatives: Add missing '\n' at end of ALTERNATIVE inline asm

Where an ALTERNATIVE is used in the middle of an inline asm block, this
would otherwise lead to the following instruction being appended directly
to the trailing ".popsection", and a failed compile.

Fixes: 9cebed423c84 ("x86, alternative: Use .pushsection/.popsection")
Signed-off-by: David Woodhouse <dwmw@amazon.co.uk>
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Cc: gnomes@lxorguk.ukuu.org.uk
Cc: Rik van Riel <riel@redhat.com>
Cc: ak@linux.intel.com
Cc: Tim Chen <tim.c.chen@linux.intel.com>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Paul Turner <pjt@google.com>
Cc: Jiri Kosina <jikos@kernel.org>
Cc: Andy Lutomirski <luto@amacapital.net>
Cc: Dave Hansen <dave.hansen@intel.com>
Cc: Kees Cook <keescook@google.com>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Greg Kroah-Hartman <gregkh@linux-foundation.org>
Cc: stable@vger.kernel.org
Link: https://lkml.kernel.org/r/20180104143710.8961-8-dwmw@amazon.co.uk
6 years agomfd: rtsx: Release IRQ during shutdown
Sinan Kaya [Wed, 3 Jan 2018 12:32:45 +0000 (07:32 -0500)]
mfd: rtsx: Release IRQ during shutdown

'Commit cc27b735ad3a ("PCI/portdrv: Turn off PCIe services during
shutdown")' revealed a resource leak in rtsx_pci driver during shutdown.

Issue shows up as a warning during shutdown as follows:

remove_proc_entry: removing non-empty directory 'irq/17', leaking at least
'rtsx_pci'
WARNING: CPU: 0 PID: 1578 at fs/proc/generic.c:572
remove_proc_entry+0x11d/0x130
Modules linked in <long list but none that are out-of-tree>
...
Call Trace:
unregister_irq_proc
free_desc
irq_free_descs
mp_unmap_irq
acpi_unregister_gsi_apic
acpi_pci_irq_disable
do_pci_disable_device
pci_disable_device
device_shutdown
kernel_restart
Sys_reboot

Even though rtsx_pci driver implements a shutdown callback, it is not
releasing the interrupt that it registered during probe. This is causing
the ACPI layer to complain that the shared IRQ is in use while freeing
IRQ.

This code releases the IRQ to prevent resource leak and eliminate the
warning.

Fixes: cc27b735ad3a ("PCI/portdrv: Turn off PCIe services during shutdown")
Link: https://bugzilla.kernel.org/show_bug.cgi?id=198141
Reported-by: Chris Clayton <chris2553@googlemail.com>
Signed-off-by: Sinan Kaya <okaya@codeaurora.org>
Reviewed-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
Signed-off-by: Lee Jones <lee.jones@linaro.org>
6 years agoMerge tag 'drm-fixes-for-v4.15-rc7' of git://people.freedesktop.org/~airlied/linux
Linus Torvalds [Fri, 5 Jan 2018 02:02:55 +0000 (18:02 -0800)]
Merge tag 'drm-fixes-for-v4.15-rc7' of git://people.freedesktop.org/~airlied/linux

Pull drm fixes from Dave Airlie:
 "Just collecting some fixes to finish my hoildays :-).

  A few fixes for i915 (one documentation build fix), one ttm fix, one
  AMD display fix, one omapdrm fix, and a set of armada fixes from
  Russell.

  All seem pretty small, you can now return to your latest security news
  site"

* tag 'drm-fixes-for-v4.15-rc7' of git://people.freedesktop.org/~airlied/linux:
  drm/i915: Apply Display WA #1183 on skl, kbl, and cfl
  drm/ttm: check the return value of kzalloc
  drm/amd/display: call set csc_default if enable adjustment is false
  docs: fix, intel_guc_loader.c has been moved to intel_guc_fw.c
  omapdrm/dss/hdmi4_cec: fix interrupt handling
  documentation/gpu/i915: fix docs build error after file rename
  drm/i915: Put all non-blocking modesets onto an ordered wq
  drm/i915: Disable DC states around GMBUS on GLK
  drm/i915/psr: Fix register name mess up.
  drm/armada: fix YUV planar format framebuffer offsets
  drm/armada: improve efficiency of armada_drm_plane_calc_addrs()
  drm/armada: fix UV swap code
  drm/armada: fix SRAM powerdown
  drm/armada: fix leak of crtc structure

6 years agomailmap: update Mark Yao's email address
Jeffy Chen [Fri, 5 Jan 2018 00:18:12 +0000 (16:18 -0800)]
mailmap: update Mark Yao's email address

Change the previous employers email addresses to the current email
address.

