curl/lib/request.c

Ignoring revisions in .git-blame-ignore-revs. Click here to bypass and see the normal blame view.

455 lines
13 KiB
C
Raw Normal View History

/***************************************************************************
* _ _ ____ _
* Project ___| | | | _ \| |
* / __| | | | |_) | |
* | (__| |_| | _ <| |___
* \___|\___/|_| \_\_____|
*
* Copyright (C) Daniel Stenberg, <daniel@haxx.se>, et al.
*
* This software is licensed as described in the file COPYING, which
* you should have received as part of this distribution. The terms
* are also available at https://curl.se/docs/copyright.html.
*
* You may opt to use, copy, modify, merge, publish, distribute and/or sell
* copies of the Software, and permit persons to whom the Software is
* furnished to do so, under the terms of the COPYING file.
*
* This software is distributed on an "AS IS" basis, WITHOUT WARRANTY OF ANY
* KIND, either express or implied.
*
* SPDX-License-Identifier: curl
*
***************************************************************************/
#include "curl_setup.h"
#include "urldata.h"
#include "cfilters.h"
#include "dynbuf.h"
#include "doh.h"
lib: Curl_read/Curl_write clarifications - replace `Curl_read()`, `Curl_write()` and `Curl_nwrite()` to clarify when and at what level they operate - send/recv of transfer related data is now done via `Curl_xfer_send()/Curl_xfer_recv()` which no longer has socket/socketindex as parameter. It decides on the transfer setup of `conn->sockfd` and `conn->writesockfd` on which connection filter chain to operate. - send/recv on a specific connection filter chain is done via `Curl_conn_send()/Curl_conn_recv()` which get the socket index as parameter. - rename `Curl_setup_transfer()` to `Curl_xfer_setup()` for naming consistency - clarify that the special CURLE_AGAIN hangling to return `CURLE_OK` with length 0 only applies to `Curl_xfer_send()` and CURLE_AGAIN is returned by all other send() variants. - fix a bug in websocket `curl_ws_recv()` that mixed up data when it arrived in more than a single chunk (to be made into a sperate PR, also) Added as documented [in CLIENT-READER.md](https://github.com/curl/curl/blob/5b1f31dfbab8aef467c419c68aa06dc738cb75d4/docs/CLIENT-READERS.md). - old `Curl_buffer_send()` completely replaced by new `Curl_req_send()` - old `Curl_fillreadbuffer()` replaced with `Curl_client_read()` - HTTP chunked uploads are now formatted in a client reader added when needed. - FTP line-end conversions are done in a client reader added when needed. - when sending requests headers, remaining buffer space is filled with body data for sending in "one go". This is independent of the request body size. Resolves #12938 as now small and large requests have the same code path. Changes done to test cases: - test513: now fails before sending request headers as this initial "client read" triggers the setup fault. Behaves now the same as in hyper build - test547, test555, test1620: fix the length check in the lib code to only fail for reads *smaller* than expected. This was a bug in the test code that never triggered in the old implementation. Closes #12969
2024-02-15 23:22:53 +08:00
#include "multiif.h"
lib: Curl_read/Curl_write clarifications - replace `Curl_read()`, `Curl_write()` and `Curl_nwrite()` to clarify when and at what level they operate - send/recv of transfer related data is now done via `Curl_xfer_send()/Curl_xfer_recv()` which no longer has socket/socketindex as parameter. It decides on the transfer setup of `conn->sockfd` and `conn->writesockfd` on which connection filter chain to operate. - send/recv on a specific connection filter chain is done via `Curl_conn_send()/Curl_conn_recv()` which get the socket index as parameter. - rename `Curl_setup_transfer()` to `Curl_xfer_setup()` for naming consistency - clarify that the special CURLE_AGAIN hangling to return `CURLE_OK` with length 0 only applies to `Curl_xfer_send()` and CURLE_AGAIN is returned by all other send() variants. - fix a bug in websocket `curl_ws_recv()` that mixed up data when it arrived in more than a single chunk The method for sending not just raw bytes, but bytes that are either "headers" or "body". The send abstraction stack, to to bottom, now is: * `Curl_req_send()`: has parameter to indicate amount of header bytes, buffers all data. * `Curl_xfer_send()`: knows on which socket index to send, returns amount of bytes sent. * `Curl_conn_send()`: called with socket index, returns amount of bytes sent. In addition there is `Curl_req_flush()` for writing out all buffered bytes. `Curl_req_send()` is active for requests without body, `Curl_buffer_send()` still being used for others. This is because the special quirks need to be addressed in future parts: * `expect-100` handling * `Curl_fillreadbuffer()` needs to add directly to the new `data->req.sendbuf` * special body handlings, like `chunked` encodings and line end conversions will be moved into something like a Client Reader. In functions of the pattern `CURLcode xxx_send(..., ssize_t *written)`, replace the `ssize_t` with a `size_t`. It makes no sense to allow for negative values as the returned `CURLcode` already specifies error conditions. This allows easier handling of lengths without casting. Closes #12964
2024-02-15 23:22:53 +08:00
#include "progress.h"
#include "request.h"
#include "sendf.h"
lib: Curl_read/Curl_write clarifications - replace `Curl_read()`, `Curl_write()` and `Curl_nwrite()` to clarify when and at what level they operate - send/recv of transfer related data is now done via `Curl_xfer_send()/Curl_xfer_recv()` which no longer has socket/socketindex as parameter. It decides on the transfer setup of `conn->sockfd` and `conn->writesockfd` on which connection filter chain to operate. - send/recv on a specific connection filter chain is done via `Curl_conn_send()/Curl_conn_recv()` which get the socket index as parameter. - rename `Curl_setup_transfer()` to `Curl_xfer_setup()` for naming consistency - clarify that the special CURLE_AGAIN hangling to return `CURLE_OK` with length 0 only applies to `Curl_xfer_send()` and CURLE_AGAIN is returned by all other send() variants. - fix a bug in websocket `curl_ws_recv()` that mixed up data when it arrived in more than a single chunk The method for sending not just raw bytes, but bytes that are either "headers" or "body". The send abstraction stack, to to bottom, now is: * `Curl_req_send()`: has parameter to indicate amount of header bytes, buffers all data. * `Curl_xfer_send()`: knows on which socket index to send, returns amount of bytes sent. * `Curl_conn_send()`: called with socket index, returns amount of bytes sent. In addition there is `Curl_req_flush()` for writing out all buffered bytes. `Curl_req_send()` is active for requests without body, `Curl_buffer_send()` still being used for others. This is because the special quirks need to be addressed in future parts: * `expect-100` handling * `Curl_fillreadbuffer()` needs to add directly to the new `data->req.sendbuf` * special body handlings, like `chunked` encodings and line end conversions will be moved into something like a Client Reader. In functions of the pattern `CURLcode xxx_send(..., ssize_t *written)`, replace the `ssize_t` with a `size_t`. It makes no sense to allow for negative values as the returned `CURLcode` already specifies error conditions. This allows easier handling of lengths without casting. Closes #12964
2024-02-15 23:22:53 +08:00
#include "transfer.h"
#include "url.h"
/* The last 3 #include files should be in this order */
#include "curl_printf.h"
#include "curl_memory.h"
#include "memdebug.h"
void Curl_req_init(struct SingleRequest *req)
{
memset(req, 0, sizeof(*req));
}
CURLcode Curl_req_soft_reset(struct SingleRequest *req,
struct Curl_easy *data)
{
CURLcode result;
req->done = FALSE;
req->upload_done = FALSE;
req->download_done = FALSE;
req->ignorebody = FALSE;
req->shutdown = FALSE;
req->bytecount = 0;
req->writebytecount = 0;
req->header = TRUE; /* assume header */
req->headerline = 0;
req->headerbytecount = 0;
req->allheadercount = 0;
req->deductheadercount = 0;
result = Curl_client_start(data);
if(result)
return result;
lib: Curl_read/Curl_write clarifications - replace `Curl_read()`, `Curl_write()` and `Curl_nwrite()` to clarify when and at what level they operate - send/recv of transfer related data is now done via `Curl_xfer_send()/Curl_xfer_recv()` which no longer has socket/socketindex as parameter. It decides on the transfer setup of `conn->sockfd` and `conn->writesockfd` on which connection filter chain to operate. - send/recv on a specific connection filter chain is done via `Curl_conn_send()/Curl_conn_recv()` which get the socket index as parameter. - rename `Curl_setup_transfer()` to `Curl_xfer_setup()` for naming consistency - clarify that the special CURLE_AGAIN hangling to return `CURLE_OK` with length 0 only applies to `Curl_xfer_send()` and CURLE_AGAIN is returned by all other send() variants. - fix a bug in websocket `curl_ws_recv()` that mixed up data when it arrived in more than a single chunk The method for sending not just raw bytes, but bytes that are either "headers" or "body". The send abstraction stack, to to bottom, now is: * `Curl_req_send()`: has parameter to indicate amount of header bytes, buffers all data. * `Curl_xfer_send()`: knows on which socket index to send, returns amount of bytes sent. * `Curl_conn_send()`: called with socket index, returns amount of bytes sent. In addition there is `Curl_req_flush()` for writing out all buffered bytes. `Curl_req_send()` is active for requests without body, `Curl_buffer_send()` still being used for others. This is because the special quirks need to be addressed in future parts: * `expect-100` handling * `Curl_fillreadbuffer()` needs to add directly to the new `data->req.sendbuf` * special body handlings, like `chunked` encodings and line end conversions will be moved into something like a Client Reader. In functions of the pattern `CURLcode xxx_send(..., ssize_t *written)`, replace the `ssize_t` with a `size_t`. It makes no sense to allow for negative values as the returned `CURLcode` already specifies error conditions. This allows easier handling of lengths without casting. Closes #12964
