-const char parsers_rcs[] = "$Id: parsers.c,v 1.269 2012/11/24 14:09:11 fabiankeil Exp $";
/*********************************************************************
*
* File : $Source: /cvsroot/ijbswa/current/parsers.c,v $
*
* Purpose : Declares functions to parse/crunch headers and pages.
*
- * Copyright : Written by and Copyright (C) 2001-2012 the
+ * Copyright : Written by and Copyright (C) 2001-2017 the
* Privoxy team. http://www.privoxy.org/
*
* Based on the Internet Junkbuster originally written
#include "strptime.h"
#endif
-const char parsers_h_rcs[] = PARSERS_H_VERSION;
-
static char *get_header_line(struct iob *iob);
static jb_err scan_headers(struct client_state *csp);
static jb_err header_tagger(struct client_state *csp, char *header);
static jb_err parse_header_time(const char *header_time, time_t *result);
+static jb_err parse_time_header(const char *header, time_t *result);
static jb_err crumble (struct client_state *csp, char **header);
static jb_err filter_header (struct client_state *csp, char **header);
static jb_err crunch_client_header (struct client_state *csp, char **header);
static jb_err client_x_filter (struct client_state *csp, char **header);
static jb_err client_range (struct client_state *csp, char **header);
+static jb_err client_expect (struct client_state *csp, char **header);
static jb_err server_set_cookie (struct client_state *csp, char **header);
static jb_err server_connection (struct client_state *csp, char **header);
static jb_err server_content_type (struct client_state *csp, char **header);
#ifdef FEATURE_CONNECTION_KEEP_ALIVE
static jb_err server_proxy_connection_adder(struct client_state *csp);
#endif /* def FEATURE_CONNECTION_KEEP_ALIVE */
+static jb_err proxy_authentication(struct client_state *csp, char **header);
static jb_err create_forged_referrer(char **header, const char *hostport);
static jb_err create_fake_referrer(char **header, const char *fake_referrer);
{ "Request-Range:", 14, client_range },
{ "If-Range:", 9, client_range },
{ "X-Filter:", 9, client_x_filter },
+ { "Proxy-Authorization:", 20, proxy_authentication },
#if 0
{ "Transfer-Encoding:", 18, client_transfer_encoding },
#endif
+ { "Expect:", 7, client_expect },
{ "*", 0, crunch_client_header },
{ "*", 0, filter_header },
{ NULL, 0, NULL }
{ "Transfer-Encoding:", 18, server_transfer_coding },
{ "content-disposition:", 20, server_content_disposition },
{ "Last-Modified:", 14, server_last_modified },
+ { "Proxy-Authenticate:", 19, proxy_authentication },
{ "*", 0, crunch_server_header },
{ "*", 0, filter_header },
{ NULL, 0, NULL }
void clear_iob(struct iob *iob)
{
free(iob->buf);
- memset(iob, '\0', sizeof(*iob));;
+ memset(iob, '\0', sizeof(*iob));
}
int status; /* return status of the inflate() call */
z_stream zstr; /* used by calls to zlib */
+#ifdef FUZZ
+ assert(csp->iob->cur - csp->iob->buf >= 0);
+ assert(csp->iob->eod - csp->iob->cur >= 0);
+#else
assert(csp->iob->cur - csp->iob->buf > 0);
assert(csp->iob->eod - csp->iob->cur > 0);
+#endif
bufsize = csp->iob->size;
skip_size = (size_t)(csp->iob->cur - csp->iob->buf);
* This is to protect the parsing of gzipped data,
* but it should(?) be valid for deflated data also.
*/
- log_error(LOG_LEVEL_ERROR, "Buffer too small decompressing iob");
+ log_error(LOG_LEVEL_ERROR,
+ "Insufficient data to start decompression. Bytes in buffer: %d",
+ csp->iob->eod - csp->iob->cur);
return JB_ERR_COMPRESS;
}
* Make sure the new uncompressed iob obeys some minimal
* consistency conditions.
