fs/netfs: Add folio fscache functions

Match the page writeback functions by adding
folio_start_fscache(), folio_end_fscache(), folio_wait_fscache() and
folio_wait_fscache_killable().  Remove set_page_private_2().  Also rewrite
the kernel-doc to describe when to use the function rather than what the
function does, and include the kernel-doc in the appropriate rst file.
Saves 31 bytes of text in netfs_rreq_unlock() due to set_page_fscache()
calling page_folio() once instead of three times.

Signed-off-by: Matthew Wilcox (Oracle) <willy@infradead.org>
Acked-by: Vlastimil Babka <vbabka@suse.cz>
Reviewed-by: William Kucharski <william.kucharski@oracle.com>
Reviewed-by: Christoph Hellwig <hch@lst.de>
Acked-by: Kirill A. Shutemov <kirill.shutemov@linux.intel.com>
Acked-by: Mike Rapoport <rppt@linux.ibm.com>
Reviewed-by: David Howells <dhowells@redhat.com>

mm/filemap: Add folio private_2 functions

end_page_private_2() becomes folio_end_private_2(),
wait_on_page_private_2() becomes folio_wait_private_2() and
wait_on_page_private_2_killable() becomes folio_wait_private_2_killable().

Adjust the fscache equivalents to call page_folio() before calling these
functions to avoid adding wrappers.  Ends up costing 1 byte of text
in ceph & netfs, but the core shrinks by three calls to page_folio().

Signed-off-by: Matthew Wilcox (Oracle) <willy@infradead.org>
Acked-by: Vlastimil Babka <vbabka@suse.cz>
Reviewed-by: William Kucharski <william.kucharski@oracle.com>
Reviewed-by: Christoph Hellwig <hch@lst.de>
Reviewed-by: David Howells <dhowells@redhat.com>
Acked-by: Kirill A. Shutemov <kirill.shutemov@linux.intel.com>

mm/filemap: Convert page wait queues to be folios

Reinforce that page flags are actually in the head page by changing the
type from page to folio.  Increases the size of cachefiles by two bytes,
but the kernel core is unchanged in size.

Signed-off-by: Matthew Wilcox (Oracle) <willy@infradead.org>
Reviewed-by: Christoph Hellwig <hch@lst.de>
Acked-by: Jeff Layton <jlayton@kernel.org>
Acked-by: Kirill A. Shutemov <kirill.shutemov@linux.intel.com>
Acked-by: Vlastimil Babka <vbabka@suse.cz>
Reviewed-by: William Kucharski <william.kucharski@oracle.com>
Reviewed-by: David Howells <dhowells@redhat.com>

mm/filemap: Add folio_wake_bit()

Convert wake_up_page_bit() to folio_wake_bit().  All callers have a folio,
so use it directly.  Saves 66 bytes of text in end_page_private_2().

Signed-off-by: Matthew Wilcox (Oracle) <willy@infradead.org>
Reviewed-by: Christoph Hellwig <hch@lst.de>
Acked-by: Jeff Layton <jlayton@kernel.org>
Acked-by: Kirill A. Shutemov <kirill.shutemov@linux.intel.com>
Acked-by: Vlastimil Babka <vbabka@suse.cz>
Reviewed-by: William Kucharski <william.kucharski@oracle.com>
Reviewed-by: David Howells <dhowells@redhat.com>
Acked-by: Mike Rapoport <rppt@linux.ibm.com>

mm/filemap: Add folio_wait_bit()

Rename wait_on_page_bit() to folio_wait_bit().  We must always wait on
the folio, otherwise we won't be woken up due to the tail page hashing
to a different bucket from the head page.

This commit shrinks the kernel by 770 bytes, mostly due to moving
the page waitqueue lookup into folio_wait_bit_common().

Signed-off-by: Matthew Wilcox (Oracle) <willy@infradead.org>
Reviewed-by: Christoph Hellwig <hch@lst.de>
Acked-by: Jeff Layton <jlayton@kernel.org>
Acked-by: Kirill A. Shutemov <kirill.shutemov@linux.intel.com>
Acked-by: Vlastimil Babka <vbabka@suse.cz>
Reviewed-by: William Kucharski <william.kucharski@oracle.com>
Reviewed-by: David Howells <dhowells@redhat.com>
Acked-by: Mike Rapoport <rppt@linux.ibm.com>

mm/writeback: Add folio_wait_stable()

Move wait_for_stable_page() into the folio compatibility file.
folio_wait_stable() avoids a call to compound_head() and is 14 bytes
smaller than wait_for_stable_page() was.  The net text size grows by 16
bytes as a result of this patch.  We can also remove thp_head() as this
was the last user.

Signed-off-by: Matthew Wilcox (Oracle) <willy@infradead.org>
Reviewed-by: Christoph Hellwig <hch@lst.de>
Acked-by: Jeff Layton <jlayton@kernel.org>
Acked-by: Kirill A. Shutemov <kirill.shutemov@linux.intel.com>
Acked-by: Vlastimil Babka <vbabka@suse.cz>
Reviewed-by: William Kucharski <william.kucharski@oracle.com>
Reviewed-by: David Howells <dhowells@redhat.com>

mm/writeback: Add folio_wait_writeback()

wait_on_page_writeback_killable() only has one caller, so convert it to
call folio_wait_writeback_killable().  For the wait_on_page_writeback()
callers, add a compatibility wrapper around folio_wait_writeback().

Turning PageWriteback() into folio_test_writeback() eliminates a call
to compound_head() which saves 8 bytes and 15 bytes in the two
functions.  Unfortunately, that is more than offset by adding the
wait_on_page_writeback compatibility wrapper for a net increase in text
of 7 bytes.

Signed-off-by: Matthew Wilcox (Oracle) <willy@infradead.org>
Reviewed-by: Christoph Hellwig <hch@lst.de>
Acked-by: Jeff Layton <jlayton@kernel.org>
Acked-by: Kirill A. Shutemov <kirill.shutemov@linux.intel.com>
Acked-by: Vlastimil Babka <vbabka@suse.cz>
Reviewed-by: William Kucharski <william.kucharski@oracle.com>
Acked-by: Mike Rapoport <rppt@linux.ibm.com>
Reviewed-by: David Howells <dhowells@redhat.com>

mm/filemap: Add folio_end_writeback()

Add an end_page_writeback() wrapper function for users that are not yet
converted to folios.

folio_end_writeback() is less than half the size of end_page_writeback()
at just 105 bytes compared to 228 bytes, due to removing all the
compound_head() calls.  The 30 byte wrapper function makes this a net
saving of 93 bytes.

Signed-off-by: Matthew Wilcox (Oracle) <willy@infradead.org>
Reviewed-by: Christoph Hellwig <hch@lst.de>
Acked-by: Jeff Layton <jlayton@kernel.org>
Acked-by: Kirill A. Shutemov <kirill.shutemov@linux.intel.com>
Acked-by: Vlastimil Babka <vbabka@suse.cz>
Reviewed-by: William Kucharski <william.kucharski@oracle.com>
Reviewed-by: David Howells <dhowells@redhat.com>
Acked-by: Mike Rapoport <rppt@linux.ibm.com>

mm/swap: Add folio_rotate_reclaimable()

Convert rotate_reclaimable_page() to folio_rotate_reclaimable().  This
eliminates all five of the calls to compound_head() in this function,
saving 75 bytes at the cost of adding 15 bytes to its one caller,
end_page_writeback().  We also save 36 bytes from pagevec_move_tail_fn()
due to using folios there.  Net 96 bytes savings.

Also move its declaration to mm/internal.h as it's only used by filemap.c.

Signed-off-by: Matthew Wilcox (Oracle) <willy@infradead.org>
Acked-by: Vlastimil Babka <vbabka@suse.cz>
Reviewed-by: William Kucharski <william.kucharski@oracle.com>
Reviewed-by: Christoph Hellwig <hch@lst.de>
Acked-by: Kirill A. Shutemov <kirill.shutemov@linux.intel.com>
Acked-by: Mike Rapoport <rppt@linux.ibm.com>
Reviewed-by: David Howells <dhowells@redhat.com>

mm/filemap: Add __folio_lock_or_retry()

Convert __lock_page_or_retry() to __folio_lock_or_retry().  This actually
saves 4 bytes in the only caller of lock_page_or_retry() (due to better
register allocation) and saves the 14 byte cost of calling page_folio()
in __folio_lock_or_retry() for a total saving of 18 bytes.  Also use
a bool for the return type.

Signed-off-by: Matthew Wilcox (Oracle) <willy@infradead.org>
Reviewed-by: Christoph Hellwig <hch@lst.de>
Acked-by: Jeff Layton <jlayton@kernel.org>
Acked-by: Kirill A. Shutemov <kirill.shutemov@linux.intel.com>
Reviewed-by: William Kucharski <william.kucharski@oracle.com>
Acked-by: Mike Rapoport <rppt@linux.ibm.com>
Reviewed-by: David Howells <dhowells@redhat.com>
Acked-by: Vlastimil Babka <vbabka@suse.cz>

mm/filemap: Add folio_wait_locked()

Also add folio_wait_locked_killable().  Turn wait_on_page_locked() and
wait_on_page_locked_killable() into wrappers.  This eliminates a call
to compound_head() from each call-site, reducing text size by 193 bytes
for me.

Signed-off-by: Matthew Wilcox (Oracle) <willy@infradead.org>
Reviewed-by: Christoph Hellwig <hch@lst.de>
Acked-by: Jeff Layton <jlayton@kernel.org>
Acked-by: Kirill A. Shutemov <kirill.shutemov@linux.intel.com>
Acked-by: Vlastimil Babka <vbabka@suse.cz>
Reviewed-by: William Kucharski <william.kucharski@oracle.com>
Reviewed-by: David Howells <dhowells@redhat.com>
Acked-by: Mike Rapoport <rppt@linux.ibm.com>

mm/filemap: Add __folio_lock_async()

There aren't any actual callers of lock_page_async(), so remove it.
Convert filemap_update_page() to call __folio_lock_async().

__folio_lock_async() is 21 bytes smaller than __lock_page_async(),
but the real savings come from using a folio in filemap_update_page(),
shrinking it from 515 bytes to 404 bytes, saving 110 bytes.  The text
shrinks by 132 bytes in total.

Signed-off-by: Matthew Wilcox (Oracle) <willy@infradead.org>
Reviewed-by: Christoph Hellwig <hch@lst.de>
Acked-by: Jeff Layton <jlayton@kernel.org>
Acked-by: Kirill A. Shutemov <kirill.shutemov@linux.intel.com>
Acked-by: Vlastimil Babka <vbabka@suse.cz>
Reviewed-by: William Kucharski <william.kucharski@oracle.com>
Reviewed-by: David Howells <dhowells@redhat.com>
Acked-by: Mike Rapoport <rppt@linux.ibm.com>

mm/filemap: Add folio_lock_killable()

This is like lock_page_killable() but for use by callers who
know they have a folio.  Convert __lock_page_killable() to be
__folio_lock_killable().  This saves one call to compound_head() per
contended call to lock_page_killable().

__folio_lock_killable() is 19 bytes smaller than __lock_page_killable()
was.  filemap_fault() shrinks by 74 bytes and __lock_page_or_retry()
shrinks by 71 bytes.  That's a total of 164 bytes of text saved.

Signed-off-by: Matthew Wilcox (Oracle) <willy@infradead.org>
Reviewed-by: Christoph Hellwig <hch@lst.de>
Acked-by: Jeff Layton <jlayton@kernel.org>
Acked-by: Kirill A. Shutemov <kirill.shutemov@linux.intel.com>
Acked-by: Vlastimil Babka <vbabka@suse.cz>
Reviewed-by: William Kucharski <william.kucharski@oracle.com>
Acked-by: Mike Rapoport <rppt@linux.ibm.com>
Reviewed-by: David Howells <dhowells@redhat.com>

mm/filemap: Add folio_lock()

This is like lock_page() but for use by callers who know they have a folio.
Convert __lock_page() to be __folio_lock().  This saves one call to
compound_head() per contended call to lock_page().

