You can not select more than 25 topics Topics must start with a letter or number, can include dashes ('-') and can be up to 35 characters long.
 
 
 
 
 
 

167 lines
8.4 KiB

From 5bebcd06287be9024f0fba25f350393f02e050c1 Mon Sep 17 00:00:00 2001
From: Willy Tarreau <w@1wt.eu>
Date: Thu, 10 Jul 2014 19:06:10 +0200
Subject: [PATCH 24/25] BUG/MAJOR: http: correctly rewind the request body
after start of forwarding
Daniel Dubovik reported an interesting bug showing that the request body
processing was still not 100% fixed. If a POST request contained short
enough data to be forwarded at once before trying to establish the
connection to the server, we had no way to correctly rewind the body.
The first visible case is that balancing on a header does not always work
on such POST requests since the header cannot be found. But there are even
nastier implications which are that http-send-name-header would apply to
the wrong location and possibly even affect part of the request's body
due to an incorrect rewinding.
There are two options to fix the problem :
- first one is to force the HTTP_MSG_F_WAIT_CONN flag on all hash-based
balancing algorithms and http-send-name-header, but there's always a
risk that any new algorithm forgets to set it ;
- the second option is to account for the amount of skipped data before
the connection establishes so that we always know the position of the
request's body relative to the buffer's origin.
The second option is much more reliable and fits very well in the spirit
of the past changes to fix forwarding. Indeed, at the moment we have
msg->sov which points to the start of the body before headers are forwarded
and which equals zero afterwards (so it still points to the start of the
body before forwarding data). A minor change consists in always making it
point to the start of the body even after data have been forwarded. It means
that it can get a negative value (so we need to change its type to signed)..
In order to avoid wrapping, we only do this as long as the other side of
the buffer is not connected yet.
Doing this definitely fixes the issues above for the requests. Since the
response cannot be rewound we don't need to perform any change there.
This bug was introduced/remained unfixed in 1.5-dev23 so the fix must be
backported to 1.5.
(cherry picked from commit bb2e669f9e73531ac9cc9277b40066b701eec918)
---
doc/internals/body-parsing.txt | 20 +++++++++++++-------
include/types/proto_http.h | 11 ++++++-----
src/proto_http.c | 9 +++++++--
3 files changed, 26 insertions(+), 14 deletions(-)
diff --git a/doc/internals/body-parsing.txt b/doc/internals/body-parsing.txt
index e9c8b4b..5baa549 100644
--- a/doc/internals/body-parsing.txt
+++ b/doc/internals/body-parsing.txt
@@ -67,12 +67,17 @@ msg.next : points to the next byte to inspect. This offset is automatically
automatically adjusted to the number of bytes already inspected.
msg.sov : start of value. First character of the header's value in the header
- states, start of the body in the data states until headers are
- forwarded. This offset is automatically adjusted when inserting or
- removing some headers. In data states, it always constains the size
- of the whole HTTP headers (including the trailing CRLF) that needs
- to be forwarded before the first byte of body. Once the headers are
- forwarded, this value drops to zero.
+ states, start of the body in the data states. Strictly positive
+ values indicate that headers were not forwarded yet (<buf.p> is
+ before the start of the body), and null or positive values are seen
+ after headers are forwarded (<buf.p> is at or past the start of the
+ body). The value stops changing when data start to leave the buffer
+ (in order to avoid integer overflows). So the maximum possible range
+ is -<buf.size> to +<buf.size>. This offset is automatically adjusted
+ when inserting or removing some headers. It is useful to rewind the
+ request buffer to the beginning of the body at any phase. The
+ response buffer does not really use it since it is immediately
+ forwarded to the client.
msg.sol : start of line. Points to the beginning of the current header line
while parsing headers. It is cleared to zero in the BODY state,
@@ -97,7 +102,8 @@ msg.eol : end of line. Points to the CRLF or LF of the current header line
states nor by forwarding.
The beginning of the message headers can always be found this way even after
-headers have been forwarded :
+headers or data have been forwarded, provided that everything is still present
+in the buffer :
headers = buf.p + msg->sov - msg->eoh - msg->eol
diff --git a/include/types/proto_http.h b/include/types/proto_http.h
index 12e1141..c53c7fd 100644
--- a/include/types/proto_http.h
+++ b/include/types/proto_http.h
@@ -329,7 +329,8 @@ enum {
* message or a response message.
*
* The values there are a little bit obscure, because their meaning can change
- * during the parsing :
+ * during the parsing. Please read carefully doc/internal/body-parsing.txt if
+ * you need to manipulate them. Quick reminder :
*
* - eoh (End of Headers) : relative offset in the buffer of first byte that
* is not part of a completely processed header.
@@ -344,9 +345,9 @@ enum {
* - sov (start of value) : Before HTTP_MSG_BODY, points to the value of
* the header being parsed. Starting from
* HTTP_MSG_BODY, will point to the start of the
- * body (relative to buffer's origin), or to data
- * following a chunk size. Thus <sov> bytes of
- * headers will have to be sent only once.
+ * body (relative to buffer's origin). It can be
+ * negative when forwarding data. It stops growing
+ * once data start to leave the buffer.
*
* - next (parse pointer) : next relative byte to be parsed. Always points
* to a byte matching the current state.
@@ -372,7 +373,7 @@ struct http_msg {
/* 6 bytes unused here */
struct channel *chn; /* pointer to the channel transporting the message */
unsigned int next; /* pointer to next byte to parse, relative to buf->p */
- unsigned int sov; /* current header: start of value */
+ int sov; /* current header: start of value ; data: start of body */
unsigned int eoh; /* End Of Headers, relative to buffer */
unsigned int sol; /* start of current line during parsing otherwise zero */
unsigned int eol; /* end of line */
diff --git a/src/proto_http.c b/src/proto_http.c
index 4a862b0..94afed7 100644
--- a/src/proto_http.c
+++ b/src/proto_http.c
@@ -5315,7 +5315,7 @@ int http_request_forward_body(struct session *s, struct channel *req, int an_bit
* an "Expect: 100-continue" header.
*/
- if (msg->sov) {
+ if (msg->sov > 0) {
/* we have msg->sov which points to the first byte of message
* body, and req->buf.p still points to the beginning of the
* message. We forward the headers now, as we don't need them
@@ -5429,6 +5429,8 @@ int http_request_forward_body(struct session *s, struct channel *req, int an_bit
* such as last chunk of data or trailers.
*/
b_adv(req->buf, msg->next);
+ if (unlikely(!(s->rep->flags & CF_READ_ATTACHED)))
+ msg->sov -= msg->next;
msg->next = 0;
/* for keep-alive we don't want to forward closes on DONE */
@@ -5479,6 +5481,9 @@ int http_request_forward_body(struct session *s, struct channel *req, int an_bit
missing_data:
/* we may have some pending data starting at req->buf->p */
b_adv(req->buf, msg->next);
+ if (unlikely(!(s->rep->flags & CF_READ_ATTACHED)))
+ msg->sov -= msg->next + MIN(msg->chunk_len, req->buf->i);
+
msg->next = 0;
msg->chunk_len -= channel_forward(req, msg->chunk_len);
@@ -6493,7 +6498,7 @@ int http_response_forward_body(struct session *s, struct channel *res, int an_bi
/* in most states, we should abort in case of early close */
channel_auto_close(res);
- if (msg->sov) {
+ if (msg->sov > 0) {
/* we have msg->sov which points to the first byte of message
* body, and res->buf.p still points to the beginning of the
* message. We forward the headers now, as we don't need them
--
1.8.5.5