*/
- if ((csp->iob->buf < csp->iob->cur)
+ if ((csp->iob->buf <= csp->iob->cur)
&& (csp->iob->cur <= csp->iob->eod)
&& (csp->iob->eod <= csp->iob->buf + csp->iob->size))
{
*
* Function : normalize_lws
*
- * Description : Reduces unquoted linear white space in headers
- * to a single space in accordance with RFC 2616 2.2.
+ * Description : Reduces unquoted linear whitespace in headers to
+ * a single space in accordance with RFC 7230 3.2.4.
* This simplifies parsing and filtering later on.
*
- * XXX: Remove log messages before
- * the next stable release?
- *
* Parameters :
- * 1 : header = A header with linear white space to reduce.
+ * 1 : header = A header with linear whitespace to reduce.
*
* Returns : N/A
*
{
q++;
}
- log_error(LOG_LEVEL_HEADER, "Reducing white space in '%s'", header);
+ log_error(LOG_LEVEL_HEADER, "Reducing whitespace in '%s'", header);
string_move(p+1, q);
}
return;
}
+
/*********************************************************************
*
* Function : sed
* server and header filtering.
*
* Returns : JB_ERR_OK in case off success, or
- * JB_ERR_MEMORY on out-of-memory error.
+ * JB_ERR_MEMORY on some out-of-memory errors, or
+ * JB_ERR_PARSE in case of fatal parse errors.
*
*********************************************************************/
jb_err sed(struct client_state *csp, int filter_server_headers)
const add_header_func_ptr *f;
jb_err err = JB_ERR_OK;
+ scan_headers(csp);
+
if (filter_server_headers)
{
v = server_patterns;
f = add_server_headers;
+ check_negative_tag_patterns(csp, PATTERN_SPEC_NO_RESPONSE_TAG_PATTERN);
}
else
{
v = client_patterns;
f = add_client_headers;
+ check_negative_tag_patterns(csp, PATTERN_SPEC_NO_REQUEST_TAG_PATTERN);
}
- scan_headers(csp);
-
- while ((err == JB_ERR_OK) && (v->str != NULL))
+ while (v->str != NULL)
{
- for (p = csp->headers->first; (err == JB_ERR_OK) && (p != NULL); p = p->next)
+ for (p = csp->headers->first; p != NULL; p = p->next)
{
/* Header crunch()ed in previous run? -> ignore */
if (p->str == NULL) continue;
(v->len == CHECK_EVERY_HEADER_REMAINING))
{
err = v->parser(csp, &(p->str));
+ if (err != JB_ERR_OK)
+ {
+ return err;
+ }
}
}
v++;
*********************************************************************/
static jb_err header_tagger(struct client_state *csp, char *header)
{
- int wanted_filter_type;
+ enum filter_type wanted_filter_type;
int multi_action_index;
- int i;
pcrs_job *job;
- struct file_list *fl;
struct re_filterfile_spec *b;
struct list_entry *tag_name;
multi_action_index = ACTION_MULTI_CLIENT_HEADER_TAGGER;
}
- if (filters_available(csp) == FALSE)
+ if (list_is_empty(csp->action->multi[multi_action_index])
+ || filters_available(csp) == FALSE)
{
- log_error(LOG_LEVEL_ERROR, "Inconsistent configuration: "
- "tagging enabled, but no taggers available.");
+ /* Return early if no taggers apply or if none are available. */
return JB_ERR_OK;
}
- for (i = 0; i < MAX_AF_FILES; i++)
+ /* Execute all applying taggers */
+ for (tag_name = csp->action->multi[multi_action_index]->first;
+ NULL != tag_name; tag_name = tag_name->next)
{
- fl = csp->rlist[i];
- if ((NULL == fl) || (NULL == fl->f))
+ char *modified_tag = NULL;
+ char *tag = header;
+ size_t size = header_length;
+ pcrs_job *joblist;
+
+ b = get_filter(csp, tag_name->str, wanted_filter_type);
+ if (b == NULL)
{
- /*
- * Either there are no filter files
- * left, or this filter file just
- * contains no valid filters.