Saves 455 bytes of text; mostly from improved register allocation and
inlining decisions.  __folio_lock is 59 bytes while __lock_page was 79.

Signed-off-by: Matthew Wilcox (Oracle) <willy@infradead.org>
Reviewed-by: Christoph Hellwig <hch@lst.de>
Acked-by: Jeff Layton <jlayton@kernel.org>
Acked-by: Kirill A. Shutemov <kirill.shutemov@linux.intel.com>
Acked-by: Vlastimil Babka <vbabka@suse.cz>
Reviewed-by: William Kucharski <william.kucharski@oracle.com>
Reviewed-by: David Howells <dhowells@redhat.com>
Acked-by: Mike Rapoport <rppt@linux.ibm.com>

mm/filemap: Add folio_unlock()

Convert unlock_page() to call folio_unlock().  By using a folio we
avoid a call to compound_head().  This shortens the function from 39
bytes to 25 and removes 4 instructions on x86-64.  Because we still
have unlock_page(), it's a net increase of 16 bytes of text for the
kernel as a whole, but any path that uses folio_unlock() will execute
4 fewer instructions.

Signed-off-by: Matthew Wilcox (Oracle) <willy@infradead.org>
Reviewed-by: Christoph Hellwig <hch@lst.de>
Acked-by: Jeff Layton <jlayton@kernel.org>
Acked-by: Kirill A. Shutemov <kirill.shutemov@linux.intel.com>
Reviewed-by: William Kucharski <william.kucharski@oracle.com>
Reviewed-by: David Howells <dhowells@redhat.com>
Acked-by: Mike Rapoport <rppt@linux.ibm.com>
Acked-by: Vlastimil Babka <vbabka@suse.cz>

mm/util: Add folio_mapping() and folio_file_mapping()

These are the folio equivalent of page_mapping() and page_file_mapping().
Add an out-of-line page_mapping() wrapper around folio_mapping()
in order to prevent the page_folio() call from bloating every caller
of page_mapping().  Adjust page_file_mapping() and page_mapping_file()
to use folios internally.  Rename __page_file_mapping() to
swapcache_mapping() and change it to take a folio.

This ends up saving 122 bytes of text overall.  folio_mapping() is
45 bytes shorter than page_mapping() was, but the new page_mapping()
wrapper is 30 bytes.  The major reduction is a few bytes less in dozens
of nfs functions (which call page_file_mapping()).  Most of these appear
to be a slight change in gcc's register allocation decisions, which allow:

   48 8b 56 08         mov    0x8(%rsi),%rdx
   48 8d 42 ff         lea    -0x1(%rdx),%rax
   83 e2 01            and    $0x1,%edx
   48 0f 44 c6         cmove  %rsi,%rax

to become:

   48 8b 46 08         mov    0x8(%rsi),%rax
   48 8d 78 ff         lea    -0x1(%rax),%rdi
   a8 01               test   $0x1,%al
   48 0f 44 fe         cmove  %rsi,%rdi

for a reduction of a single byte.  Once the NFS client is converted to
use folios, this entire sequence will disappear.

Also add folio_mapping() documentation.

Signed-off-by: Matthew Wilcox (Oracle) <willy@infradead.org>
Reviewed-by: Christoph Hellwig <hch@lst.de>
Acked-by: Jeff Layton <jlayton@kernel.org>
Acked-by: Kirill A. Shutemov <kirill.shutemov@linux.intel.com>
Acked-by: Vlastimil Babka <vbabka@suse.cz>
Reviewed-by: William Kucharski <william.kucharski@oracle.com>
Reviewed-by: David Howells <dhowells@redhat.com>

mm/filemap: Add folio_pos() and folio_file_pos()

These are just wrappers around page_offset() and page_file_offset()
respectively.  No change to generated code.

Signed-off-by: Matthew Wilcox (Oracle) <willy@infradead.org>
Reviewed-by: Christoph Hellwig <hch@lst.de>
Acked-by: Jeff Layton <jlayton@kernel.org>
Acked-by: Kirill A. Shutemov <kirill.shutemov@linux.intel.com>
Acked-by: Vlastimil Babka <vbabka@suse.cz>
Reviewed-by: William Kucharski <william.kucharski@oracle.com>
Reviewed-by: David Howells <dhowells@redhat.com>

mm/filemap: Add folio_next_index()

This helper returns the page index of the next folio in the file (ie
the end of this folio, plus one).

No changes to generated code.

Signed-off-by: Matthew Wilcox (Oracle) <willy@infradead.org>
Reviewed-by: Christoph Hellwig <hch@lst.de>
Acked-by: Jeff Layton <jlayton@kernel.org>
Acked-by: Kirill A. Shutemov <kirill.shutemov@linux.intel.com>
Acked-by: Vlastimil Babka <vbabka@suse.cz>
Reviewed-by: William Kucharski <william.kucharski@oracle.com>
Reviewed-by: David Howells <dhowells@redhat.com>
Acked-by: Mike Rapoport <rppt@linux.ibm.com>

mm/filemap: Add folio_index(), folio_file_page() and folio_contains()

folio_index() is the equivalent of page_index() for folios.
folio_file_page() is the equivalent of find_subpage().
folio_contains() is the equivalent of thp_contains().

No changes to generated code.

Signed-off-by: Matthew Wilcox (Oracle) <willy@infradead.org>
Reviewed-by: Christoph Hellwig <hch@lst.de>
Acked-by: Jeff Layton <jlayton@kernel.org>
Acked-by: Kirill A. Shutemov <kirill.shutemov@linux.intel.com>
Acked-by: Vlastimil Babka <vbabka@suse.cz>
Reviewed-by: William Kucharski <william.kucharski@oracle.com>
Reviewed-by: David Howells <dhowells@redhat.com>
Acked-by: Mike Rapoport <rppt@linux.ibm.com>

mm: Handle per-folio private data

Add folio_get_private() which mirrors page_private() -- ie folio private
data is the same as page private data.  The only difference is that these
return a void * instead of an unsigned long, which matches the majority
of users.

Turn attach_page_private() into folio_attach_private() and reimplement
attach_page_private() as a wrapper.  No filesystem which uses page private
data currently supports compound pages, so we're free to define the rules.
attach_page_private() may only be called on a head page; if you want
to add private data to a tail page, you can call set_page_private()
directly (and shouldn't increment the page refcount!  That should be
done when adding private data to the head page / folio).

This saves 813 bytes of text with the distro-derived config that I'm
testing due to removing the calls to compound_head() in get_page()
& put_page().

Signed-off-by: Matthew Wilcox (Oracle) <willy@infradead.org>
Reviewed-by: Christoph Hellwig <hch@lst.de>
Acked-by: Jeff Layton <jlayton@kernel.org>
Acked-by: Kirill A. Shutemov <kirill.shutemov@linux.intel.com>
Acked-by: Vlastimil Babka <vbabka@suse.cz>
Reviewed-by: William Kucharski <william.kucharski@oracle.com>
Reviewed-by: David Howells <dhowells@redhat.com>
Acked-by: Mike Rapoport <rppt@linux.ibm.com>

mm/lru: Add folio LRU functions

Handle arbitrary-order folios being added to the LRU.  By definition,
all pages being added to the LRU were already head or base pages, but
call page_folio() on them anyway to get the type right and avoid the
buried calls to compound_head().

Saves 783 bytes of kernel text; no functions grow.

Signed-off-by: Matthew Wilcox (Oracle) <willy@infradead.org>
Reviewed-by: Yu Zhao <yuzhao@google.com>
Reviewed-by: Christoph Hellwig <hch@lst.de>
Reviewed-by: David Howells <dhowells@redhat.com>
Acked-by: Kirill A. Shutemov <kirill.shutemov@linux.intel.com>
Acked-by: Mike Rapoport <rppt@linux.ibm.com>
Acked-by: Vlastimil Babka <vbabka@suse.cz>

mm: Add folio flag manipulation functions

These new functions are the folio analogues of the various PageFlags
functions.  If CONFIG_DEBUG_VM_PGFLAGS is enabled, we check the folio
is not a tail page at every invocation.  This will also catch the
PagePoisoned case as a poisoned page has every bit set, which would
include PageTail.

This saves 1684 bytes of text with the distro-derived config that
I'm testing due to removing a double call to compound_head() in
PageSwapCache().

Signed-off-by: Matthew Wilcox (Oracle) <willy@infradead.org>
Reviewed-by: Christoph Hellwig <hch@lst.de>
Acked-by: Jeff Layton <jlayton@kernel.org>
Acked-by: Kirill A. Shutemov <kirill.shutemov@linux.intel.com>
Acked-by: Vlastimil Babka <vbabka@suse.cz>
Reviewed-by: William Kucharski <william.kucharski@oracle.com>
Reviewed-by: David Howells <dhowells@redhat.com>
Acked-by: Mike Rapoport <rppt@linux.ibm.com>

mm: Add folio_try_get_rcu()

This is the equivalent of page_cache_get_speculative().  Also add
folio_ref_try_add_rcu (the equivalent of page_cache_add_speculative)
and folio_get_unless_zero() (the equivalent of get_page_unless_zero()).

The new kernel-doc attempts to explain from the user's point of view
when to use folio_try_get_rcu() and when to use folio_get_unless_zero(),
because there seems to be some confusion currently between the users of
page_cache_get_speculative() and get_page_unless_zero().

Reimplement page_cache_add_speculative() and page_cache_get_speculative()
as wrappers around the folio equivalents, but leave get_page_unless_zero()
alone for now.  This commit reduces text size by 3 bytes due to slightly
different register allocation & instruction selections.

Signed-off-by: Matthew Wilcox (Oracle) <willy@infradead.org>
Acked-by: Vlastimil Babka <vbabka@suse.cz>
Reviewed-by: William Kucharski <william.kucharski@oracle.com>
Reviewed-by: Christoph Hellwig <hch@lst.de>
Acked-by: Kirill A. Shutemov <kirill.shutemov@linux.intel.com>
Acked-by: Mike Rapoport <rppt@linux.ibm.com>

mm: Add folio_get()

If we know we have a folio, we can call folio_get() instead
of get_page() and save the overhead of calling compound_head().
No change to generated code.

Signed-off-by: Matthew Wilcox (Oracle) <willy@infradead.org>
Reviewed-by: Zi Yan <ziy@nvidia.com>
Reviewed-by: Christoph Hellwig <hch@lst.de>
Acked-by: Jeff Layton <jlayton@kernel.org>
Acked-by: Kirill A. Shutemov <kirill.shutemov@linux.intel.com>
Acked-by: Vlastimil Babka <vbabka@suse.cz>
Reviewed-by: William Kucharski <william.kucharski@oracle.com>
Reviewed-by: David Howells <dhowells@redhat.com>
Acked-by: Mike Rapoport <rppt@linux.ibm.com>

mm: Add folio_put()

If we know we have a folio, we can call folio_put() instead of put_page()
and save the overhead of calling compound_head().  Also skips the
devmap checks.

This commit looks like it should be a no-op, but actually saves 684 bytes
of text with the distro-derived config that I'm testing.  Some functions
grow a little while others shrink.  I presume the compiler is making
different inlining decisions.