2024-02-15 23:22:53 +08:00
if(!req->sendbuf_init) {
Curl_bufq_init2(&req->sendbuf, data->set.upload_buffer_size, 1,
BUFQ_OPT_SOFT_LIMIT);
req->sendbuf_init = TRUE;
}
else {
Curl_bufq_reset(&req->sendbuf);
if(data->set.upload_buffer_size != req->sendbuf.chunk_size) {
Curl_bufq_free(&req->sendbuf);
Curl_bufq_init2(&req->sendbuf, data->set.upload_buffer_size, 1,
BUFQ_OPT_SOFT_LIMIT);
}
}
return CURLE_OK;
}
CURLcode Curl_req_start(struct SingleRequest *req,
struct Curl_easy *data)
{
req->start = Curl_now();
return Curl_req_soft_reset(req, data);
}
static CURLcode req_flush(struct Curl_easy *data);
CURLcode Curl_req_done(struct SingleRequest *req,
struct Curl_easy *data, bool aborted)
{
(void)req;
lib: Curl_read/Curl_write clarifications - replace `Curl_read()`, `Curl_write()` and `Curl_nwrite()` to clarify when and at what level they operate - send/recv of transfer related data is now done via `Curl_xfer_send()/Curl_xfer_recv()` which no longer has socket/socketindex as parameter. It decides on the transfer setup of `conn->sockfd` and `conn->writesockfd` on which connection filter chain to operate. - send/recv on a specific connection filter chain is done via `Curl_conn_send()/Curl_conn_recv()` which get the socket index as parameter. - rename `Curl_setup_transfer()` to `Curl_xfer_setup()` for naming consistency - clarify that the special CURLE_AGAIN hangling to return `CURLE_OK` with length 0 only applies to `Curl_xfer_send()` and CURLE_AGAIN is returned by all other send() variants. - fix a bug in websocket `curl_ws_recv()` that mixed up data when it arrived in more than a single chunk The method for sending not just raw bytes, but bytes that are either "headers" or "body". The send abstraction stack, to to bottom, now is: * `Curl_req_send()`: has parameter to indicate amount of header bytes, buffers all data. * `Curl_xfer_send()`: knows on which socket index to send, returns amount of bytes sent. * `Curl_conn_send()`: called with socket index, returns amount of bytes sent. In addition there is `Curl_req_flush()` for writing out all buffered bytes. `Curl_req_send()` is active for requests without body, `Curl_buffer_send()` still being used for others. This is because the special quirks need to be addressed in future parts: * `expect-100` handling * `Curl_fillreadbuffer()` needs to add directly to the new `data->req.sendbuf` * special body handlings, like `chunked` encodings and line end conversions will be moved into something like a Client Reader. In functions of the pattern `CURLcode xxx_send(..., ssize_t *written)`, replace the `ssize_t` with a `size_t`. It makes no sense to allow for negative values as the returned `CURLcode` already specifies error conditions. This allows easier handling of lengths without casting. Closes #12964
2024-02-15 23:22:53 +08:00
if(!aborted)
(void)req_flush(data);
lib: Curl_read/Curl_write clarifications - replace `Curl_read()`, `Curl_write()` and `Curl_nwrite()` to clarify when and at what level they operate - send/recv of transfer related data is now done via `Curl_xfer_send()/Curl_xfer_recv()` which no longer has socket/socketindex as parameter. It decides on the transfer setup of `conn->sockfd` and `conn->writesockfd` on which connection filter chain to operate. - send/recv on a specific connection filter chain is done via `Curl_conn_send()/Curl_conn_recv()` which get the socket index as parameter. - rename `Curl_setup_transfer()` to `Curl_xfer_setup()` for naming consistency - clarify that the special CURLE_AGAIN hangling to return `CURLE_OK` with length 0 only applies to `Curl_xfer_send()` and CURLE_AGAIN is returned by all other send() variants. - fix a bug in websocket `curl_ws_recv()` that mixed up data when it arrived in more than a single chunk (to be made into a sperate PR, also) Added as documented [in CLIENT-READER.md](https://github.com/curl/curl/blob/5b1f31dfbab8aef467c419c68aa06dc738cb75d4/docs/CLIENT-READERS.md). - old `Curl_buffer_send()` completely replaced by new `Curl_req_send()` - old `Curl_fillreadbuffer()` replaced with `Curl_client_read()` - HTTP chunked uploads are now formatted in a client reader added when needed. - FTP line-end conversions are done in a client reader added when needed. - when sending requests headers, remaining buffer space is filled with body data for sending in "one go". This is independent of the request body size. Resolves #12938 as now small and large requests have the same code path. Changes done to test cases: - test513: now fails before sending request headers as this initial "client read" triggers the setup fault. Behaves now the same as in hyper build - test547, test555, test1620: fix the length check in the lib code to only fail for reads *smaller* than expected. This was a bug in the test code that never triggered in the old implementation. Closes #12969
2024-02-15 23:22:53 +08:00
Curl_client_reset(data);
#ifndef CURL_DISABLE_DOH
Curl_doh_close(data);
#endif
return CURLE_OK;
}
void Curl_req_hard_reset(struct SingleRequest *req, struct Curl_easy *data)
{
struct curltime t0 = {0, 0};
lib: Curl_read/Curl_write clarifications - replace `Curl_read()`, `Curl_write()` and `Curl_nwrite()` to clarify when and at what level they operate - send/recv of transfer related data is now done via `Curl_xfer_send()/Curl_xfer_recv()` which no longer has socket/socketindex as parameter. It decides on the transfer setup of `conn->sockfd` and `conn->writesockfd` on which connection filter chain to operate. - send/recv on a specific connection filter chain is done via `Curl_conn_send()/Curl_conn_recv()` which get the socket index as parameter. - rename `Curl_setup_transfer()` to `Curl_xfer_setup()` for naming consistency - clarify that the special CURLE_AGAIN hangling to return `CURLE_OK` with length 0 only applies to `Curl_xfer_send()` and CURLE_AGAIN is returned by all other send() variants. - fix a bug in websocket `curl_ws_recv()` that mixed up data when it arrived in more than a single chunk The method for sending not just raw bytes, but bytes that are either "headers" or "body". The send abstraction stack, to to bottom, now is: * `Curl_req_send()`: has parameter to indicate amount of header bytes, buffers all data. * `Curl_xfer_send()`: knows on which socket index to send, returns amount of bytes sent. * `Curl_conn_send()`: called with socket index, returns amount of bytes sent. In addition there is `Curl_req_flush()` for writing out all buffered bytes. `Curl_req_send()` is active for requests without body, `Curl_buffer_send()` still being used for others. This is because the special quirks need to be addressed in future parts: * `expect-100` handling * `Curl_fillreadbuffer()` needs to add directly to the new `data->req.sendbuf` * special body handlings, like `chunked` encodings and line end conversions will be moved into something like a Client Reader. In functions of the pattern `CURLcode xxx_send(..., ssize_t *written)`, replace the `ssize_t` with a `size_t`. It makes no sense to allow for negative values as the returned `CURLcode` already specifies error conditions. This allows easier handling of lengths without casting. Closes #12964
2024-02-15 23:22:53 +08:00
/* This is a bit ugly. `req->p` is a union and we assume we can
* free this safely without leaks. */
Curl_safefree(req->p.ftp);
Curl_safefree(req->newurl);
lib: Curl_read/Curl_write clarifications - replace `Curl_read()`, `Curl_write()` and `Curl_nwrite()` to clarify when and at what level they operate - send/recv of transfer related data is now done via `Curl_xfer_send()/Curl_xfer_recv()` which no longer has socket/socketindex as parameter. It decides on the transfer setup of `conn->sockfd` and `conn->writesockfd` on which connection filter chain to operate. - send/recv on a specific connection filter chain is done via `Curl_conn_send()/Curl_conn_recv()` which get the socket index as parameter. - rename `Curl_setup_transfer()` to `Curl_xfer_setup()` for naming consistency - clarify that the special CURLE_AGAIN hangling to return `CURLE_OK` with length 0 only applies to `Curl_xfer_send()` and CURLE_AGAIN is returned by all other send() variants. - fix a bug in websocket `curl_ws_recv()` that mixed up data when it arrived in more than a single chunk (to be made into a sperate PR, also) Added as documented [in CLIENT-READER.md](https://github.com/curl/curl/blob/5b1f31dfbab8aef467c419c68aa06dc738cb75d4/docs/CLIENT-READERS.md). - old `Curl_buffer_send()` completely replaced by new `Curl_req_send()` - old `Curl_fillreadbuffer()` replaced with `Curl_client_read()` - HTTP chunked uploads are now formatted in a client reader added when needed. - FTP line-end conversions are done in a client reader added when needed. - when sending requests headers, remaining buffer space is filled with body data for sending in "one go". This is independent of the request body size. Resolves #12938 as now small and large requests have the same code path. Changes done to test cases: - test513: now fails before sending request headers as this initial "client read" triggers the setup fault. Behaves now the same as in hyper build - test547, test555, test1620: fix the length check in the lib code to only fail for reads *smaller* than expected. This was a bug in the test code that never triggered in the old implementation. Closes #12969
2024-02-15 23:22:53 +08:00
Curl_client_reset(data);
if(req->sendbuf_init)
Curl_bufq_reset(&req->sendbuf);
#ifndef CURL_DISABLE_DOH
Curl_doh_close(data);
#endif
/* Can no longer memset() this struct as we need to keep some state */
req->size = -1;
req->maxdownload = -1;
req->bytecount = 0;
req->writebytecount = 0;
req->start = t0;
req->headerbytecount = 0;
req->allheadercount = 0;
req->deductheadercount = 0;
req->headerline = 0;
req->offset = 0;
req->httpcode = 0;
req->keepon = 0;
req->upgr101 = UPGR101_INIT;
req->timeofdoc = 0;
req->location = NULL;
req->newurl = NULL;
#ifndef CURL_DISABLE_COOKIES
req->setcookies = 0;
#endif
req->header = FALSE;
req->content_range = FALSE;
req->download_done = FALSE;
req->eos_written = FALSE;
req->eos_read = FALSE;
req->upload_done = FALSE;
req->upload_aborted = FALSE;
req->ignorebody = FALSE;
req->http_bodyless = FALSE;
