linux.git/fs/afs, branch v5.15

Merge tag 'misc-fixes-20211007' of git://git.kernel.org/pub/scm/linux/kernel/git/dhowells/linux-fs

2021-10-07T18:20:08+00:00

Pull netfslib, cachefiles and afs fixes from David Howells:

 - Fix another couple of oopses in cachefiles tracing stemming from the
   possibility of passing in a NULL object pointer

 - Fix netfs_clear_unread() to set READ on the iov_iter so that source
   it is passed to doesn't do the wrong thing (some drivers look at the
   flag on iov_iter rather than other available information to determine
   the direction)

 - Fix afs_launder_page() to write back at the correct file position on
   the server so as not to corrupt data

* tag 'misc-fixes-20211007' of git://git.kernel.org/pub/scm/linux/kernel/git/dhowells/linux-fs:
  afs: Fix afs_launder_page() to set correct start file position
  netfs: Fix READ/WRITE confusion when calling iov_iter_xarray()
  cachefiles: Fix oops with cachefiles_cull() due to NULL object

afs: Fix afs_launder_page() to set correct start file position

2021-10-05T10:22:06+00:00

Fix afs_launder_page() to set the starting position of the StoreData RPC at
the offset into the page at which the modified data starts instead of at
the beginning of the page (the iov_iter is correctly offset).

The offset got lost during the conversion to passing an iov_iter into
afs_store_data().

Changes:
ver #2:
 - Use page_offset() rather than manually calculating it[1].

Fixes: bd80d8a80e12 ("afs: Use ITER_XARRAY for writing")
Signed-off-by: David Howells 
Reviewed-by: Jeffrey Altman 
cc: Marc Dionne 
cc: linux-afs@lists.infradead.org
Link: https://lore.kernel.org/r/YST/0e92OdSH0zjg@casper.infradead.org/ [1]
Link: https://lore.kernel.org/r/162880783179.3421678.7795105718190440134.stgit@warthog.procyon.org.uk/ # v1
Link: https://lore.kernel.org/r/162937512409.1449272.18441473411207824084.stgit@warthog.procyon.org.uk/ # v1
Link: https://lore.kernel.org/r/162981148752.1901565.3663780601682206026.stgit@warthog.procyon.org.uk/ # v1
Link: https://lore.kernel.org/r/163005741670.2472992.2073548908229887941.stgit@warthog.procyon.org.uk/ # v2
Link: https://lore.kernel.org/r/163221839087.3143591.14278359695763025231.stgit@warthog.procyon.org.uk/ # v2
Link: https://lore.kernel.org/r/163292980654.4004896.7134735179887998551.stgit@warthog.procyon.org.uk/ # v2

afs: Fix kerneldoc warning shown up by W=1

2021-10-04T21:04:44+00:00

Fix a kerneldoc warning in afs due to a partially documented internal
function by removing the kerneldoc marker.

Signed-off-by: David Howells 
Reviewed-by: Jeff Layton 
cc: Marc Dionne 
cc: linux-afs@lists.infradead.org
cc: linux-fsdevel@vger.kernel.org
cc: linux-doc@vger.kernel.org
Link: https://lore.kernel.org/r/163214005516.2945267.7000234432243167892.stgit@warthog.procyon.org.uk/ # rfc v1
Link: https://lore.kernel.org/r/163281899704.2790286.9177774252843775348.stgit@warthog.procyon.org.uk/ # rfc v2

afs: Fix updating of i_blocks on file/dir extension

2021-09-13T08:14:21+00:00

When an afs file or directory is modified locally such that the total file
size is extended, i_blocks needs to be recalculated too.

Fix this by making afs_write_end() and afs_edit_dir_add() call
afs_set_i_size() rather than setting inode->i_size directly as that also
recalculates inode->i_blocks.

This can be tested by creating and writing into directories and files and
then examining them with du.  Without this change, directories show a 4
blocks (they start out at 2048 bytes) and files show 0 blocks; with this
change, they should show a number of blocks proportional to the file size
rounded up to 1024.