Signed-off-by: Matthew Wilcox (Oracle) <willy@infradead.org>
Reviewed-by: Zi Yan <ziy@nvidia.com>
Reviewed-by: Christoph Hellwig <hch@lst.de>
Acked-by: Jeff Layton <jlayton@kernel.org>
Acked-by: Kirill A. Shutemov <kirill.shutemov@linux.intel.com>
Acked-by: Vlastimil Babka <vbabka@suse.cz>
Reviewed-by: William Kucharski <william.kucharski@oracle.com>
Reviewed-by: David Howells <dhowells@redhat.com>
Acked-by: Mike Rapoport <rppt@linux.ibm.com>

mm: Add folio reference count functions

These functions mirror their page reference counterparts.  Also add
the kernel-doc to the mm-api and correct the return type of
page_ref_add_unless() to bool.  No change to generated code.

Signed-off-by: Matthew Wilcox (Oracle) <willy@infradead.org>
Reviewed-by: Christoph Hellwig <hch@lst.de>
Acked-by: Jeff Layton <jlayton@kernel.org>
Acked-by: Kirill A. Shutemov <kirill.shutemov@linux.intel.com>
Acked-by: Vlastimil Babka <vbabka@suse.cz>
Reviewed-by: William Kucharski <william.kucharski@oracle.com>
Reviewed-by: David Howells <dhowells@redhat.com>
Acked-by: Mike Rapoport <rppt@linux.ibm.com>

mm/debug: Add VM_BUG_ON_FOLIO() and VM_WARN_ON_ONCE_FOLIO()

These are the folio equivalents of VM_BUG_ON_PAGE and
VM_WARN_ON_ONCE_PAGE.  No change to generated code.

Signed-off-by: Matthew Wilcox (Oracle) <willy@infradead.org>
Reviewed-by: Zi Yan <ziy@nvidia.com>
Reviewed-by: Christoph Hellwig <hch@lst.de>
Acked-by: Jeff Layton <jlayton@kernel.org>
Acked-by: Kirill A. Shutemov <kirill.shutemov@linux.intel.com>
Acked-by: Vlastimil Babka <vbabka@suse.cz>
Reviewed-by: William Kucharski <william.kucharski@oracle.com>
Reviewed-by: David Howells <dhowells@redhat.com>
Acked-by: Mike Rapoport <rppt@linux.ibm.com>

mm/vmstat: Add functions to account folio statistics

Allow page counters to be more readily modified by callers which have
a folio.  Name these wrappers with 'stat' instead of 'state' as requested
by Linus here:
https://lore.kernel.org/linux-mm/CAHk-=wj847SudR-kt+46fT3+xFFgiwpgThvm7DJWGdi4cVrbnQ@mail.gmail.com/
No change to generated code.

Signed-off-by: Matthew Wilcox (Oracle) <willy@infradead.org>
Reviewed-by: Christoph Hellwig <hch@lst.de>
Acked-by: Jeff Layton <jlayton@kernel.org>
Acked-by: Kirill A. Shutemov <kirill.shutemov@linux.intel.com>
Acked-by: Vlastimil Babka <vbabka@suse.cz>
Reviewed-by: William Kucharski <william.kucharski@oracle.com>
Reviewed-by: David Howells <dhowells@redhat.com>
Acked-by: Mike Rapoport <rppt@linux.ibm.com>

mm: Add folio_pgdat(), folio_zone() and folio_zonenum()

These are just convenience wrappers for callers with folios; pgdat and
zone can be reached from tail pages as well as head pages.  No change
to generated code.

Signed-off-by: Matthew Wilcox (Oracle) <willy@infradead.org>
Reviewed-by: Zi Yan <ziy@nvidia.com>
Reviewed-by: Christoph Hellwig <hch@lst.de>
Acked-by: Jeff Layton <jlayton@kernel.org>
Acked-by: Kirill A. Shutemov <kirill.shutemov@linux.intel.com>
Acked-by: Vlastimil Babka <vbabka@suse.cz>
Reviewed-by: William Kucharski <william.kucharski@oracle.com>
Reviewed-by: David Howells <dhowells@redhat.com>
Acked-by: Mike Rapoport <rppt@linux.ibm.com>

mm: Introduce struct folio

A struct folio is a new abstraction to replace the venerable struct page.
A function which takes a struct folio argument declares that it will
operate on the entire (possibly compound) page, not just PAGE_SIZE bytes.
In return, the caller guarantees that the pointer it is passing does
not point to a tail page.  No change to generated code.

Signed-off-by: Matthew Wilcox (Oracle) <willy@infradead.org>
Acked-by: Jeff Layton <jlayton@kernel.org>
Acked-by: Kirill A. Shutemov <kirill.shutemov@linux.intel.com>
Acked-by: Vlastimil Babka <vbabka@suse.cz>
Reviewed-by: William Kucharski <william.kucharski@oracle.com>
Reviewed-by: Christoph Hellwig <hch@lst.de>
Reviewed-by: David Howells <dhowells@redhat.com>
Acked-by: Mike Rapoport <rppt@linux.ibm.com>

mm: Convert get_page_unless_zero() to return bool

atomic_add_unless() returns bool, so remove the widening casts to int
in page_ref_add_unless() and get_page_unless_zero().  This causes gcc
to produce slightly larger code in isolate_migratepages_block(), but
it's not clear that it's worse code.  Net +19 bytes of text.

Signed-off-by: Matthew Wilcox (Oracle) <willy@infradead.org>
Reviewed-by: Christoph Hellwig <hch@lst.de>
Acked-by: Kirill A. Shutemov <kirill.shutemov@linux.intel.com>
Reviewed-by: David Howells <dhowells@redhat.com>
Acked-by: Vlastimil Babka <vbabka@suse.cz>

Linux 5.15-rc3

Merge tag '5.15-rc2-ksmbd-fixes' of git://git.samba.org/ksmbd

Pull ksmbd fixes from Steve French:
 "Five fixes for the ksmbd kernel server, including three security
  fixes:

   - remove follow symlinks support

   - use LOOKUP_BENEATH to prevent out of share access

   - SMB3 compounding security fix

   - fix for returning the default streams correctly, fixing a bug when
     writing ppt or doc files from some clients

   - logging more clearly that ksmbd is experimental (at module load
     time)"

* tag '5.15-rc2-ksmbd-fixes' of git://git.samba.org/ksmbd:
  ksmbd: use LOOKUP_BENEATH to prevent the out of share access
  ksmbd: remove follow symlinks support
  ksmbd: check protocol id in ksmbd_verify_smb_message()
  ksmbd: add default data stream name in FILE_STREAM_INFORMATION
  ksmbd: log that server is experimental at module load

Merge tag 'edac_urgent_for_v5.15_rc3' of git://git./linux/kernel/git/ras/ras

Pull EDAC fixes from Borislav Petkov:
 "Fix two EDAC drivers using the wrong value type for the DIMM mode"

* tag 'edac_urgent_for_v5.15_rc3' of git://git.kernel.org/pub/scm/linux/kernel/git/ras/ras:
  EDAC/dmc520: Assign the proper type to dimm->edac_mode
  EDAC/synopsys: Fix wrong value type assignment for edac_mode

Merge tag 'thermal-v5.15-rc3' of git://git./linux/kernel/git/thermal/linux

Pull thermal fixes from Daniel Lezcano:

 - Fix thermal shutdown after a suspend/resume due to a wrong TCC value
   restored on Intel platform (Antoine Tenart)

 - Fix potential buffer overflow when building the list of policies. The
   buffer size is not updated after writing to it (Dan Carpenter)

 - Fix wrong check against IS_ERR instead of NULL (Ansuel Smith)

* tag 'thermal-v5.15-rc3' of git://git.kernel.org/pub/scm/linux/kernel/git/thermal/linux:
  thermal/drivers/tsens: Fix wrong check for tzd in irq handlers
  thermal/core: Potential buffer overflow in thermal_build_list_of_policies()
  thermal/drivers/int340x: Do not set a wrong tcc offset on resume

Merge tag 'x86-urgent-2021-09-26' of git://git./linux/kernel/git/tip/tip

Pull x86 fixes from Thomas Gleixner:
 "A set of fixes for X86:

   - Prevent sending the wrong signal when protection keys are enabled
     and the kernel handles a fault in the vsyscall emulation.

   - Invoke early_reserve_memory() before invoking e820_memory_setup()
     which is required to make the Xen dom0 e820 hooks work correctly.

   - Use the correct data type for the SETZ operand in the EMQCMDS
     instruction wrapper.

   - Prevent undefined behaviour to the potential unaligned accesss in
     the instruction decoder library"

* tag 'x86-urgent-2021-09-26' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip:
  x86/insn, tools/x86: Fix undefined behavior due to potential unaligned accesses
  x86/asm: Fix SETZ size enqcmds() build failure
  x86/setup: Call early_reserve_memory() earlier
  x86/fault: Fix wrong signal when vsyscall fails with pkey

Merge tag 'timers-urgent-2021-09-26' of git://git./linux/kernel/git/tip/tip

Pull timer fix from Thomas Gleixner:
 "A single fix for the recently introduced regression in posix CPU
  timers which failed to stop the timer when requested. That caused
  unexpected signals to be sent to the process/thread causing
  malfunction"

* tag 'timers-urgent-2021-09-26' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip:
  posix-cpu-timers: Prevent spuriously armed 0-value itimer

Merge tag 'irq-urgent-2021-09-26' of git://git./linux/kernel/git/tip/tip

Pull irq fixes from Thomas Gleixner:
 "A set of fixes for interrupt chip drivers:

   - Work around a bad GIC integration on a Renesas platform which can't
     handle byte-sized MMIO access

   - Plug a potential memory leak in the GICv4 driver

   - Fix a regression in the Armada 370-XP IPI code which was caused by
     issuing EOI instack of ACK.

   - A couple of small fixes here and there"

* tag 'irq-urgent-2021-09-26' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip:
  irqchip/gic: Work around broken Renesas integration
  irqchip/renesas-rza1: Use semicolons instead of commas
  irqchip/gic-v3-its: Fix potential VPE leak on error
  irqchip/goldfish-pic: Select GENERIC_IRQ_CHIP to fix build
  irqchip/mbigen: Repair non-kernel-doc notation
  irqdomain: Change the type of 'size' in __irq_domain_add() to be consistent
  irqchip/armada-370-xp: Fix ack/eoi breakage
  Documentation: Fix irq-domain.rst build warning

Merge branch 'akpm' (patches from Andrew)

Merge misc fixes from Andrew Morton:
 "16 patches.

  Subsystems affected by this patch series: xtensa, sh, ocfs2, scripts,
  lib, and mm (memory-failure, kasan, damon, shmem, tools, pagecache,
  debug, and pagemap)"

* emailed patches from Andrew Morton <akpm@linux-foundation.org>:
  mm: fix uninitialized use in overcommit_policy_handler
  mm/memory_failure: fix the missing pte_unmap() call
  kasan: always respect CONFIG_KASAN_STACK
  sh: pgtable-3level: fix cast to pointer from integer of different size
  mm/debug: sync up latest migrate_reason to migrate_reason_names
  mm/debug: sync up MR_CONTIG_RANGE and MR_LONGTERM_PIN
  mm: fs: invalidate bh_lrus for only cold path
  lib/zlib_inflate/inffast: check config in C to avoid unused function warning
  tools/vm/page-types: remove dependency on opt_file for idle page tracking
  scripts/sorttable: riscv: fix undeclared identifier 'EM_RISCV' error
  ocfs2: drop acl cache for directories too
  mm/shmem.c: fix judgment error in shmem_is_huge()
  xtensa: increase size of gcc stack frame check
  mm/damon: don't use strnlen() with known-bogus source length
  kasan: fix Kconfig check of CC_HAS_WORKING_NOSANITIZE_ADDRESS
  mm, hwpoison: add is_free_buddy_page() in HWPoisonHandlable()

Merge tag 'scsi-fixes' of git://git./linux/kernel/git/jejb/scsi

Pull SCSI fixes from James Bottomley:
 "Thirty-three fixes, I'm afraid.

  Essentially the build up from the last couple of weeks while I've been
  dealling with Linux Plumbers conference infrastructure issues. It's
  mostly the usual assortment of spelling fixes and minor corrections.