req->chunk = FALSE;
req->ignore_cl = FALSE;
req->upload_chunky = FALSE;
req->getheader = FALSE;
req->no_body = data->set.opt_no_body;
req->authneg = FALSE;
req->shutdown = FALSE;
#ifdef USE_HYPER
req->bodywritten = FALSE;
#endif
}
void Curl_req_free(struct SingleRequest *req, struct Curl_easy *data)
{
/* This is a bit ugly. `req->p` is a union and we assume we can
* free this safely without leaks. */
Curl_safefree(req->p.ftp);
Curl_safefree(req->newurl);
lib: Curl_read/Curl_write clarifications - replace `Curl_read()`, `Curl_write()` and `Curl_nwrite()` to clarify when and at what level they operate - send/recv of transfer related data is now done via `Curl_xfer_send()/Curl_xfer_recv()` which no longer has socket/socketindex as parameter. It decides on the transfer setup of `conn->sockfd` and `conn->writesockfd` on which connection filter chain to operate. - send/recv on a specific connection filter chain is done via `Curl_conn_send()/Curl_conn_recv()` which get the socket index as parameter. - rename `Curl_setup_transfer()` to `Curl_xfer_setup()` for naming consistency - clarify that the special CURLE_AGAIN hangling to return `CURLE_OK` with length 0 only applies to `Curl_xfer_send()` and CURLE_AGAIN is returned by all other send() variants. - fix a bug in websocket `curl_ws_recv()` that mixed up data when it arrived in more than a single chunk The method for sending not just raw bytes, but bytes that are either "headers" or "body". The send abstraction stack, to to bottom, now is: * `Curl_req_send()`: has parameter to indicate amount of header bytes, buffers all data. * `Curl_xfer_send()`: knows on which socket index to send, returns amount of bytes sent. * `Curl_conn_send()`: called with socket index, returns amount of bytes sent. In addition there is `Curl_req_flush()` for writing out all buffered bytes. `Curl_req_send()` is active for requests without body, `Curl_buffer_send()` still being used for others. This is because the special quirks need to be addressed in future parts: * `expect-100` handling * `Curl_fillreadbuffer()` needs to add directly to the new `data->req.sendbuf` * special body handlings, like `chunked` encodings and line end conversions will be moved into something like a Client Reader. In functions of the pattern `CURLcode xxx_send(..., ssize_t *written)`, replace the `ssize_t` with a `size_t`. It makes no sense to allow for negative values as the returned `CURLcode` already specifies error conditions. This allows easier handling of lengths without casting. Closes #12964
2024-02-15 23:22:53 +08:00
if(req->sendbuf_init)
Curl_bufq_free(&req->sendbuf);
Curl_client_cleanup(data);
#ifndef CURL_DISABLE_DOH
Curl_doh_cleanup(data);
#endif
}
static CURLcode xfer_send(struct Curl_easy *data,
const char *buf, size_t blen,
size_t hds_len, size_t *pnwritten)
lib: Curl_read/Curl_write clarifications - replace `Curl_read()`, `Curl_write()` and `Curl_nwrite()` to clarify when and at what level they operate - send/recv of transfer related data is now done via `Curl_xfer_send()/Curl_xfer_recv()` which no longer has socket/socketindex as parameter. It decides on the transfer setup of `conn->sockfd` and `conn->writesockfd` on which connection filter chain to operate. - send/recv on a specific connection filter chain is done via `Curl_conn_send()/Curl_conn_recv()` which get the socket index as parameter. - rename `Curl_setup_transfer()` to `Curl_xfer_setup()` for naming consistency - clarify that the special CURLE_AGAIN hangling to return `CURLE_OK` with length 0 only applies to `Curl_xfer_send()` and CURLE_AGAIN is returned by all other send() variants. - fix a bug in websocket `curl_ws_recv()` that mixed up data when it arrived in more than a single chunk The method for sending not just raw bytes, but bytes that are either "headers" or "body". The send abstraction stack, to to bottom, now is: * `Curl_req_send()`: has parameter to indicate amount of header bytes, buffers all data. * `Curl_xfer_send()`: knows on which socket index to send, returns amount of bytes sent. * `Curl_conn_send()`: called with socket index, returns amount of bytes sent. In addition there is `Curl_req_flush()` for writing out all buffered bytes. `Curl_req_send()` is active for requests without body, `Curl_buffer_send()` still being used for others. This is because the special quirks need to be addressed in future parts: * `expect-100` handling * `Curl_fillreadbuffer()` needs to add directly to the new `data->req.sendbuf` * special body handlings, like `chunked` encodings and line end conversions will be moved into something like a Client Reader. In functions of the pattern `CURLcode xxx_send(..., ssize_t *written)`, replace the `ssize_t` with a `size_t`. It makes no sense to allow for negative values as the returned `CURLcode` already specifies error conditions. This allows easier handling of lengths without casting. Closes #12964
2024-02-15 23:22:53 +08:00
{
CURLcode result = CURLE_OK;
bool eos = FALSE;
lib: Curl_read/Curl_write clarifications - replace `Curl_read()`, `Curl_write()` and `Curl_nwrite()` to clarify when and at what level they operate - send/recv of transfer related data is now done via `Curl_xfer_send()/Curl_xfer_recv()` which no longer has socket/socketindex as parameter. It decides on the transfer setup of `conn->sockfd` and `conn->writesockfd` on which connection filter chain to operate. - send/recv on a specific connection filter chain is done via `Curl_conn_send()/Curl_conn_recv()` which get the socket index as parameter. - rename `Curl_setup_transfer()` to `Curl_xfer_setup()` for naming consistency - clarify that the special CURLE_AGAIN hangling to return `CURLE_OK` with length 0 only applies to `Curl_xfer_send()` and CURLE_AGAIN is returned by all other send() variants. - fix a bug in websocket `curl_ws_recv()` that mixed up data when it arrived in more than a single chunk The method for sending not just raw bytes, but bytes that are either "headers" or "body". The send abstraction stack, to to bottom, now is: * `Curl_req_send()`: has parameter to indicate amount of header bytes, buffers all data. * `Curl_xfer_send()`: knows on which socket index to send, returns amount of bytes sent. * `Curl_conn_send()`: called with socket index, returns amount of bytes sent. In addition there is `Curl_req_flush()` for writing out all buffered bytes. `Curl_req_send()` is active for requests without body, `Curl_buffer_send()` still being used for others. This is because the special quirks need to be addressed in future parts: * `expect-100` handling * `Curl_fillreadbuffer()` needs to add directly to the new `data->req.sendbuf` * special body handlings, like `chunked` encodings and line end conversions will be moved into something like a Client Reader. In functions of the pattern `CURLcode xxx_send(..., ssize_t *written)`, replace the `ssize_t` with a `size_t`. It makes no sense to allow for negative values as the returned `CURLcode` already specifies error conditions. This allows easier handling of lengths without casting. Closes #12964
2024-02-15 23:22:53 +08:00
*pnwritten = 0;
DEBUGASSERT(hds_len <= blen);
#ifdef DEBUGBUILD
lib: Curl_read/Curl_write clarifications - replace `Curl_read()`, `Curl_write()` and `Curl_nwrite()` to clarify when and at what level they operate - send/recv of transfer related data is now done via `Curl_xfer_send()/Curl_xfer_recv()` which no longer has socket/socketindex as parameter. It decides on the transfer setup of `conn->sockfd` and `conn->writesockfd` on which connection filter chain to operate. - send/recv on a specific connection filter chain is done via `Curl_conn_send()/Curl_conn_recv()` which get the socket index as parameter. - rename `Curl_setup_transfer()` to `Curl_xfer_setup()` for naming consistency - clarify that the special CURLE_AGAIN hangling to return `CURLE_OK` with length 0 only applies to `Curl_xfer_send()` and CURLE_AGAIN is returned by all other send() variants. - fix a bug in websocket `curl_ws_recv()` that mixed up data when it arrived in more than a single chunk The method for sending not just raw bytes, but bytes that are either "headers" or "body". The send abstraction stack, to to bottom, now is: * `Curl_req_send()`: has parameter to indicate amount of header bytes, buffers all data. * `Curl_xfer_send()`: knows on which socket index to send, returns amount of bytes sent. * `Curl_conn_send()`: called with socket index, returns amount of bytes sent. In addition there is `Curl_req_flush()` for writing out all buffered bytes. `Curl_req_send()` is active for requests without body, `Curl_buffer_send()` still being used for others. This is because the special quirks need to be addressed in future parts: * `expect-100` handling * `Curl_fillreadbuffer()` needs to add directly to the new `data->req.sendbuf` * special body handlings, like `chunked` encodings and line end conversions will be moved into something like a Client Reader. In functions of the pattern `CURLcode xxx_send(..., ssize_t *written)`, replace the `ssize_t` with a `size_t`. It makes no sense to allow for negative values as the returned `CURLcode` already specifies error conditions. This allows easier handling of lengths without casting. Closes #12964
2024-02-15 23:22:53 +08:00
{
/* Allow debug builds to override this logic to force short initial
sends */
size_t body_len = blen - hds_len;