- *
- * Continue to be sure we don't miss
- * valid filter files that are chained
- * after empty or invalid ones.
- */
continue;
}
- /* For all filters, */
- for (b = fl->f; b; b = b->next)
+ joblist = b->joblist;
+
+ if (b->dynamic) joblist = compile_dynamic_pcrs_job_list(csp, b);
+
+ if (NULL == joblist)
+ {
+ log_error(LOG_LEVEL_RE_FILTER,
+ "Tagger %s has empty joblist. Nothing to do.", b->name);
+ continue;
+ }
+
+ /* execute their pcrs_joblist on the header. */
+ for (job = joblist; NULL != job; job = job->next)
{
- if (b->type != wanted_filter_type)
+ const int hits = pcrs_execute(job, tag, size, &modified_tag, &size);
+
+ if (0 < hits)
{
- /* skip the ones we don't care about, */
- continue;
+ /* Success, continue with the modified version. */
+ if (tag != header)
+ {
+ freez(tag);
+ }
+ tag = modified_tag;
}
- /* leaving only taggers that could apply, of which we use the ones, */
- for (tag_name = csp->action->multi[multi_action_index]->first;
- NULL != tag_name; tag_name = tag_name->next)
+ else
{
- /* that do apply, and */
- if (strcmp(b->name, tag_name->str) == 0)
+ /* Tagger doesn't match */
+ if (0 > hits)
{
- char *modified_tag = NULL;
- char *tag = header;
- size_t size = header_length;
- pcrs_job *joblist = b->joblist;
+ /* Regex failure, log it but continue anyway. */
+ assert(NULL != header);
+ log_error(LOG_LEVEL_ERROR,
+ "Problems with tagger \'%s\' and header \'%s\': %s",
+ b->name, *header, pcrs_strerror(hits));
+ }
+ freez(modified_tag);
+ }
+ }
+
+ if (b->dynamic) pcrs_free_joblist(joblist);
- if (b->dynamic) joblist = compile_dynamic_pcrs_job_list(csp, b);
+ /* If this tagger matched */
+ if (tag != header)
+ {
+ if (0 == size)
+ {
+ /*
+ * There is no technical limitation which makes
+ * it impossible to use empty tags, but I assume
+ * no one would do it intentionally.
+ */
+ freez(tag);
+ log_error(LOG_LEVEL_INFO,
+ "Tagger \'%s\' created an empty tag. Ignored.", b->name);
+ continue;
+ }
- if (NULL == joblist)
+ if (!list_contains_item(csp->tags, tag))
+ {
+ if (JB_ERR_OK != enlist(csp->tags, tag))
+ {
+ log_error(LOG_LEVEL_ERROR,
+ "Insufficient memory to add tag \'%s\', "
+ "based on tagger \'%s\' and header \'%s\'",
+ tag, b->name, *header);
+ }
+ else
+ {
+ char *action_message;
+ /*
+ * update the action bits right away, to make
+ * tagging based on tags set by earlier taggers
+ * of the same kind possible.
+ */
+ if (update_action_bits_for_tag(csp, tag))
{
- log_error(LOG_LEVEL_RE_FILTER,
- "Tagger %s has empty joblist. Nothing to do.", b->name);
- continue;
+ action_message = "Action bits updated accordingly.";
}
-
- /* execute their pcrs_joblist on the header. */
- for (job = joblist; NULL != job; job = job->next)
+ else
{
- const int hits = pcrs_execute(job, tag, size, &modified_tag, &size);
-
- if (0 < hits)
- {
- /* Success, continue with the modified version. */
- if (tag != header)
- {
- freez(tag);
- }
- tag = modified_tag;
- }
- else
- {
- /* Tagger doesn't match */
- if (0 > hits)
- {
- /* Regex failure, log it but continue anyway. */
- assert(NULL != header);
- log_error(LOG_LEVEL_ERROR,
- "Problems with tagger \'%s\' and header \'%s\': %s",
- b->name, *header, pcrs_strerror(hits));
- }
- freez(modified_tag);
- }
+ action_message = "No action bits update necessary.";
}
- if (b->dynamic) pcrs_free_joblist(joblist);
-
- /* If this tagger matched */
- if (tag != header)
- {
- if (0 == size)
- {
- /*
- * There is to technical limitation which makes
- * it impossible to use empty tags, but I assume
- * no one would do it intentionally.