  The only core relevant changes are to the sd driver to reduce the spin
  up message spew and fix a small memory leak on the freeing path"

* tag 'scsi-fixes' of git://git.kernel.org/pub/scm/linux/kernel/git/jejb/scsi: (33 commits)
  scsi: ses: Retry failed Send/Receive Diagnostic commands
  scsi: target: Fix spelling mistake "CONFLIFT" -> "CONFLICT"
  scsi: lpfc: Fix gcc -Wstringop-overread warning, again
  scsi: lpfc: Use correct scnprintf() limit
  scsi: lpfc: Fix sprintf() overflow in lpfc_display_fpin_wwpn()
  scsi: core: Remove 'current_tag'
  scsi: acornscsi: Remove tagged queuing vestiges
  scsi: fas216: Kill scmd->tag
  scsi: qla2xxx: Restore initiator in dual mode
  scsi: ufs: core: Unbreak the reset handler
  scsi: sd_zbc: Support disks with more than 2**32 logical blocks
  scsi: ufs: core: Revert "scsi: ufs: Synchronize SCSI and UFS error handling"
  scsi: bsg: Fix device unregistration
  scsi: sd: Make sd_spinup_disk() less noisy
  scsi: ufs: ufs-pci: Fix Intel LKF link stability
  scsi: mpt3sas: Clean up some inconsistent indenting
  scsi: megaraid: Clean up some inconsistent indenting
  scsi: sr: Fix spelling mistake "does'nt" -> "doesn't"
  scsi: Remove SCSI CDROM MAINTAINERS entry
  scsi: megaraid: Fix Coccinelle warning
  ...

Merge tag 'io_uring-5.15-2021-09-25' of git://git.kernel.dk/linux-block

Pull io_uring fixes from Jens Axboe:
 "This one looks a bit bigger than it is, but that's mainly because 2/3
  of it is enabling IORING_OP_CLOSE to close direct file descriptors.

  We've had a few folks using them and finding it confusing that the way
  to close them is through using -1 for file update, this just brings
  API symmetry for direct descriptors. Hence I think we should just do
  this now and have a better API for 5.15 release. There's some room for
  de-duplicating the close code, but we're leaving that for the next
  merge window.

  Outside of that, just small fixes:

   - Poll race fixes (Hao)

   - io-wq core dump exit fix (me)

   - Reschedule around potentially intensive tctx and buffer iterators
     on teardown (me)

   - Fix for always ending up punting files update to io-wq (me)

   - Put the provided buffer meta data under memcg accounting (me)

   - Tweak for io_write(), removing dead code that was added with the
     iterator changes in this release (Pavel)"

* tag 'io_uring-5.15-2021-09-25' of git://git.kernel.dk/linux-block:
  io_uring: make OP_CLOSE consistent with direct open
  io_uring: kill extra checks in io_write()
  io_uring: don't punt files update to io-wq unconditionally
  io_uring: put provided buffer meta data under memcg accounting
  io_uring: allow conditional reschedule for intensive iterators
  io_uring: fix potential req refcount underflow
  io_uring: fix missing set of EPOLLONESHOT for CQ ring overflow
  io_uring: fix race between poll completion and cancel_hash insertion
  io-wq: ensure we exit if thread group is exiting

Merge tag 'block-5.15-2021-09-25' of git://git.kernel.dk/linux-block

Pull block fixes from Jens Axboe:

 - NVMe pull request via Christoph:
      - keep ctrl->namespaces ordered (Christoph Hellwig)
      - fix incorrect h2cdata pdu offset accounting in nvme-tcp (Sagi
        Grimberg)
      - handled updated hw_queues in nvme-fc more carefully (Daniel
        Wagner, James Smart)

 - md lock order fix (Christoph)

 - fallocate locking fix (Ming)

 - blktrace UAF fix (Zhihao)

 - rq-qos bio tracking fix (Ming)

* tag 'block-5.15-2021-09-25' of git://git.kernel.dk/linux-block:
  block: hold ->invalidate_lock in blkdev_fallocate
  blktrace: Fix uaf in blk_trace access after removing by sysfs
  block: don't call rq_qos_ops->done_bio if the bio isn't tracked
  md: fix a lock order reversal in md_alloc
  nvme: keep ctrl->namespaces ordered
  nvme-tcp: fix incorrect h2cdata pdu offset accounting
  nvme-fc: remove freeze/unfreeze around update_nr_hw_queues
  nvme-fc: avoid race between time out and tear down
  nvme-fc: update hardware queues before using them

Merge tag 'for-linus-5.15b-rc3-tag' of git://git./linux/kernel/git/xen/tip

Pull xen fixes from Juergen Gross:
 "Some minor cleanups and fixes of some theoretical bugs, as well as a
  fix of a bug introduced in 5.15-rc1"

* tag 'for-linus-5.15b-rc3-tag' of git://git.kernel.org/pub/scm/linux/kernel/git/xen/tip:
  xen/x86: fix PV trap handling on secondary processors
  xen/balloon: fix balloon kthread freezing
  swiotlb-xen: this is PV-only on x86
  xen/pci-swiotlb: reduce visibility of symbols
  PCI: only build xen-pcifront in PV-enabled environments
  swiotlb-xen: ensure to issue well-formed XENMEM_exchange requests
  Xen/gntdev: don't ignore kernel unmapping error
  xen/x86: drop redundant zeroing from cpu_initialize_context()

Merge tag 'linux-kselftest-fixes-5.15-rc3' of git://git./linux/kernel/git/shuah/linux-kselftest

Pull Kselftest fixes from Shuah Khan:

 - fix to Kselftest common framework header install to run before other
   targets for it work correctly in parallel build case.

 - fixes to kvm test to not ignore fscanf() returns which could result
   in inconsistent test behavior and failures.

* tag 'linux-kselftest-fixes-5.15-rc3' of git://git.kernel.org/pub/scm/linux/kernel/git/shuah/linux-kselftest:
  selftests: kvm: fix get_run_delay() ignoring fscanf() return warn
  selftests: kvm: move get_run_delay() into lib/test_util
  selftests:kvm: fix get_trans_hugepagesz() ignoring fscanf() return warn
  selftests:kvm: fix get_warnings_count() ignoring fscanf() return warn
  selftests: be sure to make khdr before other targets

Merge tag 'erofs-for-5.15-rc3-fixes' of git://git./linux/kernel/git/xiang/erofs

Pull erofs fixes from Gao Xiang:
 "Two bugfixes to fix the 4KiB blockmap chunk format availability and a
  dangling pointer usage. There is also a trivial cleanup to clarify
  compacted_2b if compacted_4b_initial > totalidx.

  Summary:

   - fix the dangling pointer use in erofs_lookup tracepoint

   - fix unsupported chunk format check

   - zero out compacted_2b if compacted_4b_initial > totalidx"

* tag 'erofs-for-5.15-rc3-fixes' of git://git.kernel.org/pub/scm/linux/kernel/git/xiang/erofs:
  erofs: clear compacted_2b if compacted_4b_initial > totalidx
  erofs: fix misbehavior of unsupported chunk format check
  erofs: fix up erofs_lookup tracepoint

Merge tag '5.15-rc2-smb3-fixes' of git://git.samba.org/sfrench/cifs-2.6

Pull cifs fixes from Steve French:
 "Six small cifs/smb3 fixes, two for stable:

   - important fix for deferred close (found by a git functional test)
     related to attribute caching on close.

   - four (two cosmetic, two more serious) small fixes for problems
     pointed out by smatch via Dan Carpenter

   - fix for comment formatting problems pointed out by W=1"

* tag '5.15-rc2-smb3-fixes' of git://git.samba.org/sfrench/cifs-2.6:
  cifs: fix incorrect check for null pointer in header_assemble
  smb3: correct server pointer dereferencing check to be more consistent
  smb3: correct smb3 ACL security descriptor
  cifs: Clear modified attribute bit from inode flags
  cifs: Deal with some warnings from W=1
  cifs: fix a sign extension bug

Merge tag 'char-misc-5.15-rc3' of git://git./linux/kernel/git/gregkh/char-misc

Pull char/misc driver fixes from Greg KH:
 "Here are some small char and misc driver fixes for 5.15-rc3.

  Nothing huge in here, just fixes for a number of small issues that
  have been reported. These include:

   - habanalabs race conditions and other bugs fixed

   - binder driver fixes

   - fpga driver fixes

   - coresight build warning fix

   - nvmem driver fix

   - comedi memory leak fix

   - bcm-vk tty race fix

   - other tiny driver fixes

  All of these have been in linux-next for a while with no reported
  issues"

* tag 'char-misc-5.15-rc3' of git://git.kernel.org/pub/scm/linux/kernel/git/gregkh/char-misc: (21 commits)
  comedi: Fix memory leak in compat_insnlist()
  nvmem: NVMEM_NINTENDO_OTP should depend on WII
  misc: bcm-vk: fix tty registration race
  fpga: dfl: Avoid reads to AFU CSRs during enumeration
  fpga: machxo2-spi: Fix missing error code in machxo2_write_complete()
  fpga: machxo2-spi: Return an error on failure
  habanalabs: expose a single cs seq in staged submissions
  habanalabs: fix wait offset handling
  habanalabs: rate limit multi CS completion errors
  habanalabs/gaudi: fix LBW RR configuration
  habanalabs: Fix spelling mistake "FEADBACK" -> "FEEDBACK"
  habanalabs: fail collective wait when not supported
  habanalabs/gaudi: use direct MSI in single mode
  habanalabs: fix kernel OOPs related to staged cs
  habanalabs: fix potential race in interrupt wait ioctl
  mcb: fix error handling in mcb_alloc_bus()
  misc: genwqe: Fixes DMA mask setting
  coresight: syscfg: Fix compiler warning
  nvmem: core: Add stubs for nvmem_cell_read_variable_le_u32/64 if !CONFIG_NVMEM
  binder: make sure fd closes complete
  ...

Merge tag 'staging-5.15-rc3' of git://git./linux/kernel/git/gregkh/staging

Pull staging driver fixes from Greg KH:
 "Here are two small staging driver fixes for 5.15-rc3:

   - greybus tty use-after-free bugfix

   - r8188eu ioctl overlap build warning fix

  Note, the r8188eu ioctl has been entirely removed for 5.16-rc1, but
  it's good to get this fixed now for people using this in 5.15.

  Both of these have been in linux-next for a while with no reported
  issues"

* tag 'staging-5.15-rc3' of git://git.kernel.org/pub/scm/linux/kernel/git/gregkh/staging:
  staging: r8188eu: fix -Wrestrict warnings
  staging: greybus: uart: fix tty use after free

Merge tag 'tty-5.15-rc3' of git://git./linux/kernel/git/gregkh/tty

Pull tty/serial fixes from Greg KH:
 "Here are four small tty/serial driver fixes for 5.15-rc3. They
  include:

   - remove an export now that no one is using it anymore

   - mvebu-uart tx_empty callback fix

   - 8250_omap bugfix

   - synclink_gt build fix

  All of these have been in linux-next for a while with no reported
  issues"

* tag 'tty-5.15-rc3' of git://git.kernel.org/pub/scm/linux/kernel/git/gregkh/tty:
  tty: unexport tty_ldisc_release
  tty: synclink_gt: rename a conflicting function name
  serial: mvebu-uart: fix driver's tx_empty callback
  serial: 8250: 8250_omap: Fix RX_LVL register offset

Merge tag 'usb-5.15-rc3' of git://git./linux/kernel/git/gregkh/usb

Pull USB driver fixes from Greg KH:
 "Here are some USB driver fixes and new device ids for 5.15-rc3.