Link: http://lkml.kernel.org/r/20171229121726.31589-1-jeffy.chen@rock-chips.com
Signed-off-by: Jeffy Chen <jeffy.chen@rock-chips.com>
Acked-by: Martin Kepplinger <martink@posteo.de>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
6 years agouserfaultfd: clear the vma->vm_userfaultfd_ctx if UFFD_EVENT_FORK fails
Andrea Arcangeli [Fri, 5 Jan 2018 00:18:09 +0000 (16:18 -0800)]
userfaultfd: clear the vma->vm_userfaultfd_ctx if UFFD_EVENT_FORK fails

The previous fix in commit 384632e67e08 ("userfaultfd: non-cooperative:
fix fork use after free") corrected the refcounting in case of
UFFD_EVENT_FORK failure for the fork userfault paths.

That still didn't clear the vma->vm_userfaultfd_ctx of the vmas that
were set to point to the aborted new uffd ctx earlier in
dup_userfaultfd.

Link: http://lkml.kernel.org/r/20171223002505.593-2-aarcange@redhat.com
Signed-off-by: Andrea Arcangeli <aarcange@redhat.com>
Reported-by: syzbot <syzkaller@googlegroups.com>
Reviewed-by: Mike Rapoport <rppt@linux.vnet.ibm.com>
Cc: Eric Biggers <ebiggers3@gmail.com>
Cc: Dmitry Vyukov <dvyukov@google.com>
Cc: <stable@vger.kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
6 years agomm/sparse.c: wrong allocation for mem_section
Baoquan He [Fri, 5 Jan 2018 00:18:06 +0000 (16:18 -0800)]
mm/sparse.c: wrong allocation for mem_section

In commit 83e3c48729d9 ("mm/sparsemem: Allocate mem_section at runtime
for CONFIG_SPARSEMEM_EXTREME=y") mem_section is allocated at runtime to
save memory.

It allocates the first dimension of array with sizeof(struct mem_section).

It costs extra memory, should be sizeof(struct mem_section *).

Fix it.

Link: http://lkml.kernel.org/r/1513932498-20350-1-git-send-email-bhe@redhat.com
Fixes: 83e3c48729 ("mm/sparsemem: Allocate mem_section at runtime for CONFIG_SPARSEMEM_EXTREME=y")
Signed-off-by: Baoquan He <bhe@redhat.com>
Tested-by: Dave Young <dyoung@redhat.com>
Acked-by: Kirill A. Shutemov <kirill.shutemov@linux.intel.com>
Cc: Kirill A. Shutemov <kirill.shutemov@linux.intel.com>
Cc: Ingo Molnar <mingo@kernel.org>
Cc: Andy Lutomirski <luto@amacapital.net>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: Atsushi Kumagai <ats-kumagai@wm.jp.nec.com>
Cc: <stable@vger.kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
6 years agomm/zsmalloc.c: include fs.h
Sergey Senozhatsky [Fri, 5 Jan 2018 00:18:02 +0000 (16:18 -0800)]
mm/zsmalloc.c: include fs.h

`struct file_system_type' and alloc_anon_inode() function are defined in
fs.h, include it directly.

Link: http://lkml.kernel.org/r/20171219104219.3017-1-sergey.senozhatsky@gmail.com
Signed-off-by: Sergey Senozhatsky <sergey.senozhatsky@gmail.com>
Cc: Minchan Kim <minchan@kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
6 years agomm/debug.c: provide useful debugging information for VM_BUG
Matthew Wilcox [Fri, 5 Jan 2018 00:17:59 +0000 (16:17 -0800)]
mm/debug.c: provide useful debugging information for VM_BUG

With the recent addition of hashed kernel pointers, places which need to
produce useful debug output have to specify %px, not %p.  This patch
fixes all the VM debug to use %px.  This is appropriate because it's
debug output that the user should never be able to trigger, and kernel
developers need to see the actual pointers.