lib: Curl_read/Curl_write clarifications - replace `Curl_read()`, `Curl_write()` and `Curl_nwrite()` to clarify when and at what level they operate - send/recv of transfer related data is now done via `Curl_xfer_send()/Curl_xfer_recv()` which no longer has socket/socketindex as parameter. It decides on the transfer setup of `conn->sockfd` and `conn->writesockfd` on which connection filter chain to operate. - send/recv on a specific connection filter chain is done via `Curl_conn_send()/Curl_conn_recv()` which get the socket index as parameter. - rename `Curl_setup_transfer()` to `Curl_xfer_setup()` for naming consistency - clarify that the special CURLE_AGAIN hangling to return `CURLE_OK` with length 0 only applies to `Curl_xfer_send()` and CURLE_AGAIN is returned by all other send() variants. - fix a bug in websocket `curl_ws_recv()` that mixed up data when it arrived in more than a single chunk The method for sending not just raw bytes, but bytes that are either "headers" or "body". The send abstraction stack, to to bottom, now is: * `Curl_req_send()`: has parameter to indicate amount of header bytes, buffers all data. * `Curl_xfer_send()`: knows on which socket index to send, returns amount of bytes sent. * `Curl_conn_send()`: called with socket index, returns amount of bytes sent. In addition there is `Curl_req_flush()` for writing out all buffered bytes. `Curl_req_send()` is active for requests without body, `Curl_buffer_send()` still being used for others. This is because the special quirks need to be addressed in future parts: * `expect-100` handling * `Curl_fillreadbuffer()` needs to add directly to the new `data->req.sendbuf` * special body handlings, like `chunked` encodings and line end conversions will be moved into something like a Client Reader. In functions of the pattern `CURLcode xxx_send(..., ssize_t *written)`, replace the `ssize_t` with a `size_t`. It makes no sense to allow for negative values as the returned `CURLcode` already specifies error conditions. This allows easier handling of lengths without casting. Closes #12964
2024-02-15 23:22:53 +08:00
char *p = getenv("CURL_SMALLREQSEND");
if(p) {
size_t body_small = (size_t)strtoul(p, NULL, 10);
if(body_small && body_small < body_len)
blen = hds_len + body_small;
lib: Curl_read/Curl_write clarifications - replace `Curl_read()`, `Curl_write()` and `Curl_nwrite()` to clarify when and at what level they operate - send/recv of transfer related data is now done via `Curl_xfer_send()/Curl_xfer_recv()` which no longer has socket/socketindex as parameter. It decides on the transfer setup of `conn->sockfd` and `conn->writesockfd` on which connection filter chain to operate. - send/recv on a specific connection filter chain is done via `Curl_conn_send()/Curl_conn_recv()` which get the socket index as parameter. - rename `Curl_setup_transfer()` to `Curl_xfer_setup()` for naming consistency - clarify that the special CURLE_AGAIN hangling to return `CURLE_OK` with length 0 only applies to `Curl_xfer_send()` and CURLE_AGAIN is returned by all other send() variants. - fix a bug in websocket `curl_ws_recv()` that mixed up data when it arrived in more than a single chunk The method for sending not just raw bytes, but bytes that are either "headers" or "body". The send abstraction stack, to to bottom, now is: * `Curl_req_send()`: has parameter to indicate amount of header bytes, buffers all data. * `Curl_xfer_send()`: knows on which socket index to send, returns amount of bytes sent. * `Curl_conn_send()`: called with socket index, returns amount of bytes sent. In addition there is `Curl_req_flush()` for writing out all buffered bytes. `Curl_req_send()` is active for requests without body, `Curl_buffer_send()` still being used for others. This is because the special quirks need to be addressed in future parts: * `expect-100` handling * `Curl_fillreadbuffer()` needs to add directly to the new `data->req.sendbuf` * special body handlings, like `chunked` encodings and line end conversions will be moved into something like a Client Reader. In functions of the pattern `CURLcode xxx_send(..., ssize_t *written)`, replace the `ssize_t` with a `size_t`. It makes no sense to allow for negative values as the returned `CURLcode` already specifies error conditions. This allows easier handling of lengths without casting. Closes #12964
2024-02-15 23:22:53 +08:00
}
}
#endif
/* Make sure this does not send more body bytes than what the max send
lib: Curl_read/Curl_write clarifications - replace `Curl_read()`, `Curl_write()` and `Curl_nwrite()` to clarify when and at what level they operate - send/recv of transfer related data is now done via `Curl_xfer_send()/Curl_xfer_recv()` which no longer has socket/socketindex as parameter. It decides on the transfer setup of `conn->sockfd` and `conn->writesockfd` on which connection filter chain to operate. - send/recv on a specific connection filter chain is done via `Curl_conn_send()/Curl_conn_recv()` which get the socket index as parameter. - rename `Curl_setup_transfer()` to `Curl_xfer_setup()` for naming consistency - clarify that the special CURLE_AGAIN hangling to return `CURLE_OK` with length 0 only applies to `Curl_xfer_send()` and CURLE_AGAIN is returned by all other send() variants. - fix a bug in websocket `curl_ws_recv()` that mixed up data when it arrived in more than a single chunk The method for sending not just raw bytes, but bytes that are either "headers" or "body". The send abstraction stack, to to bottom, now is: * `Curl_req_send()`: has parameter to indicate amount of header bytes, buffers all data. * `Curl_xfer_send()`: knows on which socket index to send, returns amount of bytes sent. * `Curl_conn_send()`: called with socket index, returns amount of bytes sent. In addition there is `Curl_req_flush()` for writing out all buffered bytes. `Curl_req_send()` is active for requests without body, `Curl_buffer_send()` still being used for others. This is because the special quirks need to be addressed in future parts: * `expect-100` handling * `Curl_fillreadbuffer()` needs to add directly to the new `data->req.sendbuf` * special body handlings, like `chunked` encodings and line end conversions will be moved into something like a Client Reader. In functions of the pattern `CURLcode xxx_send(..., ssize_t *written)`, replace the `ssize_t` with a `size_t`. It makes no sense to allow for negative values as the returned `CURLcode` already specifies error conditions. This allows easier handling of lengths without casting. Closes #12964
2024-02-15 23:22:53 +08:00
speed says. The headers do not count to the max speed. */
if(data->set.max_send_speed) {
size_t body_bytes = blen - hds_len;
if((curl_off_t)body_bytes > data->set.max_send_speed)
blen = hds_len + (size_t)data->set.max_send_speed;
}
if(data->req.eos_read &&
(Curl_bufq_is_empty(&data->req.sendbuf) ||
Curl_bufq_len(&data->req.sendbuf) == blen)) {
DEBUGF(infof(data, "sending last upload chunk of %zu bytes", blen));
eos = TRUE;
}
result = Curl_xfer_send(data, buf, blen, eos, pnwritten);
lib: Curl_read/Curl_write clarifications - replace `Curl_read()`, `Curl_write()` and `Curl_nwrite()` to clarify when and at what level they operate - send/recv of transfer related data is now done via `Curl_xfer_send()/Curl_xfer_recv()` which no longer has socket/socketindex as parameter. It decides on the transfer setup of `conn->sockfd` and `conn->writesockfd` on which connection filter chain to operate. - send/recv on a specific connection filter chain is done via `Curl_conn_send()/Curl_conn_recv()` which get the socket index as parameter. - rename `Curl_setup_transfer()` to `Curl_xfer_setup()` for naming consistency - clarify that the special CURLE_AGAIN hangling to return `CURLE_OK` with length 0 only applies to `Curl_xfer_send()` and CURLE_AGAIN is returned by all other send() variants. - fix a bug in websocket `curl_ws_recv()` that mixed up data when it arrived in more than a single chunk The method for sending not just raw bytes, but bytes that are either "headers" or "body". The send abstraction stack, to to bottom, now is: * `Curl_req_send()`: has parameter to indicate amount of header bytes, buffers all data. * `Curl_xfer_send()`: knows on which socket index to send, returns amount of bytes sent. * `Curl_conn_send()`: called with socket index, returns amount of bytes sent. In addition there is `Curl_req_flush()` for writing out all buffered bytes. `Curl_req_send()` is active for requests without body, `Curl_buffer_send()` still being used for others. This is because the special quirks need to be addressed in future parts: * `expect-100` handling * `Curl_fillreadbuffer()` needs to add directly to the new `data->req.sendbuf` * special body handlings, like `chunked` encodings and line end conversions will be moved into something like a Client Reader. In functions of the pattern `CURLcode xxx_send(..., ssize_t *written)`, replace the `ssize_t` with a `size_t`. It makes no sense to allow for negative values as the returned `CURLcode` already specifies error conditions. This allows easier handling of lengths without casting. Closes #12964
2024-02-15 23:22:53 +08:00
if(!result && *pnwritten) {
if(hds_len)
Curl_debug(data, CURLINFO_HEADER_OUT, (char *)buf,
CURLMIN(hds_len, *pnwritten));
if(*pnwritten > hds_len) {
size_t body_len = *pnwritten - hds_len;
Curl_debug(data, CURLINFO_DATA_OUT, (char *)buf + hds_len, body_len);
data->req.writebytecount += body_len;
Curl_pgrsSetUploadCounter(data, data->req.writebytecount);
}
}
return result;
}
static CURLcode req_send_buffer_flush(struct Curl_easy *data)
{
CURLcode result = CURLE_OK;
const unsigned char *buf;
size_t blen;
while(Curl_bufq_peek(&data->req.sendbuf, &buf, &blen)) {
size_t nwritten, hds_len = CURLMIN(data->req.sendbuf_hds_len, blen);
result = xfer_send(data, (const char *)buf, blen, hds_len, &nwritten);
lib: Curl_read/Curl_write clarifications - replace `Curl_read()`, `Curl_write()` and `Curl_nwrite()` to clarify when and at what level they operate - send/recv of transfer related data is now done via `Curl_xfer_send()/Curl_xfer_recv()` which no longer has socket/socketindex as parameter. It decides on the transfer setup of `conn->sockfd` and `conn->writesockfd` on which connection filter chain to operate. - send/recv on a specific connection filter chain is done via `Curl_conn_send()/Curl_conn_recv()` which get the socket index as parameter. - rename `Curl_setup_transfer()` to `Curl_xfer_setup()` for naming consistency - clarify that the special CURLE_AGAIN hangling to return `CURLE_OK` with length 0 only applies to `Curl_xfer_send()` and CURLE_AGAIN is returned by all other send() variants. - fix a bug in websocket `curl_ws_recv()` that mixed up data when it arrived in more than a single chunk The method for sending not just raw bytes, but bytes that are either "headers" or "body". The send abstraction stack, to to bottom, now is: * `Curl_req_send()`: has parameter to indicate amount of header bytes, buffers all data. * `Curl_xfer_send()`: knows on which socket index to send, returns amount of bytes sent. * `Curl_conn_send()`: called with socket index, returns amount of bytes sent. In addition there is `Curl_req_flush()` for writing out all buffered bytes. `Curl_req_send()` is active for requests without body, `Curl_buffer_send()` still being used for others. This is because the special quirks need to be addressed in future parts: * `expect-100` handling * `Curl_fillreadbuffer()` needs to add directly to the new `data->req.sendbuf` * special body handlings, like `chunked` encodings and line end conversions will be moved into something like a Client Reader. In functions of the pattern `CURLcode xxx_send(..., ssize_t *written)`, replace the `ssize_t` with a `size_t`. It makes no sense to allow for negative values as the returned `CURLcode` already specifies error conditions. This allows easier handling of lengths without casting. Closes #12964