  They include:

   - usb-storage quirk additions

   - usb-serial new device ids

   - usb-serial driver fixes

   - USB roothub registration bugfix to resolve a long-reported issue

   - usb gadget driver fixes for a large number of small things

   - dwc2 driver fixes

  All of these have been in linux-next for a while with no reported
  issues"

* tag 'usb-5.15-rc3' of git://git.kernel.org/pub/scm/linux/kernel/git/gregkh/usb: (28 commits)
  USB: serial: option: add device id for Foxconn T99W265
  USB: serial: cp210x: add ID for GW Instek GDM-834x Digital Multimeter
  USB: serial: cp210x: add part-number debug printk
  USB: serial: cp210x: fix dropped characters with CP2102
  MAINTAINERS: usb, update Peter Korsgaard's entries
  usb: musb: tusb6010: uninitialized data in tusb_fifo_write_unaligned()
  usb-storage: Add quirk for ScanLogic SL11R-IDE older than 2.6c
  Re-enable UAS for LaCie Rugged USB3-FW with fk quirk
  USB: serial: option: remove duplicate USB device ID
  USB: serial: mos7840: remove duplicated 0xac24 device ID
  arm64: dts: qcom: ipq8074: remove USB tx-fifo-resize property
  usb: gadget: f_uac2: Populate SS descriptors' wBytesPerInterval
  usb: gadget: f_uac2: Add missing companion descriptor for feedback EP
  usb: dwc2: gadget: Fix ISOC transfer complete handling for DDMA
  usb: core: hcd: Modularize HCD stop configuration in usb_stop_hcd()
  xhci: Set HCD flag to defer primary roothub registration
  usb: core: hcd: Add support for deferring roothub registration
  usb: dwc2: gadget: Fix ISOC flow for BDMA and Slave
  usb: dwc3: core: balance phy init and exit
  Revert "USB: bcma: Add a check for devm_gpiod_get"
  ...

ksmbd: use LOOKUP_BENEATH to prevent the out of share access

instead of removing '..' in a given path, call
kern_path with LOOKUP_BENEATH flag to prevent
the out of share access.

ran various test on this:
smb2-cat-async smb://127.0.0.1/homes/../out_of_share
smb2-cat-async smb://127.0.0.1/homes/foo/../../out_of_share
smbclient //127.0.0.1/homes -c "mkdir ../foo2"
smbclient //127.0.0.1/homes -c "rename bar ../bar"

Cc: Ronnie Sahlberg <ronniesahlberg@gmail.com>
Cc: Ralph Boehme <slow@samba.org>
Tested-by: Steve French <smfrench@gmail.com>
Tested-by: Namjae Jeon <linkinjeon@kernel.org>
Acked-by: Namjae Jeon <linkinjeon@kernel.org>
Signed-off-by: Hyunchul Lee <hyc.lee@gmail.com>
Signed-off-by: Steve French <stfrench@microsoft.com>

mm: fix uninitialized use in overcommit_policy_handler

We get an unexpected value of /proc/sys/vm/overcommit_memory after
running the following program:

  int main()
  {
      int fd = open("/proc/sys/vm/overcommit_memory", O_RDWR);
      write(fd, "1", 1);
      write(fd, "2", 1);
      close(fd);
  }

write(fd, "2", 1) will pass *ppos = 1 to proc_dointvec_minmax.
proc_dointvec_minmax will return 0 without setting new_policy.

  t.data = &new_policy;
  ret = proc_dointvec_minmax(&t, write, buffer, lenp, ppos)
      -->do_proc_dointvec
         -->__do_proc_dointvec
              if (write) {
                if (proc_first_pos_non_zero_ignore(ppos, table))
                  goto out;

  sysctl_overcommit_memory = new_policy;

so sysctl_overcommit_memory will be set to an uninitialized value.

Check whether new_policy has been changed by proc_dointvec_minmax.

Link: https://lkml.kernel.org/r/20210923020524.13289-1-chenjun102@huawei.com
Fixes: 56f3547bfa4d ("mm: adjust vm_committed_as_batch according to vm overcommit policy")
Signed-off-by: Chen Jun <chenjun102@huawei.com>
Acked-by: Michal Hocko <mhocko@suse.com>
Reviewed-by: Feng Tang <feng.tang@intel.com>
Reviewed-by: Kefeng Wang <wangkefeng.wang@huawei.com>
Cc: Rui Xiang <rui.xiang@huawei.com>
Cc: <stable@vger.kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

mm/memory_failure: fix the missing pte_unmap() call

The paired pte_unmap() call is missing before the
dev_pagemap_mapping_shift() returns.  So fix it.

David says:
 "I guess this code never runs on 32bit / highmem, that's why we didn't
  notice so far".

[akpm@linux-foundation.org: cleanup]

Link: https://lkml.kernel.org/r/20210923122642.4999-1-zhengqi.arch@bytedance.com
Signed-off-by: Qi Zheng <zhengqi.arch@bytedance.com>
Reviewed-by: David Hildenbrand <david@redhat.com>
Cc: Naoya Horiguchi <naoya.horiguchi@nec.com>
Cc: Muchun Song <songmuchun@bytedance.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

kasan: always respect CONFIG_KASAN_STACK

Currently, the asan-stack parameter is only passed along if
CFLAGS_KASAN_SHADOW is not empty, which requires KASAN_SHADOW_OFFSET to
be defined in Kconfig so that the value can be checked.  In RISC-V's
case, KASAN_SHADOW_OFFSET is not defined in Kconfig, which means that
asan-stack does not get disabled with clang even when CONFIG_KASAN_STACK
is disabled, resulting in large stack warnings with allmodconfig:

  drivers/video/fbdev/omap2/omapfb/displays/panel-lgphilips-lb035q02.c:117:12: error: stack frame size (14400) exceeds limit (2048) in function 'lb035q02_connect' [-Werror,-Wframe-larger-than]
  static int lb035q02_connect(struct omap_dss_device *dssdev)
             ^
  1 error generated.

Ensure that the value of CONFIG_KASAN_STACK is always passed along to
the compiler so that these warnings do not happen when
CONFIG_KASAN_STACK is disabled.

Link: https://github.com/ClangBuiltLinux/linux/issues/1453
References: 6baec880d7a5 ("kasan: turn off asan-stack for clang-8 and earlier")
Link: https://lkml.kernel.org/r/20210922205525.570068-1-nathan@kernel.org
Signed-off-by: Nathan Chancellor <nathan@kernel.org>
Reviewed-by: Marco Elver <elver@google.com>
Cc: Andrey Ryabinin <ryabinin.a.a@gmail.com>
Cc: Alexander Potapenko <glider@google.com>
Cc: Andrey Konovalov <andreyknvl@gmail.com>
Cc: Dmitry Vyukov <dvyukov@google.com>
Cc: Nick Desaulniers <ndesaulniers@google.com>
Cc: Arnd Bergmann <arnd@arndb.de>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

sh: pgtable-3level: fix cast to pointer from integer of different size

If X2TLB=y (CPU_SHX2=y or CPU_SHX3=y, e.g. migor_defconfig), pgd_t.pgd
is "unsigned long long", causing:

    In file included from arch/sh/include/asm/pgtable.h:13,
     from include/linux/pgtable.h:6,
     from include/linux/mm.h:33,
     from arch/sh/kernel/asm-offsets.c:14:
    arch/sh/include/asm/pgtable-3level.h: In function `pud_pgtable':
    arch/sh/include/asm/pgtable-3level.h:37:9: warning: cast to pointer from integer of different size [-Wint-to-pointer-cast]
       37 |  return (pmd_t *)pud_val(pud);
  |         ^

Fix this by adding an intermediate cast to "unsigned long", which is
basically what the old code did before.

Link: https://lkml.kernel.org/r/2c2eef3c9a2f57e5609100a4864715ccf253d30f.1631713483.git.geert+renesas@glider.be
Fixes: 9cf6fa2458443118 ("mm: rename pud_page_vaddr to pud_pgtable and make it return pmd_t *")
Signed-off-by: Geert Uytterhoeven <geert+renesas@glider.be>
Tested-by: Daniel Palmer <daniel@thingy.jp>
Acked-by: Rob Landley <rob@landley.net>
Cc: Yoshinori Sato <ysato@users.osdn.me>
Cc: Rich Felker <dalias@libc.org>
Cc: "Aneesh Kumar K . V" <aneesh.kumar@linux.ibm.com>
Cc: Jacopo Mondi <jacopo+renesas@jmondi.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

mm/debug: sync up latest migrate_reason to migrate_reason_names

Sync up MR_DEMOTION to migrate_reason_names and add a synch prompt.

Link: https://lkml.kernel.org/r/20210921064553.293905-3-o451686892@gmail.com
Fixes: 26aa2d199d6f ("mm/migrate: demote pages during reclaim")
Signed-off-by: Weizhao Ouyang <o451686892@gmail.com>
Reviewed-by: "Huang, Ying" <ying.huang@intel.com>
Reviewed-by: John Hubbard <jhubbard@nvidia.com>
Cc: Anshuman Khandual <khandual@linux.vnet.ibm.com>
Cc: Michal Hocko <mhocko@suse.com>
Cc: Pavel Tatashin <pasha.tatashin@soleen.com>
Cc: Yang Shi <yang.shi@linux.alibaba.com>
Cc: Zi Yan <ziy@nvidia.com>
Cc: Dave Hansen <dave.hansen@linux.intel.com>
Cc: Minchan Kim <minchan@kernel.org>
Cc: Mina Almasry <almasrymina@google.com>
Cc: "Matthew Wilcox (Oracle)" <willy@infradead.org>
Cc: Oscar Salvador <osalvador@suse.de>
Cc: Wei Xu <weixugc@google.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

mm/debug: sync up MR_CONTIG_RANGE and MR_LONGTERM_PIN

Sync up MR_CONTIG_RANGE and MR_LONGTERM_PIN to migrate_reason_names.

Link: https://lkml.kernel.org/r/20210921064553.293905-2-o451686892@gmail.com
Fixes: 310253514bbf ("mm/migrate: rename migration reason MR_CMA to MR_CONTIG_RANGE")
Fixes: d1e153fea2a8 ("mm/gup: migrate pinned pages out of movable zone")
Signed-off-by: Weizhao Ouyang <o451686892@gmail.com>
Reviewed-by: "Huang, Ying" <ying.huang@intel.com>
Reviewed-by: John Hubbard <jhubbard@nvidia.com>
Cc: Anshuman Khandual <khandual@linux.vnet.ibm.com>
Cc: Michal Hocko <mhocko@suse.com>
Cc: Pavel Tatashin <pasha.tatashin@soleen.com>
Cc: Yang Shi <yang.shi@linux.alibaba.com>
Cc: Zi Yan <ziy@nvidia.com>
Cc: Dave Hansen <dave.hansen@linux.intel.com>
Cc: Minchan Kim <minchan@kernel.org>
Cc: Mina Almasry <almasrymina@google.com>
Cc: "Matthew Wilcox (Oracle)" <willy@infradead.org>
Cc: Oscar Salvador <osalvador@suse.de>
Cc: Wei Xu <weixugc@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>

mm: fs: invalidate bh_lrus for only cold path

The kernel test robot reported the regression of fio.write_iops[1] with
commit 8cc621d2f45d ("mm: fs: invalidate BH LRU during page migration").

Since lru_add_drain is called frequently, invalidate bh_lrus there could
increase bh_lrus cache miss ratio, which needs more IO in the end.

This patch moves the bh_lrus invalidation from the hot path( e.g.,
zap_page_range, pagevec_release) to cold path(i.e., lru_add_drain_all,
lru_cache_disable).