Fixes: 31143d5d515e ("AFS: implement basic file write support")
Fixes: 63a4681ff39c ("afs: Locally edit directory data for mkdir/create/unlink/...")
Reported-by: Markus Suvanto 
Signed-off-by: David Howells 
Reviewed-by: Marc Dionne 
Tested-by: Markus Suvanto 
cc: linux-afs@lists.infradead.org
Link: https://lore.kernel.org/r/163113612442.352844.11162345591911691150.stgit@warthog.procyon.org.uk/

afs: Fix corruption in reads at fpos 2G-4G from an OpenAFS server

2021-09-13T08:14:21+00:00

AFS-3 has two data fetch RPC variants, FS.FetchData and FS.FetchData64, and
Linux's afs client switches between them when talking to a non-YFS server
if the read size, the file position or the sum of the two have the upper 32
bits set of the 64-bit value.

This is a problem, however, since the file position and length fields of
FS.FetchData are *signed* 32-bit values.

Fix this by capturing the capability bits obtained from the fileserver when
it's sent an FS.GetCapabilities RPC, rather than just discarding them, and
then picking out the VICED_CAPABILITY_64BITFILES flag.  This can then be
used to decide whether to use FS.FetchData or FS.FetchData64 - and also
FS.StoreData or FS.StoreData64 - rather than using upper_32_bits() to
switch on the parameter values.

This capabilities flag could also be used to limit the maximum size of the
file, but all servers must be checked for that.

Note that the issue does not exist with FS.StoreData - that uses *unsigned*
32-bit values.  It's also not a problem with Auristor servers as its
YFS.FetchData64 op uses unsigned 64-bit values.

This can be tested by cloning a git repo through an OpenAFS client to an
OpenAFS server and then doing "git status" on it from a Linux afs
client[1].  Provided the clone has a pack file that's in the 2G-4G range,
the git status will show errors like:

	error: packfile .git/objects/pack/pack-5e813c51d12b6847bbc0fcd97c2bca66da50079c.pack does not match index
	error: packfile .git/objects/pack/pack-5e813c51d12b6847bbc0fcd97c2bca66da50079c.pack does not match index

This can be observed in the server's FileLog with something like the
following appearing:

Sun Aug 29 19:31:39 2021 SRXAFS_FetchData, Fid = 2303380852.491776.3263114, Host 192.168.11.201:7001, Id 1001
Sun Aug 29 19:31:39 2021 CheckRights: len=0, for host=192.168.11.201:7001
Sun Aug 29 19:31:39 2021 FetchData_RXStyle: Pos 18446744071815340032, Len 3154
Sun Aug 29 19:31:39 2021 FetchData_RXStyle: file size 2400758866
...
Sun Aug 29 19:31:40 2021 SRXAFS_FetchData returns 5

Note the file position of 18446744071815340032.  This is the requested file
position sign-extended.

Fixes: b9b1f8d5930a ("AFS: write support fixes")
Reported-by: Markus Suvanto 
Signed-off-by: David Howells 
Reviewed-by: Marc Dionne 
Tested-by: Markus Suvanto 
cc: linux-afs@lists.infradead.org
cc: openafs-devel@openafs.org
Link: https://bugzilla.kernel.org/show_bug.cgi?id=214217#c9 [1]
Link: https://lore.kernel.org/r/951332.1631308745@warthog.procyon.org.uk/

afs: Try to avoid taking RCU read lock when checking vnode validity

2021-09-13T08:10:39+00:00

Try to avoid taking the RCU read lock when checking the validity of a
vnode's callback state.  The only thing it's needed for is to pin the
parent volume's server list whilst we search it to find the record of the
server we're currently using to see if it has been reinitialised (ie. it
sent us a CB.InitCallBackState* RPC).

Do this by the following means:

 (1) Keep an additional per-cell counter (fs_s_break) that's incremented
     each time any of the fileservers in the cell reinitialises.

     Since the new counter can be accessed without RCU from the vnode, we
     can check that first - and only if it differs, get the RCU read lock
     and check the volume's server list.

 (2) Replace afs_get_s_break_rcu() with afs_check_server_good() which now
     indicates whether the callback promise is still expected to be present
     on the server.  This does the checks as described in (1).

 (3) Restructure afs_check_validity() to take account of the change in (2).

     We can also get rid of the valid variable and just use the need_clear
     variable with the addition of the afs_cb_break_no_promise reason.

 (4) afs_check_validity() probably shouldn't be altering vnode->cb_v_break
     and vnode->cb_s_break when it doesn't have cb_lock exclusively locked.