2024-02-15 23:22:53 +08:00
if(result)
break;
Curl_bufq_skip(&data->req.sendbuf, nwritten);
lib: Curl_read/Curl_write clarifications - replace `Curl_read()`, `Curl_write()` and `Curl_nwrite()` to clarify when and at what level they operate - send/recv of transfer related data is now done via `Curl_xfer_send()/Curl_xfer_recv()` which no longer has socket/socketindex as parameter. It decides on the transfer setup of `conn->sockfd` and `conn->writesockfd` on which connection filter chain to operate. - send/recv on a specific connection filter chain is done via `Curl_conn_send()/Curl_conn_recv()` which get the socket index as parameter. - rename `Curl_setup_transfer()` to `Curl_xfer_setup()` for naming consistency - clarify that the special CURLE_AGAIN hangling to return `CURLE_OK` with length 0 only applies to `Curl_xfer_send()` and CURLE_AGAIN is returned by all other send() variants. - fix a bug in websocket `curl_ws_recv()` that mixed up data when it arrived in more than a single chunk (to be made into a sperate PR, also) Added as documented [in CLIENT-READER.md](https://github.com/curl/curl/blob/5b1f31dfbab8aef467c419c68aa06dc738cb75d4/docs/CLIENT-READERS.md). - old `Curl_buffer_send()` completely replaced by new `Curl_req_send()` - old `Curl_fillreadbuffer()` replaced with `Curl_client_read()` - HTTP chunked uploads are now formatted in a client reader added when needed. - FTP line-end conversions are done in a client reader added when needed. - when sending requests headers, remaining buffer space is filled with body data for sending in "one go". This is independent of the request body size. Resolves #12938 as now small and large requests have the same code path. Changes done to test cases: - test513: now fails before sending request headers as this initial "client read" triggers the setup fault. Behaves now the same as in hyper build - test547, test555, test1620: fix the length check in the lib code to only fail for reads *smaller* than expected. This was a bug in the test code that never triggered in the old implementation. Closes #12969
2024-02-15 23:22:53 +08:00
if(hds_len) {
lib: Curl_read/Curl_write clarifications - replace `Curl_read()`, `Curl_write()` and `Curl_nwrite()` to clarify when and at what level they operate - send/recv of transfer related data is now done via `Curl_xfer_send()/Curl_xfer_recv()` which no longer has socket/socketindex as parameter. It decides on the transfer setup of `conn->sockfd` and `conn->writesockfd` on which connection filter chain to operate. - send/recv on a specific connection filter chain is done via `Curl_conn_send()/Curl_conn_recv()` which get the socket index as parameter. - rename `Curl_setup_transfer()` to `Curl_xfer_setup()` for naming consistency - clarify that the special CURLE_AGAIN hangling to return `CURLE_OK` with length 0 only applies to `Curl_xfer_send()` and CURLE_AGAIN is returned by all other send() variants. - fix a bug in websocket `curl_ws_recv()` that mixed up data when it arrived in more than a single chunk The method for sending not just raw bytes, but bytes that are either "headers" or "body". The send abstraction stack, to to bottom, now is: * `Curl_req_send()`: has parameter to indicate amount of header bytes, buffers all data. * `Curl_xfer_send()`: knows on which socket index to send, returns amount of bytes sent. * `Curl_conn_send()`: called with socket index, returns amount of bytes sent. In addition there is `Curl_req_flush()` for writing out all buffered bytes. `Curl_req_send()` is active for requests without body, `Curl_buffer_send()` still being used for others. This is because the special quirks need to be addressed in future parts: * `expect-100` handling * `Curl_fillreadbuffer()` needs to add directly to the new `data->req.sendbuf` * special body handlings, like `chunked` encodings and line end conversions will be moved into something like a Client Reader. In functions of the pattern `CURLcode xxx_send(..., ssize_t *written)`, replace the `ssize_t` with a `size_t`. It makes no sense to allow for negative values as the returned `CURLcode` already specifies error conditions. This allows easier handling of lengths without casting. Closes #12964
2024-02-15 23:22:53 +08:00
data->req.sendbuf_hds_len -= CURLMIN(hds_len, nwritten);
lib: Curl_read/Curl_write clarifications - replace `Curl_read()`, `Curl_write()` and `Curl_nwrite()` to clarify when and at what level they operate - send/recv of transfer related data is now done via `Curl_xfer_send()/Curl_xfer_recv()` which no longer has socket/socketindex as parameter. It decides on the transfer setup of `conn->sockfd` and `conn->writesockfd` on which connection filter chain to operate. - send/recv on a specific connection filter chain is done via `Curl_conn_send()/Curl_conn_recv()` which get the socket index as parameter. - rename `Curl_setup_transfer()` to `Curl_xfer_setup()` for naming consistency - clarify that the special CURLE_AGAIN hangling to return `CURLE_OK` with length 0 only applies to `Curl_xfer_send()` and CURLE_AGAIN is returned by all other send() variants. - fix a bug in websocket `curl_ws_recv()` that mixed up data when it arrived in more than a single chunk (to be made into a sperate PR, also) Added as documented [in CLIENT-READER.md](https://github.com/curl/curl/blob/5b1f31dfbab8aef467c419c68aa06dc738cb75d4/docs/CLIENT-READERS.md). - old `Curl_buffer_send()` completely replaced by new `Curl_req_send()` - old `Curl_fillreadbuffer()` replaced with `Curl_client_read()` - HTTP chunked uploads are now formatted in a client reader added when needed. - FTP line-end conversions are done in a client reader added when needed. - when sending requests headers, remaining buffer space is filled with body data for sending in "one go". This is independent of the request body size. Resolves #12938 as now small and large requests have the same code path. Changes done to test cases: - test513: now fails before sending request headers as this initial "client read" triggers the setup fault. Behaves now the same as in hyper build - test547, test555, test1620: fix the length check in the lib code to only fail for reads *smaller* than expected. This was a bug in the test code that never triggered in the old implementation. Closes #12969
2024-02-15 23:22:53 +08:00
}
lib: Curl_read/Curl_write clarifications - replace `Curl_read()`, `Curl_write()` and `Curl_nwrite()` to clarify when and at what level they operate - send/recv of transfer related data is now done via `Curl_xfer_send()/Curl_xfer_recv()` which no longer has socket/socketindex as parameter. It decides on the transfer setup of `conn->sockfd` and `conn->writesockfd` on which connection filter chain to operate. - send/recv on a specific connection filter chain is done via `Curl_conn_send()/Curl_conn_recv()` which get the socket index as parameter. - rename `Curl_setup_transfer()` to `Curl_xfer_setup()` for naming consistency - clarify that the special CURLE_AGAIN hangling to return `CURLE_OK` with length 0 only applies to `Curl_xfer_send()` and CURLE_AGAIN is returned by all other send() variants. - fix a bug in websocket `curl_ws_recv()` that mixed up data when it arrived in more than a single chunk The method for sending not just raw bytes, but bytes that are either "headers" or "body". The send abstraction stack, to to bottom, now is: * `Curl_req_send()`: has parameter to indicate amount of header bytes, buffers all data. * `Curl_xfer_send()`: knows on which socket index to send, returns amount of bytes sent. * `Curl_conn_send()`: called with socket index, returns amount of bytes sent. In addition there is `Curl_req_flush()` for writing out all buffered bytes. `Curl_req_send()` is active for requests without body, `Curl_buffer_send()` still being used for others. This is because the special quirks need to be addressed in future parts: * `expect-100` handling * `Curl_fillreadbuffer()` needs to add directly to the new `data->req.sendbuf` * special body handlings, like `chunked` encodings and line end conversions will be moved into something like a Client Reader. In functions of the pattern `CURLcode xxx_send(..., ssize_t *written)`, replace the `ssize_t` with a `size_t`. It makes no sense to allow for negative values as the returned `CURLcode` already specifies error conditions. This allows easier handling of lengths without casting. Closes #12964
2024-02-15 23:22:53 +08:00
/* leave if we could not send all. Maybe network blocking or
* speed limits on transfer */
if(nwritten < blen)
break;
}
return result;
}
CURLcode Curl_req_set_upload_done(struct Curl_easy *data)
{
DEBUGASSERT(!data->req.upload_done);
data->req.upload_done = TRUE;
data->req.keepon &= ~(KEEP_SEND|KEEP_SEND_TIMED); /* we are done sending */
Curl_pgrsTime(data, TIMER_POSTRANSFER);
Curl_creader_done(data, data->req.upload_aborted);
if(data->req.upload_aborted) {
Curl_bufq_reset(&data->req.sendbuf);
if(data->req.writebytecount)
infof(data, "abort upload after having sent %" CURL_FORMAT_CURL_OFF_T
" bytes", data->req.writebytecount);
else
infof(data, "abort upload");
}
else if(data->req.writebytecount)
infof(data, "upload completely sent off: %" CURL_FORMAT_CURL_OFF_T
" bytes", data->req.writebytecount);
else if(!data->req.download_done) {
DEBUGASSERT(Curl_bufq_is_empty(&data->req.sendbuf));
infof(data, Curl_creader_total_length(data)?