Zhengjun Xing confirmed
 "I test the patch, the regression reduced to -2.9%"

[1] https://lore.kernel.org/lkml/20210520083144.GD14190@xsang-OptiPlex-9020/
[2] 8cc621d2f45d, mm: fs: invalidate BH LRU during page migration

Link: https://lkml.kernel.org/r/20210907212347.1977686-1-minchan@kernel.org
Signed-off-by: Minchan Kim <minchan@kernel.org>
Reported-by: kernel test robot <oliver.sang@intel.com>
Reviewed-by: Chris Goldsworthy <cgoldswo@codeaurora.org>
Tested-by: "Xing, Zhengjun" <zhengjun.xing@intel.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

lib/zlib_inflate/inffast: check config in C to avoid unused function warning

Building Linux for ppc64le with Ubuntu clang version
12.0.0-3ubuntu1~21.04.1 shows the warning below.

    arch/powerpc/boot/inffast.c:20:1: warning: unused function 'get_unaligned16' [-Wunused-function]
    get_unaligned16(const unsigned short *p)
    ^
    1 warning generated.

Fix it by moving the check from the preprocessor to C, so the compiler
sees the use.

Link: https://lkml.kernel.org/r/20210920084332.5752-1-pmenzel@molgen.mpg.de
Signed-off-by: Paul Menzel <pmenzel@molgen.mpg.de>
Reviewed-by: Nathan Chancellor <nathan@kernel.org>
Tested-by: Nathan Chancellor <nathan@kernel.org>
Cc: Nick Desaulniers <ndesaulniers@google.com>
Cc: Christophe Leroy <christophe.leroy@csgroup.eu>
Cc: Zhen Lei <thunder.leizhen@huawei.com>
Cc: Michael Ellerman <mpe@ellerman.id.au>
Cc: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Cc: Paul Mackerras <paulus@samba.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

tools/vm/page-types: remove dependency on opt_file for idle page tracking

Idle page tracking can also be used for process address space, not only
file mappings.

Without this change, using with '-i' option for process address space
encounters below errors reported.

  $ sudo ./page-types -p $(pidof bash) -i
  mark page idle: Bad file descriptor
  mark page idle: Bad file descriptor
  mark page idle: Bad file descriptor
  mark page idle: Bad file descriptor
  ...

Link: https://lkml.kernel.org/r/20210917032826.10669-1-changbin.du@gmail.com
Signed-off-by: Changbin Du <changbin.du@gmail.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

scripts/sorttable: riscv: fix undeclared identifier 'EM_RISCV' error

Fix the following build failure reported in [1] by adding a conditional
definition of EM_RISCV in order to allow cross-compilation on machines
which do not have EM_RISCV definition in their host.

   scripts/sorttable.c:352:7: error: use of undeclared identifier 'EM_RISCV'

EM_RISCV was added to <elf.h> in glibc 2.24 so builds on systems with
glibc headers < 2.24 should show this error.

[mkubecek@suse.cz: changelog addition]
Link: https://lore.kernel.org/lkml/e8965b25-f15b-c7b4-748c-d207dda9c8e8@i2se.com/
Link: https://lkml.kernel.org/r/20210913030625.4525-1-miles.chen@mediatek.com
Fixes: 54fed35fd393 ("riscv: Enable BUILDTIME_TABLE_SORT")
Signed-off-by: Miles Chen <miles.chen@mediatek.com>
Reported-by: Stefan Wahren <stefan.wahren@i2se.com>
Tested-by: Stefan Wahren <stefan.wahren@i2se.com>
Reviewed-by: Jisheng Zhang <jszhang@kernel.org>
Cc: Michal Kubecek <mkubecek@suse.cz>
Cc: Paul Walmsley <paul.walmsley@sifive.com>
Cc: Palmer Dabbelt <palmer@dabbelt.com>
Cc: Albert Ou <aou@eecs.berkeley.edu>
Cc: Markus Mayer <mmayer@broadcom.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

ocfs2: drop acl cache for directories too

ocfs2_data_convert_worker() is currently dropping any cached acl info
for FILE before down-converting meta lock.  It should also drop for
DIRECTORY.  Otherwise the second acl lookup returns the cached one (from
VFS layer) which could be already stale.

The problem we are seeing is that the acl changes on one node doesn't
get refreshed on other nodes in the following case:

  Node 1                    Node 2
  --------------            ----------------
  getfacl dir1

                            getfacl dir1    <-- this is OK

  setfacl -m u:user1:rwX dir1
  getfacl dir1   <-- see the change for user1

                            getfacl dir1    <-- can't see change for user1

Link: https://lkml.kernel.org/r/20210903012631.6099-1-wen.gang.wang@oracle.com
Signed-off-by: Wengang Wang <wen.gang.wang@oracle.com>
Reviewed-by: Joseph Qi <joseph.qi@linux.alibaba.com>
Cc: Mark Fasheh <mark@fasheh.com>
Cc: Joel Becker <jlbec@evilplan.org>
Cc: Junxiao Bi <junxiao.bi@oracle.com>
Cc: Changwei Ge <gechangwei@live.cn>
Cc: Gang He <ghe@suse.com>
Cc: Jun Piao <piaojun@huawei.com>
Cc: <stable@vger.kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

mm/shmem.c: fix judgment error in shmem_is_huge()

In the case of SHMEM_HUGE_WITHIN_SIZE, the page index is not rounded up
correctly.  When the page index points to the first page in a huge page,
round_up() cannot bring it to the end of the huge page, but to the end
of the previous one.

An example:

HPAGE_PMD_NR on my machine is 512(2 MB huge page size).  After
allcoating a 3000 KB buffer, I access it at location 2050 KB.  In
shmem_is_huge(), the corresponding index happens to be 512.  After
rounded up by HPAGE_PMD_NR, it will still be 512 which is smaller than
i_size, and shmem_is_huge() will return true.  As a result, my buffer
takes an additional huge page, and that shouldn't happen when
shmem_enabled is set to within_size.

Link: https://lkml.kernel.org/r/20210909032007.18353-1-liuyuntao10@huawei.com
Fixes: f3f0e1d2150b2b ("khugepaged: add support of collapse for tmpfs/shmem pages")
Signed-off-by: Liu Yuntao <liuyuntao10@huawei.com>
Acked-by: Kirill A. Shutemov <kirill.shutemov@linux.intel.com>
Acked-by: Hugh Dickins <hughd@google.com>
Cc: wuxu.wu <wuxu.wu@huawei.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

xtensa: increase size of gcc stack frame check

xtensa frame size is larger than the frame size for almost all other
architectures.  This results in more than 50 "the frame size of <n> is
larger than 1024 bytes" errors when trying to build xtensa:allmodconfig.

Increase frame size for xtensa to 1536 bytes to avoid compile errors due
to frame size limits.

Link: https://lkml.kernel.org/r/20210912025235.3514761-1-linux@roeck-us.net
Signed-off-by: Guenter Roeck <linux@roeck-us.net>
Reviewed-by: Max Filippov <jcmvbkbc@gmail.com>
Cc: Chris Zankel <chris@zankel.net>
Cc: David Laight <David.Laight@ACULAB.COM>
Cc: Masahiro Yamada <masahiroy@kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

mm/damon: don't use strnlen() with known-bogus source length

gcc knows the true length too, and rightfully complains.

Link: https://lkml.kernel.org/r/20210912204447.10427-1-kilobyte@angband.pl
Signed-off-by: Adam Borowski <kilobyte@angband.pl>
Cc: SeongJae Park <sj38.park@gmail.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

kasan: fix Kconfig check of CC_HAS_WORKING_NOSANITIZE_ADDRESS

In the main KASAN config option CC_HAS_WORKING_NOSANITIZE_ADDRESS is
checked for instrumentation-based modes.  However, if
HAVE_ARCH_KASAN_HW_TAGS is true all modes may still be selected.

To fix, also make the software modes depend on
CC_HAS_WORKING_NOSANITIZE_ADDRESS.

Link: https://lkml.kernel.org/r/20210910084240.1215803-1-elver@google.com
Fixes: 6a63a63ff1ac ("kasan: introduce CONFIG_KASAN_HW_TAGS")
Signed-off-by: Marco Elver <elver@google.com>
Cc: Andrey Ryabinin <ryabinin.a.a@gmail.com>
Cc: Alexander Potapenko <glider@google.com>
Cc: Andrey Konovalov <andreyknvl@gmail.com>
Cc: Dmitry Vyukov <dvyukov@google.com>
Cc: Aleksandr Nogikh <nogikh@google.com>
Cc: Taras Madan <tarasmadan@google.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

mm, hwpoison: add is_free_buddy_page() in HWPoisonHandlable()

Commit fcc00621d88b ("mm/hwpoison: retry with shake_page() for
unhandlable pages") changed the return value of __get_hwpoison_page() to
retry for transiently unhandlable cases.  However, __get_hwpoison_page()
currently fails to properly judge buddy pages as handlable, so hard/soft
offline for buddy pages always fail as "unhandlable page".  This is
totally regrettable.

So let's add is_free_buddy_page() in HWPoisonHandlable(), so that
__get_hwpoison_page() returns different return values between buddy
pages and unhandlable pages as intended.

Link: https://lkml.kernel.org/r/20210909004131.163221-1-naoya.horiguchi@linux.dev
Fixes: fcc00621d88b ("mm/hwpoison: retry with shake_page() for unhandlable pages")
Signed-off-by: Naoya Horiguchi <naoya.horiguchi@nec.com>
Acked-by: David Hildenbrand <david@redhat.com>
Reviewed-by: Yang Shi <shy828301@gmail.com>
Cc: Tony Luck <tony.luck@intel.com>
Cc: Oscar Salvador <osalvador@suse.de>
Cc: Mike Kravetz <mike.kravetz@oracle.com>
Cc: 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>

io_uring: make OP_CLOSE consistent with direct open

From recently open/accept are now able to manipulate fixed file table,
but it's inconsistent that close can't. Close the gap, keep API same as
with open/accept, i.e. via sqe->file_slot.

Signed-off-by: Pavel Begunkov <asml.silence@gmail.com>
Signed-off-by: Jens Axboe <axboe@kernel.dk>

Merge tag 'gpio-fixes-for-v5.15-rc3' of git://git./linux/kernel/git/brgl/linux

Pull gpio fixes from Bartosz Golaszewski:

 - fix a regression in GPIO ACPI on HP ElitePad 1000 G2 where the
   gpio_set_debounce_timeout() now returns a fatal error if the specific
   debounce period is not supported by the driver instead of just
   emitting a warning

 - fix return values of irq_mask/unmask() callbacks in gpio-uniphier

 - fix hwirq calculation in gpio-aspeed-sgpio

 - fix two issues in gpio-rockchip: only make the extended debounce
   support available for v2 and remove a redundant BIT() usage

* tag 'gpio-fixes-for-v5.15-rc3' of git://git.kernel.org/pub/scm/linux/kernel/git/brgl/linux:
  gpio/rockchip: fix get_direction value handling
  gpio/rockchip: extended debounce support is only available on v2
  gpio: gpio-aspeed-sgpio: Fix wrong hwirq in irq handler.
  gpio: uniphier: Fix void functions to remove return value
  gpiolib: acpi: Make set-debounce-timeout failures non fatal

Merge tag 'devprop-5.15-rc3' of git://git./linux/kernel/git/rafael/linux-pm

Pull device properties framework fix from Rafael Wysocki:
 "Fix software node refcount imbalance on device removal (Laurentiu
  Tudor)"

* tag 'devprop-5.15-rc3' of git://git.kernel.org/pub/scm/linux/kernel/git/rafael/linux-pm:
  software node: balance refcount for managed software nodes

Merge tag 'acpi-5.15-rc3' of git://git./linux/kernel/git/rafael/linux-pm

Pull ACPI fix from Rafael Wysocki:
 "Revert a recent commit related to memory management that turned out to
  be problematic (Jia He)"

* tag 'acpi-5.15-rc3' of git://git.kernel.org/pub/scm/linux/kernel/git/rafael/linux-pm:
  Revert "ACPI: Add memory semantics to acpi_os_map_memory()"

Merge tag 'arm64-fixes' of git://git./linux/kernel/git/arm64/linux

Pull arm64 fixes from Catalin Marinas:

 - It turns out that the optimised string routines merged in 5.14 are
   not safe with in-kernel MTE (KASAN_HW_TAGS) because of reading beyond
   the end of a string (strcmp, strncmp). Such reading may go across a
   16 byte tag granule and cause a tag check fault. When KASAN_HW_TAGS
   is enabled, use the generic strcmp/strncmp C implementation.