- */
- freez(tag);
- log_error(LOG_LEVEL_INFO,
- "Tagger \'%s\' created an empty tag. Ignored.",
- b->name);
- continue;
- }
-
- if (!list_contains_item(csp->tags, tag))
- {
- if (JB_ERR_OK != enlist(csp->tags, tag))
- {
- log_error(LOG_LEVEL_ERROR,
- "Insufficient memory to add tag \'%s\', "
- "based on tagger \'%s\' and header \'%s\'",
- tag, b->name, *header);
- }
- else
- {
- char *action_message;
- /*
- * update the action bits right away, to make
- * tagging based on tags set by earlier taggers
- * of the same kind possible.
- */
- if (update_action_bits_for_tag(csp, tag))
- {
- action_message = "Action bits updated accordingly.";
- }
- else
- {
- action_message = "No action bits update necessary.";
- }
-
- log_error(LOG_LEVEL_HEADER,
- "Tagger \'%s\' added tag \'%s\'. %s",
- b->name, tag, action_message);
- }
- }
- else
- {
- /* XXX: Is this log-worthy? */
- log_error(LOG_LEVEL_HEADER,
- "Tagger \'%s\' didn't add tag \'%s\'. "
- "Tag already present", b->name, tag);
- }
- freez(tag);
- } /* if the tagger matched */
- } /* if the tagger applies */
- } /* for every tagger that could apply */
- } /* for all filters */
- } /* for all filter files */
+ log_error(LOG_LEVEL_HEADER,
+ "Tagger \'%s\' added tag \'%s\'. %s",
+ b->name, tag, action_message);
+ }
+ }
+ else
+ {
+ /* XXX: Is this log-worthy? */
+ log_error(LOG_LEVEL_HEADER,
+ "Tagger \'%s\' didn't add tag \'%s\'. Tag already present",
+ b->name, tag);
+ }
+ freez(tag);
+ }
+ }
return JB_ERR_OK;
}
char *newheader = NULL;
pcrs_job *job;
- struct file_list *fl;
struct re_filterfile_spec *b;
struct list_entry *filtername;
- int i;
- int wanted_filter_type;
+ enum filter_type wanted_filter_type;
int multi_action_index;
if (csp->flags & CSP_FLAG_NO_FILTERING)
multi_action_index = ACTION_MULTI_CLIENT_HEADER_FILTER;
}
- if (filters_available(csp) == FALSE)
+ if (list_is_empty(csp->action->multi[multi_action_index])
+ || filters_available(csp) == FALSE)
{
- log_error(LOG_LEVEL_ERROR, "Inconsistent configuration: "
- "header filtering enabled, but no matching filters available.");
+ /* Return early if no filters apply or if none are available. */
return JB_ERR_OK;
}
- for (i = 0; i < MAX_AF_FILES; i++)
+ /* Execute all applying header filters */
+ for (filtername = csp->action->multi[multi_action_index]->first;
+ filtername != NULL; filtername = filtername->next)
{
- fl = csp->rlist[i];
- if ((NULL == fl) || (NULL == fl->f))
+ int current_hits = 0;
+ pcrs_job *joblist;
+
+ b = get_filter(csp, filtername->str, wanted_filter_type);
+ if (b == NULL)
{
- /*
- * Either there are no filter files
- * left, or this filter file just
- * contains no valid filters.
- *
- * Continue to be sure we don't miss
- * valid filter files that are chained
- * after empty or invalid ones.
- */
continue;
}
- /*
- * For all applying +filter actions, look if a filter by that
- * name exists and if yes, execute its pcrs_joblist on the
- * buffer.