Link: http://lkml.kernel.org/r/20171219133236.GE13680@bombadil.infradead.org
Signed-off-by: Matthew Wilcox <mawilcox@microsoft.com>
Acked-by: Michal Hocko <mhocko@suse.com>
Cc: "Tobin C. Harding" <me@tobin.cc>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
6 years agokernel/exit.c: export abort() to modules
Andrew Morton [Fri, 5 Jan 2018 00:17:56 +0000 (16:17 -0800)]
kernel/exit.c: export abort() to modules

gcc -fisolate-erroneous-paths-dereference can generate calls to abort()
from modular code too.

[arnd@arndb.de: drop duplicate exports of abort()]
Link: http://lkml.kernel.org/r/20180102103311.706364-1-arnd@arndb.de
Reported-by: Vineet Gupta <Vineet.Gupta1@synopsys.com>
Cc: Sudip Mukherjee <sudipm.mukherjee@gmail.com>
Cc: Arnd Bergmann <arnd@arndb.de>
Cc: Alexey Brodkin <Alexey.Brodkin@synopsys.com>
Cc: Russell King <rmk+kernel@armlinux.org.uk>
Cc: Jose Abreu <Jose.Abreu@synopsys.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Arnd Bergmann <arnd@arndb.de>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
6 years agomm/mprotect: add a cond_resched() inside change_pmd_range()
Anshuman Khandual [Fri, 5 Jan 2018 00:17:52 +0000 (16:17 -0800)]
mm/mprotect: add a cond_resched() inside change_pmd_range()

While testing on a large CPU system, detected the following RCU stall
many times over the span of the workload.  This problem is solved by
adding a cond_resched() in the change_pmd_range() function.

  INFO: rcu_sched detected stalls on CPUs/tasks:
   154-....: (670 ticks this GP) idle=022/140000000000000/0 softirq=2825/2825 fqs=612
   (detected by 955, t=6002 jiffies, g=4486, c=4485, q=90864)
  Sending NMI from CPU 955 to CPUs 154:
  NMI backtrace for cpu 154
  CPU: 154 PID: 147071 Comm: workload Not tainted 4.15.0-rc3+ #3
  NIP:  c0000000000b3f64 LR: c0000000000b33d4 CTR: 000000000000aa18
  REGS: 00000000a4b0fb44 TRAP: 0501   Not tainted  (4.15.0-rc3+)
  MSR:  8000000000009033 <SF,EE,ME,IR,DR,RI,LE>  CR: 22422082  XER: 00000000
  CFAR: 00000000006cf8f0 SOFTE: 1
  GPR00: 0010000000000000 c00003ef9b1cb8c0 c0000000010cc600 0000000000000000
  GPR04: 8e0000018c32b200 40017b3858fd6e00 8e0000018c32b208 40017b3858fd6e00
  GPR08: 8e0000018c32b210 40017b3858fd6e00 8e0000018c32b218 40017b3858fd6e00
  GPR12: ffffffffffffffff c00000000fb25100
  NIP [c0000000000b3f64] plpar_hcall9+0x44/0x7c
  LR [c0000000000b33d4] pSeries_lpar_flush_hash_range+0x384/0x420
  Call Trace:
    flush_hash_range+0x48/0x100
    __flush_tlb_pending+0x44/0xd0
    hpte_need_flush+0x408/0x470
    change_protection_range+0xaac/0xf10
    change_prot_numa+0x30/0xb0
    task_numa_work+0x2d0/0x3e0
    task_work_run+0x130/0x190
    do_notify_resume+0x118/0x120
    ret_from_except_lite+0x70/0x74
  Instruction dump:
  60000000 f8810028 7ca42b78 7cc53378 7ce63b78 7d074378 7d284b78 7d495378
  e9410060 e9610068 e9810070 44000022 <7d806378e9810028 f88c0000 f8ac0008

Link: http://lkml.kernel.org/r/20171214140551.5794-1-khandual@linux.vnet.ibm.com
Signed-off-by: Anshuman Khandual <khandual@linux.vnet.ibm.com>
Suggested-by: Nicholas Piggin <npiggin@gmail.com>
Acked-by: Michal Hocko <mhocko@suse.com>
Cc: <stable@vger.kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
6 years agokernel/acct.c: fix the acct->needcheck check in check_free_space()
Oleg Nesterov [Fri, 5 Jan 2018 00:17:49 +0000 (16:17 -0800)]
kernel/acct.c: fix the acct->needcheck check in check_free_space()

As Tsukada explains, the time_is_before_jiffies(acct->needcheck) check
is very wrong, we need time_is_after_jiffies() to make sys_acct() work.