 - An errata workaround for ThunderX relied on the CPU capabilities
   being enabled in a specific order. This disappeared with the
   automatic generation of the cpucaps.h file (sorted alphabetically).
   Fix it by checking the current CPU only rather than the system-wide
   capability.

 - Add system_supports_mte() checks on the kernel entry/exit path and
   thread switching to avoid unnecessary barriers and function calls on
   systems where MTE is not supported.

 - kselftests: skip arm64 tests if the required features are missing.

* tag 'arm64-fixes' of git://git.kernel.org/pub/scm/linux/kernel/git/arm64/linux:
  arm64: Restore forced disabling of KPTI on ThunderX
  kselftest/arm64: signal: Skip tests if required features are missing
  arm64: Mitigate MTE issues with str{n}cmp()
  arm64: add MTE supported check to thread switching and syscall entry/exit

Merge tag 'ceph-for-5.15-rc3' of git://github.com/ceph/ceph-client

Pull ceph fix from Ilya Dryomov:
 "A fix for a potential array out of bounds access from Dan"

* tag 'ceph-for-5.15-rc3' of git://github.com/ceph/ceph-client:
  ceph: fix off by one bugs in unsafe_request_wait()

Merge tag 'fixes_for_v5.15-rc3' of git://git./linux/kernel/git/jack/linux-fs

Pull misc filesystem fixes from Jan Kara:
 "A for ext2 sleep in atomic context in case of some fs problems and a
  cleanup of an invalidate_lock initialization"

* tag 'fixes_for_v5.15-rc3' of git://git.kernel.org/pub/scm/linux/kernel/git/jack/linux-fs:
  ext2: fix sleeping in atomic bugs on error
  mm: Fully initialize invalidate_lock, amend lock class later

Merge branch 'work.init' of git://git./linux/kernel/git/viro/vfs

Pull vfs fixes from Al Viro:
 "Followups to nodev root stuff from this merge window"

* 'work.init' of git://git.kernel.org/pub/scm/linux/kernel/git/viro/vfs:
  init: don't panic if mount_nodev_root failed
  init/do_mounts.c: Harden split_fs_names() against buffer overflow

Merge tag 'drm-fixes-2021-09-24' of git://anongit.freedesktop.org/drm/drm

Pull drm fixes from Dave Airlie:
 "Quiet week this week, just some i915 and amd fixes, just getting ready
  for my all nighter maintainer summit!

  Summary:

  i915:
   - Fix ADL-P memory bandwidth parameters
   - Fix memory corruption due to a double free
   - Fix memory leak in DMC firmware handling

  amdgpu:
   - Update MAINTAINERS entry for powerplay
   - Fix empty macros
   - SI DPM fix

  amdkfd:
   - SVM fixes
   - DMA mapping fix"

* tag 'drm-fixes-2021-09-24' of git://anongit.freedesktop.org/drm/drm:
  drm/amdkfd: fix svm_migrate_fini warning
  drm/amdkfd: handle svm migrate init error
  drm/amd/pm: Update intermediate power state for SI
  drm/amdkfd: fix dma mapping leaking warning
  drm/amdkfd: SVM map to gpus check vma boundary
  MAINTAINERS: fix up entry for AMD Powerplay
  drm/amd/display: fix empty debug macros
  drm/i915: Free all DMC payloads
  drm/i915: Move __i915_gem_free_object to ttm_bo_destroy
  drm/i915: Update memory bandwidth parameters

block: hold ->invalidate_lock in blkdev_fallocate

When running ->fallocate(), blkdev_fallocate() should hold
mapping->invalidate_lock to prevent page cache from being accessed,
otherwise stale data may be read in page cache.

Without this patch, blktests block/009 fails sometimes. With this patch,
block/009 can pass always.

Also as Jan pointed out, no pages can be created in the discarded area
while you are holding the invalidate_lock, so remove the 2nd
truncate_bdev_range().

Cc: Jan Kara <jack@suse.cz>
Signed-off-by: Ming Lei <ming.lei@redhat.com>
Reviewed-by: Jan Kara <jack@suse.cz>
Link: https://lore.kernel.org/r/20210923023751.1441091-1-ming.lei@redhat.com
Signed-off-by: Jens Axboe <axboe@kernel.dk>

blktrace: Fix uaf in blk_trace access after removing by sysfs

There is an use-after-free problem triggered by following process:

      P1(sda) P2(sdb)
echo 0 > /sys/block/sdb/trace/enable
  blk_trace_remove_queue
    synchronize_rcu
    blk_trace_free
      relay_close
rcu_read_lock
__blk_add_trace
  trace_note_tsk
  (Iterate running_trace_list)
        relay_close_buf
  relay_destroy_buf
    kfree(buf)
    trace_note(sdb's bt)
      relay_reserve
        buf->offset <- nullptr deference (use-after-free) !!!
rcu_read_unlock

[  502.714379] BUG: kernel NULL pointer dereference, address:
0000000000000010
[  502.715260] #PF: supervisor read access in kernel mode
[  502.715903] #PF: error_code(0x0000) - not-present page
[  502.716546] PGD 103984067 P4D 103984067 PUD 17592b067 PMD 0
[  502.717252] Oops: 0000 [#1] SMP
[  502.720308] RIP: 0010:trace_note.isra.0+0x86/0x360
[  502.732872] Call Trace:
[  502.733193]  __blk_add_trace.cold+0x137/0x1a3
[  502.733734]  blk_add_trace_rq+0x7b/0xd0
[  502.734207]  blk_add_trace_rq_issue+0x54/0xa0
[  502.734755]  blk_mq_start_request+0xde/0x1b0
[  502.735287]  scsi_queue_rq+0x528/0x1140
...
[  502.742704]  sg_new_write.isra.0+0x16e/0x3e0
[  502.747501]  sg_ioctl+0x466/0x1100

Reproduce method:
  ioctl(/dev/sda, BLKTRACESETUP, blk_user_trace_setup[buf_size=127])
  ioctl(/dev/sda, BLKTRACESTART)
  ioctl(/dev/sdb, BLKTRACESETUP, blk_user_trace_setup[buf_size=127])
  ioctl(/dev/sdb, BLKTRACESTART)

  echo 0 > /sys/block/sdb/trace/enable &
  // Add delay(mdelay/msleep) before kernel enters blk_trace_free()

  ioctl$SG_IO(/dev/sda, SG_IO, ...)
  // Enters trace_note_tsk() after blk_trace_free() returned
  // Use mdelay in rcu region rather than msleep(which may schedule out)

Remove blk_trace from running_list before calling blk_trace_free() by
sysfs if blk_trace is at Blktrace_running state.

Fixes: c71a896154119f ("blktrace: add ftrace plugin")
Signed-off-by: Zhihao Cheng <chengzhihao1@huawei.com>
Link: https://lore.kernel.org/r/20210923134921.109194-1-chengzhihao1@huawei.com
Signed-off-by: Jens Axboe <axboe@kernel.dk>

block: don't call rq_qos_ops->done_bio if the bio isn't tracked

rq_qos framework is only applied on request based driver, so:

1) rq_qos_done_bio() needn't to be called for bio based driver

2) rq_qos_done_bio() needn't to be called for bio which isn't tracked,
such as bios ended from error handling code.

Especially in bio_endio():

1) request queue is referred via bio->bi_bdev->bd_disk->queue, which
may be gone since request queue refcount may not be held in above two
cases

2) q->rq_qos may be freed in blk_cleanup_queue() when calling into
__rq_qos_done_bio()

Fix the potential kernel panic by not calling rq_qos_ops->done_bio if
the bio isn't tracked. This way is safe because both ioc_rqos_done_bio()
and blkcg_iolatency_done_bio() are nop if the bio isn't tracked.

Reported-by: Yu Kuai <yukuai3@huawei.com>
Cc: tj@kernel.org
Signed-off-by: Ming Lei <ming.lei@redhat.com>
Reviewed-by: Christoph Hellwig <hch@lst.de>
Acked-by: Tejun Heo <tj@kernel.org>
Link: https://lore.kernel.org/r/20210924110704.1541818-1-ming.lei@redhat.com
Signed-off-by: Jens Axboe <axboe@kernel.dk>

io_uring: kill extra checks in io_write()

We don't retry short writes and so we would never get to async setup in
io_write() in that case. Thus ret2 > 0 is always false and
iov_iter_advance() is never used. Apparently, the same is found by
Coverity, which complains on the code.

Fixes: cd65869512ab ("io_uring: use iov_iter state save/restore helpers")
Reported-by: Dave Jones <davej@codemonkey.org.uk>
Signed-off-by: Pavel Begunkov <asml.silence@gmail.com>
Link: https://lore.kernel.org/r/5b33e61034748ef1022766efc0fb8854cfcf749c.1632500058.git.asml.silence@gmail.com
Signed-off-by: Jens Axboe <axboe@kernel.dk>

io_uring: don't punt files update to io-wq unconditionally

There's no reason to punt it unconditionally, we just need to ensure that
the submit lock grabbing is conditional.

Fixes: 05f3fb3c5397 ("io_uring: avoid ring quiesce for fixed file set unregister and update")
Signed-off-by: Jens Axboe <axboe@kernel.dk>

io_uring: put provided buffer meta data under memcg accounting

For each provided buffer, we allocate a struct io_buffer to hold the
data associated with it. As a large number of buffers can be provided,
account that data with memcg.

Fixes: ddf0322db79c ("io_uring: add IORING_OP_PROVIDE_BUFFERS")
Signed-off-by: Jens Axboe <axboe@kernel.dk>

io_uring: allow conditional reschedule for intensive iterators

If we have a lot of threads and rings, the tctx list can get quite big.
This is especially true if we keep creating new threads and rings.
Likewise for the provided buffers list. Be nice and insert a conditional
reschedule point while iterating the nodes for deletion.

Link: https://lore.kernel.org/io-uring/00000000000064b6b405ccb41113@google.com/
Reported-by: syzbot+111d2a03f51f5ae73775@syzkaller.appspotmail.com
Signed-off-by: Jens Axboe <axboe@kernel.dk>

io_uring: fix potential req refcount underflow

For multishot mode, there may be cases like:

iowq                                 original context
io_poll_add
  _arm_poll()
  mask = vfs_poll() is not 0
  if mask
(2)  io_poll_complete()
  compl_unlock
   (interruption happens
    tw queued to original
    context)
                                     io_poll_task_func()
                                     compl_lock
                                 (3) done = io_poll_complete() is true
                                     compl_unlock
                                     put req ref
(1) if (poll->flags & EPOLLONESHOT)
      put req ref

EPOLLONESHOT flag in (1) may be from (2) or (3), so there are multiple
combinations that can cause ref underfow.
Let's address it by:
- check the return value in (2) as done
- change (1) to if (done)
    in this way, we only do ref put in (1) if 'oneshot flag' is from
    (2)
- do poll.done check in io_poll_task_func(), so that we won't put ref
  for the second time.

Signed-off-by: Hao Xu <haoxu@linux.alibaba.com>
Link: https://lore.kernel.org/r/20210922101238.7177-4-haoxu@linux.alibaba.com
Signed-off-by: Jens Axboe <axboe@kernel.dk>

io_uring: fix missing set of EPOLLONESHOT for CQ ring overflow

We should set EPOLLONESHOT if cqring_fill_event() returns false since
io_poll_add() decides to put req or not by it.