- */
- for (b = fl->f; b; b = b->next)
- {
- if (b->type != wanted_filter_type)
- {
- /* Skip other filter types */
- continue;
- }
-
- for (filtername = csp->action->multi[multi_action_index]->first;
- filtername ; filtername = filtername->next)
- {
- if (strcmp(b->name, filtername->str) == 0)
- {
- int current_hits = 0;
- pcrs_job *joblist = b->joblist;
-
- if (b->dynamic) joblist = compile_dynamic_pcrs_job_list(csp, b);
- if (NULL == joblist)
- {
- log_error(LOG_LEVEL_RE_FILTER, "Filter %s has empty joblist. Nothing to do.", b->name);
- continue;
- }
+ joblist = b->joblist;
- log_error(LOG_LEVEL_RE_FILTER, "filtering \'%s\' (size %d) with \'%s\' ...",
- *header, size, b->name);
+ if (b->dynamic) joblist = compile_dynamic_pcrs_job_list(csp, b);
- /* Apply all jobs from the joblist */
- for (job = joblist; NULL != job; job = job->next)
- {
- matches = pcrs_execute(job, *header, size, &newheader, &size);
- if (0 < matches)
- {
- current_hits += matches;
- log_error(LOG_LEVEL_HEADER, "Transforming \"%s\" to \"%s\"", *header, newheader);
- freez(*header);
- *header = newheader;
- }
- else if (0 == matches)
- {
- /* Filter doesn't change header */
- freez(newheader);
- }
- else
- {
- /* RegEx failure */
- log_error(LOG_LEVEL_ERROR, "Filtering \'%s\' with \'%s\' didn't work out: %s",
- *header, b->name, pcrs_strerror(matches));
- if (newheader != NULL)
- {
- log_error(LOG_LEVEL_ERROR, "Freeing what's left: %s", newheader);
- freez(newheader);
- }
- }
- }
+ if (NULL == joblist)
+ {
+ log_error(LOG_LEVEL_RE_FILTER, "Filter %s has empty joblist. Nothing to do.", b->name);
+ continue;
+ }
- if (b->dynamic) pcrs_free_joblist(joblist);
+ log_error(LOG_LEVEL_RE_FILTER, "filtering \'%s\' (size %d) with \'%s\' ...",
+ *header, size, b->name);
- log_error(LOG_LEVEL_RE_FILTER, "... produced %d hits (new size %d).", current_hits, size);
- hits += current_hits;
+ /* Apply all jobs from the joblist */
+ for (job = joblist; NULL != job; job = job->next)
+ {
+ matches = pcrs_execute(job, *header, size, &newheader, &size);
+ if (0 < matches)
+ {
+ current_hits += matches;
+ log_error(LOG_LEVEL_HEADER, "Transforming \"%s\" to \"%s\"", *header, newheader);
+ freez(*header);
+ *header = newheader;
+ }
+ else if (0 == matches)
+ {
+ /* Filter doesn't change header */
+ freez(newheader);
+ }
+ else
+ {
+ /* RegEx failure */
+ log_error(LOG_LEVEL_ERROR, "Filtering \'%s\' with \'%s\' didn't work out: %s",
+ *header, b->name, pcrs_strerror(matches));
+ if (newheader != NULL)
+ {
+ log_error(LOG_LEVEL_ERROR, "Freeing what's left: %s", newheader);
+ freez(newheader);
}
}
}
+
+ if (b->dynamic) pcrs_free_joblist(joblist);
+
+ log_error(LOG_LEVEL_RE_FILTER, "... produced %d hits (new size %d).", current_hits, size);
+ hits += current_hits;
}
/*
csp->flags |= CSP_FLAG_SERVER_KEEP_ALIVE_TIMEOUT_SET;
}
+ freez(*header);
+
return JB_ERR_OK;
}
}
+/*********************************************************************
+ *
+ * Function : proxy_authentication
+ *
+ * Description : Removes headers that are relevant for proxy
+ * authentication unless forwarding them has
+ * been explicitly requested.
+ *
+ * Parameters :
+ * 1 : csp = Current client state (buffers, headers, etc...)
+ * 2 : header = On input, pointer to header to modify.