"We are completely uploaded and fine" :
"Request completely sent off");
}
return Curl_xfer_send_close(data);
}
static CURLcode req_flush(struct Curl_easy *data)
lib: Curl_read/Curl_write clarifications - replace `Curl_read()`, `Curl_write()` and `Curl_nwrite()` to clarify when and at what level they operate - send/recv of transfer related data is now done via `Curl_xfer_send()/Curl_xfer_recv()` which no longer has socket/socketindex as parameter. It decides on the transfer setup of `conn->sockfd` and `conn->writesockfd` on which connection filter chain to operate. - send/recv on a specific connection filter chain is done via `Curl_conn_send()/Curl_conn_recv()` which get the socket index as parameter. - rename `Curl_setup_transfer()` to `Curl_xfer_setup()` for naming consistency - clarify that the special CURLE_AGAIN hangling to return `CURLE_OK` with length 0 only applies to `Curl_xfer_send()` and CURLE_AGAIN is returned by all other send() variants. - fix a bug in websocket `curl_ws_recv()` that mixed up data when it arrived in more than a single chunk The method for sending not just raw bytes, but bytes that are either "headers" or "body". The send abstraction stack, to to bottom, now is: * `Curl_req_send()`: has parameter to indicate amount of header bytes, buffers all data. * `Curl_xfer_send()`: knows on which socket index to send, returns amount of bytes sent. * `Curl_conn_send()`: called with socket index, returns amount of bytes sent. In addition there is `Curl_req_flush()` for writing out all buffered bytes. `Curl_req_send()` is active for requests without body, `Curl_buffer_send()` still being used for others. This is because the special quirks need to be addressed in future parts: * `expect-100` handling * `Curl_fillreadbuffer()` needs to add directly to the new `data->req.sendbuf` * special body handlings, like `chunked` encodings and line end conversions will be moved into something like a Client Reader. In functions of the pattern `CURLcode xxx_send(..., ssize_t *written)`, replace the `ssize_t` with a `size_t`. It makes no sense to allow for negative values as the returned `CURLcode` already specifies error conditions. This allows easier handling of lengths without casting. Closes #12964
2024-02-15 23:22:53 +08:00
{
CURLcode result;
if(!data || !data->conn)
return CURLE_FAILED_INIT;
if(!Curl_bufq_is_empty(&data->req.sendbuf)) {
result = req_send_buffer_flush(data);
if(result)
return result;
if(!Curl_bufq_is_empty(&data->req.sendbuf)) {
DEBUGF(infof(data, "Curl_req_flush(len=%zu) -> EAGAIN",
Curl_bufq_len(&data->req.sendbuf)));
lib: Curl_read/Curl_write clarifications - replace `Curl_read()`, `Curl_write()` and `Curl_nwrite()` to clarify when and at what level they operate - send/recv of transfer related data is now done via `Curl_xfer_send()/Curl_xfer_recv()` which no longer has socket/socketindex as parameter. It decides on the transfer setup of `conn->sockfd` and `conn->writesockfd` on which connection filter chain to operate. - send/recv on a specific connection filter chain is done via `Curl_conn_send()/Curl_conn_recv()` which get the socket index as parameter. - rename `Curl_setup_transfer()` to `Curl_xfer_setup()` for naming consistency - clarify that the special CURLE_AGAIN hangling to return `CURLE_OK` with length 0 only applies to `Curl_xfer_send()` and CURLE_AGAIN is returned by all other send() variants. - fix a bug in websocket `curl_ws_recv()` that mixed up data when it arrived in more than a single chunk The method for sending not just raw bytes, but bytes that are either "headers" or "body". The send abstraction stack, to to bottom, now is: * `Curl_req_send()`: has parameter to indicate amount of header bytes, buffers all data. * `Curl_xfer_send()`: knows on which socket index to send, returns amount of bytes sent. * `Curl_conn_send()`: called with socket index, returns amount of bytes sent. In addition there is `Curl_req_flush()` for writing out all buffered bytes. `Curl_req_send()` is active for requests without body, `Curl_buffer_send()` still being used for others. This is because the special quirks need to be addressed in future parts: * `expect-100` handling * `Curl_fillreadbuffer()` needs to add directly to the new `data->req.sendbuf` * special body handlings, like `chunked` encodings and line end conversions will be moved into something like a Client Reader. In functions of the pattern `CURLcode xxx_send(..., ssize_t *written)`, replace the `ssize_t` with a `size_t`. It makes no sense to allow for negative values as the returned `CURLcode` already specifies error conditions. This allows easier handling of lengths without casting. Closes #12964
2024-02-15 23:22:53 +08:00
return CURLE_AGAIN;
}
}
else if(Curl_xfer_needs_flush(data)) {
DEBUGF(infof(data, "Curl_req_flush(), xfer send_pending"));
return Curl_xfer_flush(data);
}
if(!data->req.upload_done && data->req.eos_read &&
Curl_bufq_is_empty(&data->req.sendbuf)) {
if(data->req.shutdown) {
bool done;
result = Curl_xfer_send_shutdown(data, &done);
if(result)
return result;
if(!done)
return CURLE_AGAIN;
}
return Curl_req_set_upload_done(data);
}
lib: Curl_read/Curl_write clarifications - replace `Curl_read()`, `Curl_write()` and `Curl_nwrite()` to clarify when and at what level they operate - send/recv of transfer related data is now done via `Curl_xfer_send()/Curl_xfer_recv()` which no longer has socket/socketindex as parameter. It decides on the transfer setup of `conn->sockfd` and `conn->writesockfd` on which connection filter chain to operate. - send/recv on a specific connection filter chain is done via `Curl_conn_send()/Curl_conn_recv()` which get the socket index as parameter. - rename `Curl_setup_transfer()` to `Curl_xfer_setup()` for naming consistency - clarify that the special CURLE_AGAIN hangling to return `CURLE_OK` with length 0 only applies to `Curl_xfer_send()` and CURLE_AGAIN is returned by all other send() variants. - fix a bug in websocket `curl_ws_recv()` that mixed up data when it arrived in more than a single chunk The method for sending not just raw bytes, but bytes that are either "headers" or "body". The send abstraction stack, to to bottom, now is: * `Curl_req_send()`: has parameter to indicate amount of header bytes, buffers all data. * `Curl_xfer_send()`: knows on which socket index to send, returns amount of bytes sent. * `Curl_conn_send()`: called with socket index, returns amount of bytes sent. In addition there is `Curl_req_flush()` for writing out all buffered bytes. `Curl_req_send()` is active for requests without body, `Curl_buffer_send()` still being used for others. This is because the special quirks need to be addressed in future parts: * `expect-100` handling * `Curl_fillreadbuffer()` needs to add directly to the new `data->req.sendbuf` * special body handlings, like `chunked` encodings and line end conversions will be moved into something like a Client Reader. In functions of the pattern `CURLcode xxx_send(..., ssize_t *written)`, replace the `ssize_t` with a `size_t`. It makes no sense to allow for negative values as the returned `CURLcode` already specifies error conditions. This allows easier handling of lengths without casting. Closes #12964
2024-02-15 23:22:53 +08:00
return CURLE_OK;
}
static ssize_t add_from_client(void *reader_ctx,
unsigned char *buf, size_t buflen,
CURLcode *err)
{
struct Curl_easy *data = reader_ctx;
size_t nread;
bool eos;
*err = Curl_client_read(data, (char *)buf, buflen, &nread, &eos);
if(*err)
return -1;
if(eos)
data->req.eos_read = TRUE;
return (ssize_t)nread;
}
lib: Curl_read/Curl_write clarifications - replace `Curl_read()`, `Curl_write()` and `Curl_nwrite()` to clarify when and at what level they operate - send/recv of transfer related data is now done via `Curl_xfer_send()/Curl_xfer_recv()` which no longer has socket/socketindex as parameter. It decides on the transfer setup of `conn->sockfd` and `conn->writesockfd` on which connection filter chain to operate. - send/recv on a specific connection filter chain is done via `Curl_conn_send()/Curl_conn_recv()` which get the socket index as parameter. - rename `Curl_setup_transfer()` to `Curl_xfer_setup()` for naming consistency - clarify that the special CURLE_AGAIN hangling to return `CURLE_OK` with length 0 only applies to `Curl_xfer_send()` and CURLE_AGAIN is returned by all other send() variants. - fix a bug in websocket `curl_ws_recv()` that mixed up data when it arrived in more than a single chunk (to be made into a sperate PR, also) Added as documented [in CLIENT-READER.md](https://github.com/curl/curl/blob/5b1f31dfbab8aef467c419c68aa06dc738cb75d4/docs/CLIENT-READERS.md). - old `Curl_buffer_send()` completely replaced by new `Curl_req_send()` - old `Curl_fillreadbuffer()` replaced with `Curl_client_read()` - HTTP chunked uploads are now formatted in a client reader added when needed. - FTP line-end conversions are done in a client reader added when needed. - when sending requests headers, remaining buffer space is filled with body data for sending in "one go". This is independent of the request body size. Resolves #12938 as now small and large requests have the same code path. Changes done to test cases: - test513: now fails before sending request headers as this initial "client read" triggers the setup fault. Behaves now the same as in hyper build - test547, test555, test1620: fix the length check in the lib code to only fail for reads *smaller* than expected. This was a bug in the test code that never triggered in the old implementation. Closes #12969