Ignoring the overflows, the code should "goto out" if needcheck >
jiffies, while currently it checks "needcheck < jiffies" and thus in the
likely case check_free_space() does nothing until jiffies overflow.

In particular this means that sys_acct() is simply broken, acct_on()
sets acct->needcheck = jiffies and expects that check_free_space()
should set acct->active = 1 after the free-space check, but this won't
happen if jiffies increments in between.

This was broken by commit 32dc73086015 ("get rid of timer in
kern/acct.c") in 2011, then another (correct) commit 795a2f22a8ea
("acct() should honour the limits from the very beginning") made the
problem more visible.

Link: http://lkml.kernel.org/r/20171213133940.GA6554@redhat.com
Fixes: 32dc73086015 ("get rid of timer in kern/acct.c")
Reported-by: TSUKADA Koutaro <tsukada@ascade.co.jp>
Suggested-by: TSUKADA Koutaro <tsukada@ascade.co.jp>
Signed-off-by: Oleg Nesterov <oleg@redhat.com>
Cc: Al Viro <viro@zeniv.linux.org.uk>
Cc: <stable@vger.kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
6 years agomm: check pfn_valid first in zero_resv_unavail
Dave Young [Fri, 5 Jan 2018 00:17:45 +0000 (16:17 -0800)]
mm: check pfn_valid first in zero_resv_unavail

With latest kernel I get below bug while testing kdump:

  BUG: unable to handle kernel paging request at ffffea00034b1040
  IP: zero_resv_unavail+0xbd/0x126
  PGD 37b98067 P4D 37b98067 PUD 37b97067 PMD 0
  Oops: 0002 [#1] SMP
  Modules linked in:
  CPU: 0 PID: 0 Comm: swapper Not tainted 4.15.0-rc1+ #316
  Hardware name: LENOVO 20ARS1BJ02/20ARS1BJ02, BIOS GJET92WW (2.42 ) 03/03/2017
  task: ffffffff81a0e4c0 task.stack: ffffffff81a00000
  RIP: 0010:zero_resv_unavail+0xbd/0x126
  RSP: 0000:ffffffff81a03d88 EFLAGS: 00010006
  RAX: 0000000000000000 RBX: ffffea00034b1040 RCX: 0000000000000010
  RDX: 0000000000000000 RSI: 0000000000000092 RDI: ffffea00034b1040
  RBP: 00000000000d2c41 R08: 00000000000000c0 R09: 0000000000000a0d
  R10: 0000000000000002 R11: 0000000000007f01 R12: ffffffff81a03d90
  R13: ffffea0000000000 R14: 0000000000000063 R15: 0000000000000062
  FS:  0000000000000000(0000) GS:ffffffff81c73000(0000) knlGS:0000000000000000
  CS:  0010 DS: 0000 ES: 0000 CR0: 0000000080050033
  CR2: ffffea00034b1040 CR3: 0000000037609000 CR4: 00000000000606b0
  Call Trace:
   ? free_area_init_nodes+0x640/0x664
   ? zone_sizes_init+0x58/0x72
   ? setup_arch+0xb50/0xc6c
   ? start_kernel+0x64/0x43d
   ? secondary_startup_64+0xa5/0xb0
  Code: c1 e8 0c 48 39 d8 76 27 48 89 de 48 c1 e3 06 48 c7 c7 7a 87 79 81 e8 b0 c0 3e ff 4c 01 eb b9 10 00 00 00 31 c0 48 89 df 49 ff c6 <f3> ab eb bc 6a 00 49 c7 c0 f0 93 d1 81 31 d2 83 ce ff 41 54 49
  RIP: zero_resv_unavail+0xbd/0x126 RSP: ffffffff81a03d88
  CR2: ffffea00034b1040
  ---[ end trace f5ba9e8f73c7ee26 ]---

This is introduced by commit a4a3ede2132a ("mm: zero reserved and
unavailable struct pages").