+ * On output, pointer to the modified header, or NULL
+ * to remove the header. This function frees the
+ * original string if necessary.
+ *
+ * Returns : JB_ERR_OK.
+ *
+ *********************************************************************/
+static jb_err proxy_authentication(struct client_state *csp, char **header)
+{
+ if ((csp->config->feature_flags &
+ RUNTIME_FEATURE_FORWARD_PROXY_AUTHENTICATION_HEADERS) == 0) {
+ log_error(LOG_LEVEL_HEADER,
+ "Forwarding proxy authentication headers is disabled. Crunching: %s", *header);
+ freez(*header);
+ }
+ return JB_ERR_OK;
+}
+
+
/*********************************************************************
*
* Function : client_keep_alive
static jb_err client_keep_alive(struct client_state *csp, char **header)
{
unsigned int keep_alive_timeout;
- const char *timeout_position = strstr(*header, ": ");
+ char *timeout_position;
if (!(csp->config->feature_flags & RUNTIME_FEATURE_CONNECTION_KEEP_ALIVE))
{
return JB_ERR_OK;
}
+ /* Check for parameter-less format "Keep-Alive: 100" */
+ timeout_position = strstr(*header, ": ");
if ((NULL == timeout_position)
|| (1 != sscanf(timeout_position, ": %u", &keep_alive_timeout)))
{
- log_error(LOG_LEVEL_ERROR, "Couldn't parse: %s", *header);
- }
- else
- {
- if (keep_alive_timeout < csp->config->keep_alive_timeout)
- {
- log_error(LOG_LEVEL_HEADER,
- "Reducing keep-alive timeout from %u to %u.",
- csp->config->keep_alive_timeout, keep_alive_timeout);
- csp->server_connection.keep_alive_timeout = keep_alive_timeout;
- }
- else
+ /* Assume parameter format "Keep-Alive: timeout=100" */
+ timeout_position = strstr(*header, "timeout=");
+ if ((NULL == timeout_position)
+ || (1 != sscanf(timeout_position, "timeout=%u", &keep_alive_timeout)))
{
- /* XXX: Is this log worthy? */
log_error(LOG_LEVEL_HEADER,
- "Client keep-alive timeout is %u. Sticking with %u.",
- keep_alive_timeout, csp->config->keep_alive_timeout);
+ "Couldn't parse: '%s'. Using default timeout %u",
+ *header, csp->config->keep_alive_timeout);
+ freez(*header);
+
+ return JB_ERR_OK;
}
}
+ if (keep_alive_timeout < csp->config->keep_alive_timeout)
+ {
+ log_error(LOG_LEVEL_HEADER,
+ "Reducing keep-alive timeout from %u to %u.",
+ csp->config->keep_alive_timeout, keep_alive_timeout);
+ csp->server_connection.keep_alive_timeout = keep_alive_timeout;
+ }
+ else
+ {
+ /* XXX: Is this log worthy? */
+ log_error(LOG_LEVEL_HEADER,
+ "Client keep-alive timeout is %u. Sticking with %u.",
+ keep_alive_timeout, csp->config->keep_alive_timeout);
+ freez(*header);
+ }
+
return JB_ERR_OK;
}
static jb_err get_content_length(const char *header_value, unsigned long long *length)
{
#ifdef _WIN32
- assert(sizeof(unsigned long long) > 4);
+#if SIZEOF_LONG_LONG < 8
+#error sizeof(unsigned long long) too small
+#endif
if (1 != sscanf(header_value, "%I64u", length))
#else
if (1 != sscanf(header_value, "%llu", length))
}
+/*********************************************************************
+ *
+ * Function : client_expect
+ *
+ * Description : Raise the CSP_FLAG_UNSUPPORTED_CLIENT_EXPECTATION
+ * if the Expect header value is unsupported.
+ *
+ * Rejecting unsupported expectations is a RFC 7231 5.1.1
+ * MAY and a RFC 2616 (obsolete) MUST.
+ *
+ * Parameters :
+ * 1 : csp = Current client state (buffers, headers, etc...)
