1=pod
2
3=head1 NAME
4
5BIO_should_retry, BIO_should_read, BIO_should_write,
6BIO_should_io_special, BIO_retry_type, BIO_should_retry,
7BIO_get_retry_BIO, BIO_get_retry_reason - BIO retry functions
8
9=head1 SYNOPSIS
10
11 #include <openssl/bio.h>
12
13 #define BIO_should_read(a)		((a)->flags & BIO_FLAGS_READ)
14 #define BIO_should_write(a)		((a)->flags & BIO_FLAGS_WRITE)
15 #define BIO_should_io_special(a)	((a)->flags & BIO_FLAGS_IO_SPECIAL)
16 #define BIO_retry_type(a)		((a)->flags & BIO_FLAGS_RWS)
17 #define BIO_should_retry(a)		((a)->flags & BIO_FLAGS_SHOULD_RETRY)
18
19 #define BIO_FLAGS_READ		0x01
20 #define BIO_FLAGS_WRITE	0x02
21 #define BIO_FLAGS_IO_SPECIAL	0x04
22 #define BIO_FLAGS_RWS (BIO_FLAGS_READ|BIO_FLAGS_WRITE|BIO_FLAGS_IO_SPECIAL)
23 #define BIO_FLAGS_SHOULD_RETRY	0x08
24
25 BIO *	BIO_get_retry_BIO(BIO *bio, int *reason);
26 int	BIO_get_retry_reason(BIO *bio);
27
28=head1 DESCRIPTION
29
30These functions determine why a BIO is not able to read or write data.
31They will typically be called after a failed BIO_read() or BIO_write()
32call.
33
34BIO_should_retry() is true if the call that produced this condition
35should then be retried at a later time.
36
37If BIO_should_retry() is false then the cause is an error condition.
38
39BIO_should_read() is true if the cause of the condition is that a BIO
40needs to read data.
41
42BIO_should_write() is true if the cause of the condition is that a BIO
43needs to read data.
44
45BIO_should_io_special() is true if some "special" condition, that is a
46reason other than reading or writing is the cause of the condition.
47
48BIO_retry_type() returns a mask of the cause of a retry condition
49consisting of the values B<BIO_FLAGS_READ>, B<BIO_FLAGS_WRITE>,
50B<BIO_FLAGS_IO_SPECIAL> though current BIO types will only set one of
51these.
52
53BIO_get_retry_BIO() determines the precise reason for the special
54condition, it returns the BIO that caused this condition and if 
55B<reason> is not NULL it contains the reason code. The meaning of
56the reason code and the action that should be taken depends on
57the type of BIO that resulted in this condition.
58
59BIO_get_retry_reason() returns the reason for a special condition if
60passed the relevant BIO, for example as returned by BIO_get_retry_BIO().
61
62=head1 NOTES
63
64If BIO_should_retry() returns false then the precise "error condition"
65depends on the BIO type that caused it and the return code of the BIO
66operation. For example if a call to BIO_read() on a socket BIO returns
670 and BIO_should_retry() is false then the cause will be that the
68connection closed. A similar condition on a file BIO will mean that it
69has reached EOF. Some BIO types may place additional information on
70the error queue. For more details see the individual BIO type manual
71pages.
72
73If the underlying I/O structure is in a blocking mode almost all current
74BIO types will not request a retry, because the underlying I/O
75calls will not. If the application knows that the BIO type will never
76signal a retry then it need not call BIO_should_retry() after a failed
77BIO I/O call. This is typically done with file BIOs.
78
79SSL BIOs are the only current exception to this rule: they can request a
80retry even if the underlying I/O structure is blocking, if a handshake
81occurs during a call to BIO_read(). An application can retry the failed
82call immediately or avoid this situation by setting SSL_MODE_AUTO_RETRY
83on the underlying SSL structure.
84
85While an application may retry a failed non blocking call immediately
86this is likely to be very inefficient because the call will fail
87repeatedly until data can be processed or is available. An application
88will normally wait until the necessary condition is satisfied. How
89this is done depends on the underlying I/O structure.
90
91For example if the cause is ultimately a socket and BIO_should_read()
92is true then a call to select() may be made to wait until data is
93available and then retry the BIO operation. By combining the retry
94conditions of several non blocking BIOs in a single select() call
95it is possible to service several BIOs in a single thread, though
96the performance may be poor if SSL BIOs are present because long delays
97can occur during the initial handshake process. 
98
99It is possible for a BIO to block indefinitely if the underlying I/O
100structure cannot process or return any data. This depends on the behaviour of
101the platforms I/O functions. This is often not desirable: one solution
102is to use non blocking I/O and use a timeout on the select() (or
103equivalent) call.
104
105=head1 BUGS
106
107The OpenSSL ASN1 functions cannot gracefully deal with non blocking I/O:
108that is they cannot retry after a partial read or write. This is usually
109worked around by only passing the relevant data to ASN1 functions when
110the entire structure can be read or written.
111
112=head1 SEE ALSO
113
114TBA
115