The reason is some efi reserved boot ranges is not reported in E820 ram.
In my case it is a bgrt buffer:

  efi: mem00: [Boot Data          |RUN|  |  |  |  |  |  |   |WB|WT|WC|UC] range=[0x00000000d2c41000-0x00000000d2c85fff] (0MB)

Use "add_efi_memmap" can workaround the problem with another fix:

  http://lkml.kernel.org/r/20171130052327.GA3500@dhcp-128-65.nay.redhat.com

In zero_resv_unavail it would be better to check pfn_valid first before
zero the page struct.  This fixes the problem and potential other
similar problems.  Also as Pavel Tatashin suggested checks pfn_valid at
the beginning of the section.

The range is backed by real memory.  The memory range is efi "Boot
Service Data", that means after ExitBootServices() these ranges can be
used as system ram.  But some of them need to be reserved, for example
the bgrt image address in an acpi table, if the image memory is freed
then kexec reboot will fail because kexec inherit same acpi table to
initialize the driver.

Link: http://lkml.kernel.org/r/20171201095048.GA3084@dhcp-128-65.nay.redhat.com
Fixes: a4a3ede2132a ("mm: zero reserved and unavailable struct pages")
Signed-off-by: Dave Young <dyoung@redhat.com>
Cc: Michal Hocko <mhocko@kernel.org>
Cc: Pavel Tatashin <pasha.tatashin@oracle.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
6 years agox86/tlb: Drop the _GPL from the cpu_tlbstate export
Thomas Gleixner [Thu, 4 Jan 2018 21:19:04 +0000 (22:19 +0100)]
x86/tlb: Drop the _GPL from the cpu_tlbstate export

The recent changes for PTI touch cpu_tlbstate from various tlb_flush
inlines. cpu_tlbstate is exported as GPL symbol, so this causes a
regression when building out of tree drivers for certain graphics cards.

Aside of that the export was wrong since it was introduced as it should
have been EXPORT_PER_CPU_SYMBOL_GPL().

Use the correct PER_CPU export and drop the _GPL to restore the previous
state which allows users to utilize the cards they payed for.

As always I'm really thrilled to make this kind of change to support the
#friends (or however the hot hashtag of today is spelled) from that closet
sauce graphics corp.

Fixes: 1e02ce4cccdc ("x86: Store a per-cpu shadow copy of CR4")
Fixes: 6fd166aae78c ("x86/mm: Use/Fix PCID to optimize user/kernel switches")
Reported-by: Kees Cook <keescook@google.com>
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Cc: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Andy Lutomirski <luto@kernel.org>
Cc: stable@vger.kernel.org
6 years agox86/events/intel/ds: Use the proper cache flush method for mapping ds buffers
Peter Zijlstra [Thu, 4 Jan 2018 17:07:12 +0000 (18:07 +0100)]
x86/events/intel/ds: Use the proper cache flush method for mapping ds buffers

Thomas reported the following warning:

 BUG: using smp_processor_id() in preemptible [00000000] code: ovsdb-server/4498
 caller is native_flush_tlb_single+0x57/0xc0
 native_flush_tlb_single+0x57/0xc0
 __set_pte_vaddr+0x2d/0x40
 set_pte_vaddr+0x2f/0x40
 cea_set_pte+0x30/0x40
 ds_update_cea.constprop.4+0x4d/0x70
 reserve_ds_buffers+0x159/0x410
 x86_reserve_hardware+0x150/0x160
 x86_pmu_event_init+0x3e/0x1f0
 perf_try_init_event+0x69/0x80
 perf_event_alloc+0x652/0x740
 SyS_perf_event_open+0x3f6/0xd60
 do_syscall_64+0x5c/0x190

set_pte_vaddr is used to map the ds buffers into the cpu entry area, but
there are two problems with that:

 1) The resulting flush is not supposed to be called in preemptible context

 2) The cpu entry area is supposed to be per CPU, but the debug store
    buffers are mapped for all CPUs so these mappings need to be flushed
    globally.

Add the necessary preemption protection across the mapping code and flush
TLBs globally.