Fixes: 5082620fb2ca ("io_uring: terminate multishot poll for CQ ring overflow")
Signed-off-by: Hao Xu <haoxu@linux.alibaba.com>
Link: https://lore.kernel.org/r/20210922101238.7177-3-haoxu@linux.alibaba.com
Signed-off-by: Jens Axboe <axboe@kernel.dk>

io_uring: fix race between poll completion and cancel_hash insertion

If poll arming and poll completion runs in parallel, there maybe races.
For instance, run io_poll_add in iowq and io_poll_task_func in original
context, then:

  iowq                                      original context
  io_poll_add
    vfs_poll
     (interruption happens
      tw queued to original
      context)                              io_poll_task_func
                                              generate cqe
                                              del from cancel_hash[]
    if !poll.done
      insert to cancel_hash[]

The entry left in cancel_hash[], similar case for fast poll.
Fix it by set poll.done = true when del from cancel_hash[].

Fixes: 5082620fb2ca ("io_uring: terminate multishot poll for CQ ring overflow")
Signed-off-by: Hao Xu <haoxu@linux.alibaba.com>
Link: https://lore.kernel.org/r/20210922101238.7177-2-haoxu@linux.alibaba.com
Signed-off-by: Jens Axboe <axboe@kernel.dk>

io-wq: ensure we exit if thread group is exiting

Dave reports that a coredumping workload gets stuck in 5.15-rc2, and
identified the culprit in the Fixes line below. The problem is that
relying solely on fatal_signal_pending() to gate whether to exit or not
fails miserably if a process gets eg SIGILL sent. Don't exclusively
rely on fatal signals, also check if the thread group is exiting.

Fixes: 15e20db2e0ce ("io-wq: only exit on fatal signals")
Reported-by: Dave Chinner <david@fromorbit.com>
Signed-off-by: Jens Axboe <axboe@kernel.dk>

Merge tag 'nvme-5.15-2021-09-24' of git://git.infradead.org/nvme into block-5.15

Pull NVMe fixes from Christoph:

"nvme fixes for Linux 5.15:

 - keep ctrl->namespaces ordered (me)
 - fix incorrect h2cdata pdu offset accounting in nvme-tcp
   (Sagi Grimberg)
 - handled updated hw_queues in nvme-fc more carefully (Daniel Wagner,
   James Smart)"

* tag 'nvme-5.15-2021-09-24' of git://git.infradead.org/nvme:
  nvme: keep ctrl->namespaces ordered
  nvme-tcp: fix incorrect h2cdata pdu offset accounting
  nvme-fc: remove freeze/unfreeze around update_nr_hw_queues
  nvme-fc: avoid race between time out and tear down
  nvme-fc: update hardware queues before using them

Merge tag 'irqchip-fixes-5.15-1' of git://git./linux/kernel/git/maz/arm-platforms into irq/urgent

Pull irqchip fixes from Marc Zyngier:

 - Work around a bad GIC integration on a Renesas platform, where the
   interconnect cannot deal with byte-sized MMIO accesses

 - Cleanup another Renesas driver abusing the comma operator

 - Fix a potential GICv4 memory leak on an error path

 - Make the type of 'size' consistent with the rest of the code in
   __irq_domain_add()

 - Fix a regression in the Armada 370-XP IPI path

 - Fix the build for the obviously unloved goldfish-pic

 - Some documentation fixes

Link: https://lore.kernel.org/r/20210924090933.2766857-1-maz@kernel.org

x86/insn, tools/x86: Fix undefined behavior due to potential unaligned accesses

Don't perform unaligned loads in __get_next() and __peek_nbyte_next() as
these are forms of undefined behavior:

"A pointer to an object or incomplete type may be converted to a pointer
to a different object or incomplete type. If the resulting pointer
is not correctly aligned for the pointed-to type, the behavior is
undefined."

(from http://www.open-std.org/jtc1/sc22/wg14/www/docs/n1256.pdf)

These problems were identified using the undefined behavior sanitizer
(ubsan) with the tools version of the code and perf test.

 [ bp: Massage commit message. ]

Signed-off-by: Numfor Mbiziwo-Tiapo <nums@google.com>
Signed-off-by: Ian Rogers <irogers@google.com>
Signed-off-by: Borislav Petkov <bp@suse.de>
Acked-by: Masami Hiramatsu <mhiramat@kernel.org>
Link: https://lkml.kernel.org/r/20210923161843.751834-1-irogers@google.com

Merge tag 'usb-serial-5.15-rc3' of https://git./linux/kernel/git/johan/usb-serial into usb-linus

Johan writes:

USB-serial fixes for 5.15-rc3

Here's a fix for a regression affecting some CP2102 devices and a host
of new device ids.

Included are also a couple of cleanups of duplicate device ids, which
are also tagged for stable to keep the tables in sync, and a trivial
patch to help debugging cp210x issues.

All have been in linux-next with no reported issues. Note however that
the last last two device-id commits were rebased to fix up a lore link
in a commit message (as the patch itself never made it to the list).

* tag 'usb-serial-5.15-rc3' of https://git.kernel.org/pub/scm/linux/kernel/git/johan/usb-serial:
  USB: serial: option: add device id for Foxconn T99W265
  USB: serial: cp210x: add ID for GW Instek GDM-834x Digital Multimeter
  USB: serial: cp210x: add part-number debug printk
  USB: serial: cp210x: fix dropped characters with CP2102
  USB: serial: option: remove duplicate USB device ID
  USB: serial: mos7840: remove duplicated 0xac24 device ID
  USB: serial: option: add Telit LN920 compositions

USB: serial: option: add device id for Foxconn T99W265

Adding support for Foxconn device T99W265 for enumeration with
PID 0xe0db.

usb-devices output for 0xe0db
T:  Bus=04 Lev=01 Prnt=01 Port=00 Cnt=01 Dev#= 19 Spd=5000 MxCh= 0
D:  Ver= 3.20 Cls=ef(misc ) Sub=02 Prot=01 MxPS= 9 #Cfgs=  1
P:  Vendor=0489 ProdID=e0db Rev=05.04
S:  Manufacturer=Microsoft
S:  Product=Generic Mobile Broadband Adapter
S:  SerialNumber=6c50f452
C:  #Ifs= 5 Cfg#= 1 Atr=a0 MxPwr=896mA
I:  If#=0x0 Alt= 0 #EPs= 1 Cls=02(commc) Sub=0e Prot=00 Driver=cdc_mbim
I:  If#=0x1 Alt= 1 #EPs= 2 Cls=0a(data ) Sub=00 Prot=02 Driver=cdc_mbim
I:  If#=0x2 Alt= 0 #EPs= 3 Cls=ff(vend.) Sub=ff Prot=40 Driver=option
I:  If#=0x3 Alt= 0 #EPs= 1 Cls=ff(vend.) Sub=ff Prot=ff Driver=(none)
I:  If#=0x4 Alt= 0 #EPs= 2 Cls=ff(vend.) Sub=ff Prot=30 Driver=option

if0/1: MBIM, if2:Diag, if3:GNSS, if4: Modem

Signed-off-by: Slark Xiao <slark_xiao@163.com>
Link: https://lore.kernel.org/r/20210917110106.9852-1-slark_xiao@163.com
[ johan: use USB_DEVICE_INTERFACE_CLASS(), amend comment ]
Cc: stable@vger.kernel.org
Signed-off-by: Johan Hovold <johan@kernel.org>

USB: serial: cp210x: add ID for GW Instek GDM-834x Digital Multimeter

Add the USB serial device ID for the GW Instek GDM-834x Digital Multimeter.

Signed-off-by: Uwe Brandt <uwe.brandt@gmail.com>
Link: https://lore.kernel.org/r/YUxFl3YUCPGJZd8Y@hovoldconsulting.com
Cc: stable@vger.kernel.org
Signed-off-by: Johan Hovold <johan@kernel.org>

cifs: fix incorrect check for null pointer in header_assemble

Although very unlikely that the tlink pointer would be null in this case,
get_next_mid function can in theory return null (but not an error)
so need to check for null (not for IS_ERR, which can not be returned
here).

Address warning:

        fs/smbfs_client/connect.c:2392 cifs_match_super()
        warn: 'tlink' isn't an ERR_PTR

Pointed out by Dan Carpenter via smatch code analysis tool

CC: stable@vger.kernel.org
Reported-by: Dan Carpenter <dan.carpenter@oracle.com>
Acked-by: Ronnie Sahlberg <lsahlber@redhat.com>
Signed-off-by: Steve French <stfrench@microsoft.com>

smb3: correct server pointer dereferencing check to be more consistent

Address warning:

    fs/smbfs_client/misc.c:273 header_assemble()
    warn: variable dereferenced before check 'treeCon->ses->server'

Pointed out by Dan Carpenter via smatch code analysis tool

Although the check is likely unneeded, adding it makes the code
more consistent and easier to read, as the same check is
done elsewhere in the function.

Reported-by: Dan Carpenter <dan.carpenter@oracle.com>
Acked-by: Ronnie Sahlberg <lsahlber@redhat.com>
Signed-off-by: Steve French <stfrench@microsoft.com>

Merge tag 'drm-intel-fixes-2021-09-23' of git://anongit.freedesktop.org/drm/drm-intel into drm-fixes

drm/i915 fixes for v5.15-rc3:
- Fix ADL-P memory bandwidth parameters
- Fix memory corruption due to a double free
- Fix memory leak in DMC firmware handling

Signed-off-by: Dave Airlie <airlied@redhat.com>
From: Jani Nikula <jani.nikula@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/87o88jbk3o.fsf@intel.com

Merge tag 'amd-drm-fixes-5.15-2021-09-23' of https://gitlab.freedesktop.org/agd5f/linux into drm-fixes

amd-drm-fixes-5.15-2021-09-23:

amdgpu:
- Update MAINTAINERS entry for powerplay
- Fix empty macros
- SI DPM fix

amdkfd:
- SVM fixes
- DMA mapping fix

Signed-off-by: Dave Airlie <airlied@redhat.com>
From: Alex Deucher <alexander.deucher@amd.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20210923211330.20725-1-alexander.deucher@amd.com

Merge tag 'for-5.15-rc2-tag' of git://git./linux/kernel/git/kdave/linux

Pull btrfs fixes from David Sterba:

 - regression fix for leak of transaction handle after verity rollback
   failure

 - properly reset device last error between mounts

 - improve one error handling case when checksumming bios

 - fixup confusing displayed size of space info free space

* tag 'for-5.15-rc2-tag' of git://git.kernel.org/pub/scm/linux/kernel/git/kdave/linux:
  btrfs: prevent __btrfs_dump_space_info() to underflow its free space
  btrfs: fix mount failure due to past and transient device flush error
  btrfs: fix transaction handle leak after verity rollback failure
  btrfs: replace BUG_ON() in btrfs_csum_one_bio() with proper error handling

smb3: correct smb3 ACL security descriptor

Address warning:

        fs/smbfs_client/smb2pdu.c:2425 create_sd_buf()
        warn: struct type mismatch 'smb3_acl vs cifs_acl'

Pointed out by Dan Carpenter via smatch code analysis tool

Reported-by: Dan Carpenter <dan.carpenter@oracle.com>
Acked-by: Ronnie Sahlberg <lsahlber@redhat.com>
Signed-off-by: Steve French <stfrench@microsoft.com>

Merge tag 'selinux-pr-20210923' of git://git./linux/kernel/git/pcmoore/selinux

Pull SELinux/Smack fixes from Paul Moore:
 "Another single-patch pull request for SELinux, as well as Smack.

  This fixes some credential misuse and is explained reasonably well in
  the patch description"

* tag 'selinux-pr-20210923' of git://git.kernel.org/pub/scm/linux/kernel/git/pcmoore/selinux:
  selinux,smack: fix subjective/objective credential use mixups