     Move the change to vnode->cb_v_break to __afs_break_callback().

     Delegate the change to vnode->cb_s_break to afs_select_fileserver()
     and set vnode->cb_fs_s_break there also.

 (5) afs_validate() no longer needs to get the RCU read lock around its
     call to afs_check_validity() - and can skip the call entirely if we
     don't have a promise.

Signed-off-by: David Howells 
Tested-by: Markus Suvanto 
cc: linux-afs@lists.infradead.org
Link: https://lore.kernel.org/r/163111669583.283156.1397603105683094563.stgit@warthog.procyon.org.uk/

afs: Fix mmap coherency vs 3rd-party changes

2021-09-13T08:10:39+00:00

Fix the coherency management of mmap'd data such that 3rd-party changes
become visible as soon as possible after the callback notification is
delivered by the fileserver.  This is done by the following means:

 (1) When we break a callback on a vnode specified by the CB.CallBack call
     from the server, we queue a work item (vnode->cb_work) to go and
     clobber all the PTEs mapping to that inode.

     This causes the CPU to trip through the ->map_pages() and
     ->page_mkwrite() handlers if userspace attempts to access the page(s)
     again.

     (Ideally, this would be done in the service handler for CB.CallBack,
     but the server is waiting for our reply before considering, and we
     have a list of vnodes, all of which need breaking - and the process of
     getting the mmap_lock and stripping the PTEs on all CPUs could be
     quite slow.)

 (2) Call afs_validate() from the ->map_pages() handler to check to see if
     the file has changed and to get a new callback promise from the
     server.

Also handle the fileserver telling us that it's dropping all callbacks,
possibly after it's been restarted by sending us a CB.InitCallBackState*
call by the following means:

 (3) Maintain a per-cell list of afs files that are currently mmap'd
     (cell->fs_open_mmaps).

 (4) Add a work item to each server that is invoked if there are any open
     mmaps when CB.InitCallBackState happens.  This work item goes through
     the aforementioned list and invokes the vnode->cb_work work item for
     each one that is currently using this server.

     This causes the PTEs to be cleared, causing ->map_pages() or
     ->page_mkwrite() to be called again, thereby calling afs_validate()
     again.

I've chosen to simply strip the PTEs at the point of notification reception
rather than invalidate all the pages as well because (a) it's faster, (b)
we may get a notification for other reasons than the data being altered (in
which case we don't want to clobber the pagecache) and (c) we need to ask
the server to find out - and I don't want to wait for the reply before
holding up userspace.

This was tested using the attached test program:

	#include 
	#include 
	#include 
	#include 
	#include 
	#include 
	int main(int argc, char *argv[])
	{
		size_t size = getpagesize();
		unsigned char *p;
		bool mod = (argc == 3);
		int fd;
		if (argc != 2 && argc != 3) {
			fprintf(stderr, "Format: %s  [mod]\n", argv[0]);
			exit(2);
		}
		fd = open(argv[1], mod ? O_RDWR : O_RDONLY);
		if (fd < 0) {
			perror(argv[1]);
			exit(1);
		}

		p = mmap(NULL, size, mod ? PROT_READ|PROT_WRITE : PROT_READ,
			 MAP_SHARED, fd, 0);
		if (p == MAP_FAILED) {
			perror("mmap");
			exit(1);
		}
		for (;;) {
			if (mod) {
				p[0]++;
				msync(p, size, MS_ASYNC);
				fsync(fd);
			}
			printf("%02x", p[0]);
			fflush(stdout);
			sleep(1);
		}
	}

It runs in two modes: in one mode, it mmaps a file, then sits in a loop
reading the first byte, printing it and sleeping for a second; in the
second mode it mmaps a file, then sits in a loop incrementing the first
byte and flushing, then printing and sleeping.

Two instances of this program can be run on different machines, one doing
the reading and one doing the writing.  The reader should see the changes
made by the writer, but without this patch, they aren't because validity
checking is being done lazily - only on entry to the filesystem.

Testing the InitCallBackState change is more complicated.  The server has
to be taken offline, the saved callback state file removed and then the
server restarted whilst the reading-mode program continues to run.  The
client machine then has to poke the server to trigger the InitCallBackState
call.