Fixes: c1961a4631da ("x86/events/intel/ds: Map debug buffers in cpu_entry_area")
Reported-by: Thomas Zeitlhofer <thomas.zeitlhofer+lkml@ze-it.at>
Signed-off-by: Peter Zijlstra <peterz@infradead.org>
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Tested-by: Thomas Zeitlhofer <thomas.zeitlhofer+lkml@ze-it.at>
Cc: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Cc: Hugh Dickins <hughd@google.com>
Cc: stable@vger.kernel.org
Link: https://lkml.kernel.org/r/20180104170712.GB3040@hirez.programming.kicks-ass.net
6 years agox86/kaslr: Fix the vaddr_end mess
Thomas Gleixner [Thu, 4 Jan 2018 11:32:03 +0000 (12:32 +0100)]
x86/kaslr: Fix the vaddr_end mess

vaddr_end for KASLR is only documented in the KASLR code itself and is
adjusted depending on config options. So it's not surprising that a change
of the memory layout causes KASLR to have the wrong vaddr_end. This can map
arbitrary stuff into other areas causing hard to understand problems.

Remove the whole ifdef magic and define the start of the cpu_entry_area to
be the end of the KASLR vaddr range.

Add documentation to that effect.

Fixes: 92a0f81d8957 ("x86/cpu_entry_area: Move it out of the fixmap")
Reported-by: Benjamin Gilbert <benjamin.gilbert@coreos.com>
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Tested-by: Benjamin Gilbert <benjamin.gilbert@coreos.com>
Cc: Andy Lutomirski <luto@kernel.org>
Cc: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Cc: stable <stable@vger.kernel.org>
Cc: Dave Hansen <dave.hansen@linux.intel.com>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Thomas Garnier <thgarnie@google.com>,
Cc: Alexander Kuleshov <kuleshovmail@gmail.com>
Link: https://lkml.kernel.org/r/alpine.DEB.2.20.1801041320360.1771@nanos
6 years agoMerge tag 'drm-intel-fixes-2018-01-04' of git://anongit.freedesktop.org/drm/drm-intel...
Dave Airlie [Thu, 4 Jan 2018 23:25:01 +0000 (09:25 +1000)]
Merge tag 'drm-intel-fixes-2018-01-04' of git://anongit.freedesktop.org/drm/drm-intel into drm-fixes

drm/i915 fixes for v4.15-rc7
- couple of documentation build fixes
- serialize non-blocking modesets
- prevent DMC from messing up GMBUS transfers
- PSR regression fix

* tag 'drm-intel-fixes-2018-01-04' of git://anongit.freedesktop.org/drm/drm-intel:
  drm/i915: Apply Display WA #1183 on skl, kbl, and cfl
  docs: fix, intel_guc_loader.c has been moved to intel_guc_fw.c
  documentation/gpu/i915: fix docs build error after file rename
  drm/i915: Put all non-blocking modesets onto an ordered wq
  drm/i915: Disable DC states around GMBUS on GLK
  drm/i915/psr: Fix register name mess up.

6 years agoMerge branch 'drm-fixes-4.15' of git://people.freedesktop.org/~agd5f/linux into drm...
Dave Airlie [Thu, 4 Jan 2018 23:24:26 +0000 (09:24 +1000)]
Merge branch 'drm-fixes-4.15' of git://people.freedesktop.org/~agd5f/linux into drm-fixes

- backport of a DC change which fixes a greenish tint on some RV hw
- properly handle kzalloc fail in ttm

* 'drm-fixes-4.15' of git://people.freedesktop.org/~agd5f/linux:
  drm/ttm: check the return value of kzalloc
  drm/amd/display: call set csc_default if enable adjustment is false

6 years agoMerge branch 'drm-armada-fixes-4.15' of git://git.armlinux.org.uk/~rmk/linux-arm...
Dave Airlie [Thu, 4 Jan 2018 23:24:05 +0000 (09:24 +1000)]
Merge branch 'drm-armada-fixes-4.15' of git://git.armlinux.org.uk/~rmk/linux-arm into drm-fixes

Armada fixes.