2024-02-15 23:22:53 +08:00
#ifndef USE_HYPER
static CURLcode req_send_buffer_add(struct Curl_easy *data,
const char *buf, size_t blen,
size_t hds_len)
{
CURLcode result = CURLE_OK;
ssize_t n;
n = Curl_bufq_write(&data->req.sendbuf,
(const unsigned char *)buf, blen, &result);
if(n < 0)
return result;
/* We rely on a SOFTLIMIT on sendbuf, so it can take all data in */
DEBUGASSERT((size_t)n == blen);
data->req.sendbuf_hds_len += hds_len;
return CURLE_OK;
}
CURLcode Curl_req_send(struct Curl_easy *data, struct dynbuf *req)
lib: Curl_read/Curl_write clarifications - replace `Curl_read()`, `Curl_write()` and `Curl_nwrite()` to clarify when and at what level they operate - send/recv of transfer related data is now done via `Curl_xfer_send()/Curl_xfer_recv()` which no longer has socket/socketindex as parameter. It decides on the transfer setup of `conn->sockfd` and `conn->writesockfd` on which connection filter chain to operate. - send/recv on a specific connection filter chain is done via `Curl_conn_send()/Curl_conn_recv()` which get the socket index as parameter. - rename `Curl_setup_transfer()` to `Curl_xfer_setup()` for naming consistency - clarify that the special CURLE_AGAIN hangling to return `CURLE_OK` with length 0 only applies to `Curl_xfer_send()` and CURLE_AGAIN is returned by all other send() variants. - fix a bug in websocket `curl_ws_recv()` that mixed up data when it arrived in more than a single chunk The method for sending not just raw bytes, but bytes that are either "headers" or "body". The send abstraction stack, to to bottom, now is: * `Curl_req_send()`: has parameter to indicate amount of header bytes, buffers all data. * `Curl_xfer_send()`: knows on which socket index to send, returns amount of bytes sent. * `Curl_conn_send()`: called with socket index, returns amount of bytes sent. In addition there is `Curl_req_flush()` for writing out all buffered bytes. `Curl_req_send()` is active for requests without body, `Curl_buffer_send()` still being used for others. This is because the special quirks need to be addressed in future parts: * `expect-100` handling * `Curl_fillreadbuffer()` needs to add directly to the new `data->req.sendbuf` * special body handlings, like `chunked` encodings and line end conversions will be moved into something like a Client Reader. In functions of the pattern `CURLcode xxx_send(..., ssize_t *written)`, replace the `ssize_t` with a `size_t`. It makes no sense to allow for negative values as the returned `CURLcode` already specifies error conditions. This allows easier handling of lengths without casting. Closes #12964
2024-02-15 23:22:53 +08:00
{
CURLcode result;
const char *buf;
size_t blen, nwritten;
lib: Curl_read/Curl_write clarifications - replace `Curl_read()`, `Curl_write()` and `Curl_nwrite()` to clarify when and at what level they operate - send/recv of transfer related data is now done via `Curl_xfer_send()/Curl_xfer_recv()` which no longer has socket/socketindex as parameter. It decides on the transfer setup of `conn->sockfd` and `conn->writesockfd` on which connection filter chain to operate. - send/recv on a specific connection filter chain is done via `Curl_conn_send()/Curl_conn_recv()` which get the socket index as parameter. - rename `Curl_setup_transfer()` to `Curl_xfer_setup()` for naming consistency - clarify that the special CURLE_AGAIN hangling to return `CURLE_OK` with length 0 only applies to `Curl_xfer_send()` and CURLE_AGAIN is returned by all other send() variants. - fix a bug in websocket `curl_ws_recv()` that mixed up data when it arrived in more than a single chunk The method for sending not just raw bytes, but bytes that are either "headers" or "body". The send abstraction stack, to to bottom, now is: * `Curl_req_send()`: has parameter to indicate amount of header bytes, buffers all data. * `Curl_xfer_send()`: knows on which socket index to send, returns amount of bytes sent. * `Curl_conn_send()`: called with socket index, returns amount of bytes sent. In addition there is `Curl_req_flush()` for writing out all buffered bytes. `Curl_req_send()` is active for requests without body, `Curl_buffer_send()` still being used for others. This is because the special quirks need to be addressed in future parts: * `expect-100` handling * `Curl_fillreadbuffer()` needs to add directly to the new `data->req.sendbuf` * special body handlings, like `chunked` encodings and line end conversions will be moved into something like a Client Reader. In functions of the pattern `CURLcode xxx_send(..., ssize_t *written)`, replace the `ssize_t` with a `size_t`. It makes no sense to allow for negative values as the returned `CURLcode` already specifies error conditions. This allows easier handling of lengths without casting. Closes #12964
2024-02-15 23:22:53 +08:00
if(!data || !data->conn)
return CURLE_FAILED_INIT;
buf = Curl_dyn_ptr(req);
blen = Curl_dyn_len(req);
if(!Curl_creader_total_length(data)) {
/* Request without body. Try to send directly from the buf given. */
data->req.eos_read = TRUE;
result = xfer_send(data, buf, blen, blen, &nwritten);
if(result)
return result;
buf += nwritten;
blen -= nwritten;
}
lib: Curl_read/Curl_write clarifications - replace `Curl_read()`, `Curl_write()` and `Curl_nwrite()` to clarify when and at what level they operate - send/recv of transfer related data is now done via `Curl_xfer_send()/Curl_xfer_recv()` which no longer has socket/socketindex as parameter. It decides on the transfer setup of `conn->sockfd` and `conn->writesockfd` on which connection filter chain to operate. - send/recv on a specific connection filter chain is done via `Curl_conn_send()/Curl_conn_recv()` which get the socket index as parameter. - rename `Curl_setup_transfer()` to `Curl_xfer_setup()` for naming consistency - clarify that the special CURLE_AGAIN hangling to return `CURLE_OK` with length 0 only applies to `Curl_xfer_send()` and CURLE_AGAIN is returned by all other send() variants. - fix a bug in websocket `curl_ws_recv()` that mixed up data when it arrived in more than a single chunk (to be made into a sperate PR, also) Added as documented [in CLIENT-READER.md](https://github.com/curl/curl/blob/5b1f31dfbab8aef467c419c68aa06dc738cb75d4/docs/CLIENT-READERS.md). - old `Curl_buffer_send()` completely replaced by new `Curl_req_send()` - old `Curl_fillreadbuffer()` replaced with `Curl_client_read()` - HTTP chunked uploads are now formatted in a client reader added when needed. - FTP line-end conversions are done in a client reader added when needed. - when sending requests headers, remaining buffer space is filled with body data for sending in "one go". This is independent of the request body size. Resolves #12938 as now small and large requests have the same code path. Changes done to test cases: - test513: now fails before sending request headers as this initial "client read" triggers the setup fault. Behaves now the same as in hyper build - test547, test555, test1620: fix the length check in the lib code to only fail for reads *smaller* than expected. This was a bug in the test code that never triggered in the old implementation. Closes #12969
2024-02-15 23:22:53 +08:00
if(blen) {
/* Either we have a request body, or we could not send the complete
* request in one go. Buffer the remainder and try to add as much
* body bytes as room is left in the buffer. Then flush. */
result = req_send_buffer_add(data, buf, blen, blen);
if(result)
return result;
return Curl_req_send_more(data);
}
return CURLE_OK;
}
#endif /* !USE_HYPER */
bool Curl_req_sendbuf_empty(struct Curl_easy *data)
{
return !data->req.sendbuf_init || Curl_bufq_is_empty(&data->req.sendbuf);
}
bool Curl_req_want_send(struct Curl_easy *data)
{
/* Not done and
* - KEEP_SEND and not PAUSEd.