Signed-off-by: David Howells 
Tested-by: Markus Suvanto 
cc: linux-afs@lists.infradead.org
Link: https://lore.kernel.org/r/163111668833.283156.382633263709075739.stgit@warthog.procyon.org.uk/

afs: Fix incorrect triggering of sillyrename on 3rd-party invalidation

2021-09-13T08:10:39+00:00

The AFS filesystem is currently triggering the silly-rename cleanup from
afs_d_revalidate() when it sees that a dentry has been changed by a third
party[1]. It should not be doing this as the cleanup includes deleting the
silly-rename target file on iput.

Fix this by removing the places in the d_revalidate handling that validate
anything other than the directory and the dirent. It probably should not
be looking to validate the target inode of the dentry also.

This includes removing the point in afs_d_revalidate() where the inode that
a dentry used to point to was marked as being deleted (AFS_VNODE_DELETED).
We don't know it got deleted. It could have been renamed or it could have
hard links remaining.

This was reproduced by cloning a git repo onto an afs volume on one
machine, switching to another machine and doing "git status", then
switching back to the first and doing "git status". The second status
would show weird output due to ".git/index" getting deleted by the above
mentioned mechanism.

A simpler way to do it is to do:

machine 1: touch a
machine 2: touch b; mv -f b a
machine 1: stat a

on an afs volume. The bug shows up as the stat failing with ENOENT and the
file server log showing that machine 1 deleted "a".

Fixes: 79ddbfa500b3 ("afs: Implement sillyrename for unlink and rename")
Reported-by: Markus Suvanto
Signed-off-by: David Howells
Tested-by: Markus Suvanto
cc: linux-afs@lists.infradead.org
Link: https://bugzilla.kernel.org/show_bug.cgi?id=214217#c4 [1]
Link: https://lore.kernel.org/r/163111668100.283156.3851669884664475428.stgit@warthog.procyon.org.uk/

afs: Add missing vnode validation checks

2021-09-13T08:10:39+00:00

afs_d_revalidate() should only be validating the directory entry it is
given and the directory to which that belongs; it shouldn't be validating
the inode/vnode to which that dentry points.  Besides, validation need to
be done even if we don't call afs_d_revalidate() - which might be the case
if we're starting from a file descriptor.

In order for afs_d_revalidate() to be fixed, validation points must be
added in some other places.  Certain directory operations, such as
afs_unlink(), already check this, but not all and not all file operations
either.

Note that the validation of a vnode not only checks to see if the
attributes we have are correct, but also gets a promise from the server to
notify us if that file gets changed by a third party.

Add the following checks:

 - Check the vnode we're going to make a hard link to.
 - Check the vnode we're going to move/rename.
 - Check the vnode we're going to read from.
 - Check the vnode we're going to write to.
 - Check the vnode we're going to sync.
 - Check the vnode we're going to make a mapped page writable for.

Some of these aren't strictly necessary as we're going to perform a server
operation that might get the attributes anyway from which we can determine
if something changed - though it might not get us a callback promise.

Signed-off-by: David Howells 
Tested-by: Markus Suvanto 
cc: linux-afs@lists.infradead.org
Link: https://lore.kernel.org/r/163111667354.283156.12720698333342917516.stgit@warthog.procyon.org.uk/

afs: Fix page leak

2021-09-10T21:14:51+00:00

There's a loop in afs_extend_writeback() that adds extra pages to a write
we want to make to improve the efficiency of the writeback by making it
larger.  This loop stops, however, if we hit a page we can't write back
from immediately, but it doesn't get rid of the page ref we speculatively
acquired.

This was caused by the removal of the cleanup loop when the code switched
from using find_get_pages_contig() to xarray scanning as the latter only
gets a single page at a time, not a batch.

Fix this by putting the page on a ref on an early break from the loop.
Unfortunately, we can't just add that page to the pagevec we're employing
as we'll go through that and add those pages to the RPC call.

This was found by the generic/074 test.  It leaks ~4GiB of RAM each time it
is run - which can be observed with "top".

Fixes: e87b03f5830e ("afs: Prepare for use of THPs")
Reported-by: Marc Dionne 
Signed-off-by: David Howells 
Reviewed-and-tested-by: Marc Dionne 
cc: linux-afs@lists.infradead.org
Link: https://lore.kernel.org/r/163111666635.283156.177701903478910460.stgit@warthog.procyon.org.uk/