* 'drm-armada-fixes-4.15' of git://git.armlinux.org.uk/~rmk/linux-arm:
  drm/armada: fix YUV planar format framebuffer offsets
  drm/armada: improve efficiency of armada_drm_plane_calc_addrs()
  drm/armada: fix UV swap code
  drm/armada: fix SRAM powerdown
  drm/armada: fix leak of crtc structure

6 years agoMerge tag 'omapdrm-4.15-fixes' of git://git.kernel.org/pub/scm/linux/kernel/git/tomba...
Dave Airlie [Thu, 4 Jan 2018 23:23:35 +0000 (09:23 +1000)]
Merge tag 'omapdrm-4.15-fixes' of git://git./linux/kernel/git/tomba/linux into drm-fixes

omapdrm fixes for 4.15

* Fix OMAP4 HDMI CEC interrupt handling and a possible buffer overflow

* tag 'omapdrm-4.15-fixes' of git://git.kernel.org/pub/scm/linux/kernel/git/tomba/linux:
  omapdrm/dss/hdmi4_cec: fix interrupt handling

6 years agox86/mm: Map cpu_entry_area at the same place on 4/5 level
Thomas Gleixner [Thu, 4 Jan 2018 12:01:40 +0000 (13:01 +0100)]
x86/mm: Map cpu_entry_area at the same place on 4/5 level

There is no reason for 4 and 5 level pagetables to have a different
layout. It just makes determining vaddr_end for KASLR harder than
necessary.

Fixes: 92a0f81d8957 ("x86/cpu_entry_area: Move it out of the fixmap")
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Cc: Andy Lutomirski <luto@kernel.org>
Cc: Benjamin Gilbert <benjamin.gilbert@coreos.com>
Cc: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Cc: stable <stable@vger.kernel.org>
Cc: Dave Hansen <dave.hansen@linux.intel.com>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Thomas Garnier <thgarnie@google.com>,
Cc: Alexander Kuleshov <kuleshovmail@gmail.com>
Link: https://lkml.kernel.org/r/alpine.DEB.2.20.1801041320360.1771@nanos
6 years agox86/mm: Set MODULES_END to 0xffffffffff000000
Andrey Ryabinin [Thu, 28 Dec 2017 16:06:20 +0000 (19:06 +0300)]
x86/mm: Set MODULES_END to 0xffffffffff000000

Since f06bdd4001c2 ("x86/mm: Adapt MODULES_END based on fixmap section size")
kasan_mem_to_shadow(MODULES_END) could be not aligned to a page boundary.

So passing page unaligned address to kasan_populate_zero_shadow() have two
possible effects:

1) It may leave one page hole in supposed to be populated area. After commit
  21506525fb8d ("x86/kasan/64: Teach KASAN about the cpu_entry_area") that
  hole happens to be in the shadow covering fixmap area and leads to crash:

 BUG: unable to handle kernel paging request at fffffbffffe8ee04
 RIP: 0010:check_memory_region+0x5c/0x190

 Call Trace:
  <NMI>
  memcpy+0x1f/0x50
  ghes_copy_tofrom_phys+0xab/0x180
  ghes_read_estatus+0xfb/0x280
  ghes_notify_nmi+0x2b2/0x410
  nmi_handle+0x115/0x2c0
  default_do_nmi+0x57/0x110
  do_nmi+0xf8/0x150
  end_repeat_nmi+0x1a/0x1e

Note, the crash likely disappeared after commit 92a0f81d8957, which
changed kasan_populate_zero_shadow() call the way it was before
commit 21506525fb8d.

2) Attempt to load module near MODULES_END will fail, because
   __vmalloc_node_range() called from kasan_module_alloc() will hit the
   WARN_ON(!pte_none(*pte)) in the vmap_pte_range() and bail out with error.

To fix this we need to make kasan_mem_to_shadow(MODULES_END) page aligned
which means that MODULES_END should be 8*PAGE_SIZE aligned.

The whole point of commit f06bdd4001c2 was to move MODULES_END down if
NR_CPUS is big, so the cpu_entry_area takes a lot of space.
But since 92a0f81d8957 ("x86/cpu_entry_area: Move it out of the fixmap")
the cpu_entry_area is no longer in fixmap, so we could just set
MODULES_END to a fixed 8*PAGE_SIZE aligned address.

Fixes: f06bdd4001c2 ("x86/mm: Adapt MODULES_END based on fixmap section size")
Reported-by: Jakub Kicinski <kubakici@wp.pl>
Signed-off-by: Andrey Ryabinin <aryabinin@virtuozzo.com>
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Cc: stable@vger.kernel.org
Cc: Andy Lutomirski <luto@kernel.org>
Cc: Thomas Garnier <thgarnie@google.com>
Link: https://lkml.kernel.org/r/20171228160620.23818-1-aryabinin@virtuozzo.com