* - or request has buffered data to send
* - or transfer connection has pending data to send */
return !data->req.done &&
(((data->req.keepon & KEEP_SENDBITS) == KEEP_SEND) ||
!Curl_req_sendbuf_empty(data) ||
Curl_xfer_needs_flush(data));
}
bool Curl_req_done_sending(struct Curl_easy *data)
{
return data->req.upload_done && !Curl_req_want_send(data);
}
CURLcode Curl_req_send_more(struct Curl_easy *data)
{
CURLcode result;
/* Fill our send buffer if more from client can be read. */
if(!data->req.upload_aborted &&
!data->req.eos_read &&
!(data->req.keepon & KEEP_SEND_PAUSE) &&
!Curl_bufq_is_full(&data->req.sendbuf)) {
lib: Curl_read/Curl_write clarifications - replace `Curl_read()`, `Curl_write()` and `Curl_nwrite()` to clarify when and at what level they operate - send/recv of transfer related data is now done via `Curl_xfer_send()/Curl_xfer_recv()` which no longer has socket/socketindex as parameter. It decides on the transfer setup of `conn->sockfd` and `conn->writesockfd` on which connection filter chain to operate. - send/recv on a specific connection filter chain is done via `Curl_conn_send()/Curl_conn_recv()` which get the socket index as parameter. - rename `Curl_setup_transfer()` to `Curl_xfer_setup()` for naming consistency - clarify that the special CURLE_AGAIN hangling to return `CURLE_OK` with length 0 only applies to `Curl_xfer_send()` and CURLE_AGAIN is returned by all other send() variants. - fix a bug in websocket `curl_ws_recv()` that mixed up data when it arrived in more than a single chunk (to be made into a sperate PR, also) Added as documented [in CLIENT-READER.md](https://github.com/curl/curl/blob/5b1f31dfbab8aef467c419c68aa06dc738cb75d4/docs/CLIENT-READERS.md). - old `Curl_buffer_send()` completely replaced by new `Curl_req_send()` - old `Curl_fillreadbuffer()` replaced with `Curl_client_read()` - HTTP chunked uploads are now formatted in a client reader added when needed. - FTP line-end conversions are done in a client reader added when needed. - when sending requests headers, remaining buffer space is filled with body data for sending in "one go". This is independent of the request body size. Resolves #12938 as now small and large requests have the same code path. Changes done to test cases: - test513: now fails before sending request headers as this initial "client read" triggers the setup fault. Behaves now the same as in hyper build - test547, test555, test1620: fix the length check in the lib code to only fail for reads *smaller* than expected. This was a bug in the test code that never triggered in the old implementation. Closes #12969
2024-02-15 23:22:53 +08:00
ssize_t nread = Curl_bufq_sipn(&data->req.sendbuf, 0,
add_from_client, data, &result);
if(nread < 0 && result != CURLE_AGAIN)
return result;
}
result = req_flush(data);
lib: Curl_read/Curl_write clarifications - replace `Curl_read()`, `Curl_write()` and `Curl_nwrite()` to clarify when and at what level they operate - send/recv of transfer related data is now done via `Curl_xfer_send()/Curl_xfer_recv()` which no longer has socket/socketindex as parameter. It decides on the transfer setup of `conn->sockfd` and `conn->writesockfd` on which connection filter chain to operate. - send/recv on a specific connection filter chain is done via `Curl_conn_send()/Curl_conn_recv()` which get the socket index as parameter. - rename `Curl_setup_transfer()` to `Curl_xfer_setup()` for naming consistency - clarify that the special CURLE_AGAIN hangling to return `CURLE_OK` with length 0 only applies to `Curl_xfer_send()` and CURLE_AGAIN is returned by all other send() variants. - fix a bug in websocket `curl_ws_recv()` that mixed up data when it arrived in more than a single chunk The method for sending not just raw bytes, but bytes that are either "headers" or "body". The send abstraction stack, to to bottom, now is: * `Curl_req_send()`: has parameter to indicate amount of header bytes, buffers all data. * `Curl_xfer_send()`: knows on which socket index to send, returns amount of bytes sent. * `Curl_conn_send()`: called with socket index, returns amount of bytes sent. In addition there is `Curl_req_flush()` for writing out all buffered bytes. `Curl_req_send()` is active for requests without body, `Curl_buffer_send()` still being used for others. This is because the special quirks need to be addressed in future parts: * `expect-100` handling * `Curl_fillreadbuffer()` needs to add directly to the new `data->req.sendbuf` * special body handlings, like `chunked` encodings and line end conversions will be moved into something like a Client Reader. In functions of the pattern `CURLcode xxx_send(..., ssize_t *written)`, replace the `ssize_t` with a `size_t`. It makes no sense to allow for negative values as the returned `CURLcode` already specifies error conditions. This allows easier handling of lengths without casting. Closes #12964
2024-02-15 23:22:53 +08:00
if(result == CURLE_AGAIN)
result = CURLE_OK;
lib: Curl_read/Curl_write clarifications - replace `Curl_read()`, `Curl_write()` and `Curl_nwrite()` to clarify when and at what level they operate - send/recv of transfer related data is now done via `Curl_xfer_send()/Curl_xfer_recv()` which no longer has socket/socketindex as parameter. It decides on the transfer setup of `conn->sockfd` and `conn->writesockfd` on which connection filter chain to operate. - send/recv on a specific connection filter chain is done via `Curl_conn_send()/Curl_conn_recv()` which get the socket index as parameter. - rename `Curl_setup_transfer()` to `Curl_xfer_setup()` for naming consistency - clarify that the special CURLE_AGAIN hangling to return `CURLE_OK` with length 0 only applies to `Curl_xfer_send()` and CURLE_AGAIN is returned by all other send() variants. - fix a bug in websocket `curl_ws_recv()` that mixed up data when it arrived in more than a single chunk The method for sending not just raw bytes, but bytes that are either "headers" or "body". The send abstraction stack, to to bottom, now is: * `Curl_req_send()`: has parameter to indicate amount of header bytes, buffers all data. * `Curl_xfer_send()`: knows on which socket index to send, returns amount of bytes sent. * `Curl_conn_send()`: called with socket index, returns amount of bytes sent. In addition there is `Curl_req_flush()` for writing out all buffered bytes. `Curl_req_send()` is active for requests without body, `Curl_buffer_send()` still being used for others. This is because the special quirks need to be addressed in future parts: * `expect-100` handling * `Curl_fillreadbuffer()` needs to add directly to the new `data->req.sendbuf` * special body handlings, like `chunked` encodings and line end conversions will be moved into something like a Client Reader. In functions of the pattern `CURLcode xxx_send(..., ssize_t *written)`, replace the `ssize_t` with a `size_t`. It makes no sense to allow for negative values as the returned `CURLcode` already specifies error conditions. This allows easier handling of lengths without casting. Closes #12964
2024-02-15 23:22:53 +08:00
return result;
}
CURLcode Curl_req_abort_sending(struct Curl_easy *data)
lib: Curl_read/Curl_write clarifications - replace `Curl_read()`, `Curl_write()` and `Curl_nwrite()` to clarify when and at what level they operate - send/recv of transfer related data is now done via `Curl_xfer_send()/Curl_xfer_recv()` which no longer has socket/socketindex as parameter. It decides on the transfer setup of `conn->sockfd` and `conn->writesockfd` on which connection filter chain to operate. - send/recv on a specific connection filter chain is done via `Curl_conn_send()/Curl_conn_recv()` which get the socket index as parameter. - rename `Curl_setup_transfer()` to `Curl_xfer_setup()` for naming consistency - clarify that the special CURLE_AGAIN hangling to return `CURLE_OK` with length 0 only applies to `Curl_xfer_send()` and CURLE_AGAIN is returned by all other send() variants. - fix a bug in websocket `curl_ws_recv()` that mixed up data when it arrived in more than a single chunk The method for sending not just raw bytes, but bytes that are either "headers" or "body". The send abstraction stack, to to bottom, now is: * `Curl_req_send()`: has parameter to indicate amount of header bytes, buffers all data. * `Curl_xfer_send()`: knows on which socket index to send, returns amount of bytes sent. * `Curl_conn_send()`: called with socket index, returns amount of bytes sent. In addition there is `Curl_req_flush()` for writing out all buffered bytes. `Curl_req_send()` is active for requests without body, `Curl_buffer_send()` still being used for others. This is because the special quirks need to be addressed in future parts: * `expect-100` handling * `Curl_fillreadbuffer()` needs to add directly to the new `data->req.sendbuf` * special body handlings, like `chunked` encodings and line end conversions will be moved into something like a Client Reader. In functions of the pattern `CURLcode xxx_send(..., ssize_t *written)`, replace the `ssize_t` with a `size_t`. It makes no sense to allow for negative values as the returned `CURLcode` already specifies error conditions. This allows easier handling of lengths without casting. Closes #12964
2024-02-15 23:22:53 +08:00
{
if(!data->req.upload_done) {
Curl_bufq_reset(&data->req.sendbuf);
data->req.upload_aborted = TRUE;
/* no longer KEEP_SEND and KEEP_SEND_PAUSE */
data->req.keepon &= ~KEEP_SENDBITS;
return Curl_req_set_upload_done(data);
}
return CURLE_OK;
lib: Curl_read/Curl_write clarifications - replace `Curl_read()`, `Curl_write()` and `Curl_nwrite()` to clarify when and at what level they operate - send/recv of transfer related data is now done via `Curl_xfer_send()/Curl_xfer_recv()` which no longer has socket/socketindex as parameter. It decides on the transfer setup of `conn->sockfd` and `conn->writesockfd` on which connection filter chain to operate. - send/recv on a specific connection filter chain is done via `Curl_conn_send()/Curl_conn_recv()` which get the socket index as parameter. - rename `Curl_setup_transfer()` to `Curl_xfer_setup()` for naming consistency - clarify that the special CURLE_AGAIN hangling to return `CURLE_OK` with length 0 only applies to `Curl_xfer_send()` and CURLE_AGAIN is returned by all other send() variants. - fix a bug in websocket `curl_ws_recv()` that mixed up data when it arrived in more than a single chunk The method for sending not just raw bytes, but bytes that are either "headers" or "body". The send abstraction stack, to to bottom, now is: * `Curl_req_send()`: has parameter to indicate amount of header bytes, buffers all data. * `Curl_xfer_send()`: knows on which socket index to send, returns amount of bytes sent. * `Curl_conn_send()`: called with socket index, returns amount of bytes sent. In addition there is `Curl_req_flush()` for writing out all buffered bytes. `Curl_req_send()` is active for requests without body, `Curl_buffer_send()` still being used for others. This is because the special quirks need to be addressed in future parts: * `expect-100` handling * `Curl_fillreadbuffer()` needs to add directly to the new `data->req.sendbuf` * special body handlings, like `chunked` encodings and line end conversions will be moved into something like a Client Reader. In functions of the pattern `CURLcode xxx_send(..., ssize_t *written)`, replace the `ssize_t` with a `size_t`. It makes no sense to allow for negative values as the returned `CURLcode` already specifies error conditions. This allows easier handling of lengths without casting. Closes #12964
2024-02-15 23:22:53 +08:00
}
CURLcode Curl_req_stop_send_recv(struct Curl_easy *data)
{
/* stop receiving and ALL sending as well, including PAUSE and HOLD.
* We might still be paused on receive client writes though, so
* keep those bits around. */
data->req.keepon &= ~(KEEP_RECV|KEEP_SENDBITS);
return Curl_req_abort_sending(data);
}