Logo Questions Linux Laravel Mysql Ubuntu Git Menu
 

should LOCK_EX on both read & write be atomic?

file_put_contents ( "file", "data", LOCK_EX )

for writing (which means - aquire lock and write)

file_get_contents ( "file", LOCK_EX )

for reading (which means - aquire lock and then read)

will it throw exception? raise an error? block until lock is aquired? or at least - should it? is there a chance that php will behave like this one day?

EDIT: i know that it is possible to use rename - i'd like to know answer to this...

like image 384
Kamil Tomšík Avatar asked Feb 04 '11 15:02

Kamil Tomšík


2 Answers

Since this answer is long, here's the summary: No, file_get_contents() is not atomic as it does not respect advisory locks.

About file locks in PHP:

In PHP, while on a *nix platform, filesystem locking is advisory only. Per the docs (Emphasis mine):

PHP supports a portable way of locking complete files in an advisory way (which means all accessing programs have to use the same way of locking or it will not work). By default, this function will block until the requested lock is acquired; this may be controlled (on non-Windows platforms) with the LOCK_NB option documented below.

So, as long as all of the processes that are accessing the file use this method of locking, you're fine.

However, if you're writing a static HTML file with a sane webserver, the lock will be ignored. In the middle of the write, if a request comes in, Apache will serve the partially written file. The locks will have no effect on the other process reading the lock.

The only real exception is if you use the special mount option of -o mand on the filesystem to enable mandatory locking (but that's not really used much, and can have a performance penalty).

Have a read on File Locking for some more information. Namely the section under Unix:

This means that cooperating processes may use locks to coordinate access to a file among themselves, but programs are also free to ignore locks and access the file in any way they choose to.

So, in conclusion, using LOCK_EX will create an advisory lock on the file. Any attempt to read the file will block only if the reader respects and/or checks for the lock. If they do not, the lock will be ignored (since it can be).

Try it out. In one process:

file_put_contents('test.txt', 'Foo bar');
$f = fopen('test.txt', 'a+');
if (flock($f, LOCK_EX)) {
    sleep(10);
    fseek($f, 0);
    var_dump(fgets($f, 4048));
    flock($f, LOCK_UN);
}
fclose($f);

And while it's sleeping, call this:

$f = fopen('test.txt', 'a+');
fwrite($f, 'foobar');
fclose($f);

The output will be foobar...

About file_get_contents specifically:

To your other specific question, first off, there is no LOCK_EX parameter to file_get_contents. So you can't pass that in.

Now, looking at the source code, we can see the function file_get_contents defined on line 521. There are no calls to the internal function php_stream_lock as there are when you pass file_put_contents('file', 'txt', LOCK_EX); defined on line 589 of the same file.

So, let's test it, shall we:

In file1.php:

file_put_contents('test.txt', 'Foo bar');
$f = fopen('test.txt', 'a+');
if (flock($f, LOCK_EX)) {
    sleep(10);
    fseek($f, 0);
    var_dump(fgets($f, 4048));
    flock($f, LOCK_UN);
}
fclose($f);

In file2.php:

var_dump(file_get_contents('test.txt'));

When run, file2.php returns immediately. So no, it doesn't appear that file_get_contents respects file locks at all...

like image 81
ircmaxell Avatar answered Nov 14 '22 00:11

ircmaxell


Theory questions work much better on Programmers than here.

At this point, PHP does not support atomic file locking.

Simply put, PHP doesn't support a combined fopen and flock operation, so there will always be a small window of opportunity for another process to lock a file your process has also opened before your process can lock it.

Having said that, flock will, by default, block until the lock is released. Keep in mind ircmaxell's note about advisory locks on Linux/BSD, though.

Note: For the read process, you may want to make it a LOCK_SH instead of LOCK_EX, so that multiple reader threads can lock it at the same time. Writing must always be done using LOCK_EX or risk data corruption.

Note 2: The previous note works because you can only acquire shared locks if no exclusive locks are present, but an exclusive lock requires that no locks of any sort are present before the lock can be acquired.

like image 34
Powerlord Avatar answered Nov 14 '22 00:11

Powerlord