Bug #55629 5.5.x goes into infinite loop and high cpu after error flushing io cache

Description:
This case of an error is a huge read_buffer_size, and OS gives a failure.
So, this query spins/hangs when it should return an error and quit:

---------+------+-----------+-----------------------------------------------------
 Command | Time | State     | Info
---------+------+-----------+-----------------------------------------------------
 Query   |  282 | executing | select repeat('a',@@global.read_buffer_size) into ou
 Query   |    0 | NULL      | show processlist
---------+------+-----------+-----------------------------------------------------

For this specific failure, filemon shows us:

18:27:09.453	mysqld.exe:3588	IRP_MJ_WRITE 	\\xp64\h$\outddd.txt	INSUFFICIENT RESOURCES	Offset: 0 Length: 134219775	
18:27:09.453	mysqld.exe:3588	FASTIO_WRITE	H:\outddd.txt	FAILURE	Offset: 0 Length: 134219775	

5.1.49 gave an error and didn't hang, which is friendly:
ERROR 3 (HY000): Error writing file '\\xp64\h$\outddds.txt' (Errcode: 22)

How to repeat:
start server like this:

mysqld.exe --no-defaults --console --skip-gr --skip-na --max-allowed-packet=1G 
--read-buffer-size=128M

make sure there exists some usable windows share, and run a query to save a file into it:

select repeat('a',@@global.read_buffer_size) 
into outfile '\\\\xp64\\h$\\outddd.txt'

A patch for this bug has been committed. After review, it may
be pushed to the relevant source trees for release in the next
version. You can access the patch from:

  http://lists.mysql.com/commits/119969

3092 Vladislav Vaintroub	2010-10-05
      Bug#55629 5.5.x goes into infinite loop and high cpu after
      error flushing io cache 
      
      The reason for the error was incorrect return code from
      my_win_write() in case of error on 64 bit Windows.
      
      Error should be  indicated by return code 
      (size_t)-1 == 2^64 -1, but due to cast it was
      (DWORD)-1 = 2^32 -1
      
      The caller of this function would fail to recognize the error
      and continue looping.
      
      Fix is to return correct error code (size_t)-1 in case of error
      as expected by caller.
      
      Also minimal cleanup is done : my_win_write() now uses
      the same parameter checks  as related functions (0 and
      overflow handling for count parameterr).

A patch for this bug has been committed. After review, it may
be pushed to the relevant source trees for release in the next
version. You can access the patch from:

  http://lists.mysql.com/commits/119988

3092 Vladislav Vaintroub	2010-10-05
      Bug#55629 5.5.x goes into infinite loop and high cpu after
      error flushing io cache 
      
      The reason for the error was incorrect return code from
      my_win_write() in case of error on 64 bit Windows.
      
      Error should be  indicated by return code 
      (size_t)-1 == 2^64 -1, but due to cast it was
      (DWORD)-1 = 2^32 -1
      
      The caller of this function would fail to recognize the error
      and continue looping.
      
      Fix is to return correct error code (size_t)-1 in case of error
      as expected by caller.
      
      Also minimal cleanup is done : my_win_write() now uses
      the same parameter checks  as related functions (0 and
      overflow handling for count parameter).

queued to 5.5-bugteam, trunk-merge

Pushed into mysql-5.5 5.5.7-rc (revid:sunanda.menon@sun.com-20101109182959-otkxq8vo2dcd13la) (version source revid:sunanda.menon@sun.com-20101109182959-otkxq8vo2dcd13la) (merge vers: 5.5.7-rc) (pib:21)

Pushed into mysql-trunk 5.6.99-m5 (revid:alexander.nozdrin@oracle.com-20101113155825-czmva9kg4n31anmu) (version source revid:alexander.nozdrin@oracle.com-20101113152450-2zzcm50e7i4j35v7) (merge vers: 5.6.1-m4) (pib:21)

Pushed into mysql-next-mr (revid:alexander.nozdrin@oracle.com-20101113160336-atmtmfb3mzm4pz4i) (version source revid:vasil.dimov@oracle.com-20100629074804-359l9m9gniauxr94) (pib:21)

An entry has been added to the 5.5.7 changelog:

        The server entered an infinite loop with high CPU utilization
        after an error occurred during flushing of the IO cache.