+ * 2 : header = On input, pointer to header to modify.
+ * On output, pointer to the modified header, or NULL
+ * to remove the header. This function frees the
+ * original string if necessary.
+ *
+ * Returns : JB_ERR_OK on success, or
+ *
+ *********************************************************************/
+jb_err client_expect(struct client_state *csp, char **header)
+{
+ if (0 != strcmpic(*header, "Expect: 100-continue"))
+ {
+ csp->flags |= CSP_FLAG_UNSUPPORTED_CLIENT_EXPECTATION;
+ log_error(LOG_LEVEL_HEADER,
+ "Unsupported client expectaction: %s", *header);
+ }
+
+ return JB_ERR_OK;
+
+}
+
+
/*********************************************************************
*
* Function : crumble
/* Remove header if it isn't the first Content-Type header */
if ((csp->content_type & CT_DECLARED))
{
- /*
- * Another, slightly slower, way to see if
- * we already parsed another Content-Type header.
- */
- assert(NULL != get_header_value(csp->headers, "Content-Type:"));
-
- log_error(LOG_LEVEL_ERROR,
- "Multiple Content-Type headers. Removing and ignoring: \'%s\'",
- *header);
- freez(*header);
-
+ if (content_filters_enabled(csp->action))
+ {
+ /*
+ * Making sure the client interprets the content the same way
+ * Privoxy did is only relevant if Privoxy modified it.
+ *
+ * Checking for this is "hard" as it's not yet known when
+ * this function is called, thus go shopping and and just
+ * check if Privoxy could filter it.
+ *
+ * The main thing is that we don't mess with the headers
+ * unless the user signalled that it's acceptable.
+ */
+ log_error(LOG_LEVEL_HEADER,
+ "Multiple Content-Type headers detected. "
+ "Removing and ignoring: %s",
+ *header);
+ freez(*header);
+ }
return JB_ERR_OK;
}
*/
if ((csp->content_type & CT_TEXT) || (csp->action->flags & ACTION_FORCE_TEXT_MODE))
{
+ jb_err err;
freez(*header);
*header = strdup_or_die("Content-Type: ");
- string_append(header, csp->action->string[ACTION_STRING_CONTENT_TYPE]);
- if (header == NULL)
+ err = string_append(header, csp->action->string[ACTION_STRING_CONTENT_TYPE]);
+ if (JB_ERR_OK != err)
{
log_error(LOG_LEVEL_HEADER, "Insufficient memory to replace Content-Type!");
return JB_ERR_MEMORY;
/*
* Log a warning if the user expects the content to be filtered.
*/
- if ((csp->rlist != NULL) &&
- (!list_is_empty(csp->action->multi[ACTION_MULTI_FILTER])))
+ if (content_filters_enabled(csp->action))
{
log_error(LOG_LEVEL_INFO,
"SDCH-compressed content detected, content filtering disabled. "
/*
* Log a warning if the user expects the content to be filtered.
*/
- if ((csp->rlist != NULL) &&
- (!list_is_empty(csp->action->multi[ACTION_MULTI_FILTER])))
+ if (content_filters_enabled(csp->action))
{
log_error(LOG_LEVEL_INFO,
"Compressed content detected, content filtering disabled. "
}
else if (0 == strcmpic(newval, "randomize"))
{
- const char *header_time = *header + sizeof("Last-Modified:");
-
log_error(LOG_LEVEL_HEADER, "Randomizing: %s", *header);
- if (JB_ERR_OK != parse_header_time(header_time, &last_modified))
+ if (JB_ERR_OK != parse_time_header(*header, &last_modified))
{
- log_error(LOG_LEVEL_HEADER, "Couldn't parse: %s in %s (crunching!)", header_time, *header);
+ log_error(LOG_LEVEL_HEADER,
+ "Couldn't parse time in %s (crunching!)", *header);
freez(*header);
}
else
* port information, parse and evaluate the Host
* header field.
*
- * Also, kill ill-formed HOST: headers as sent by
- * Apple's iTunes software when used with a proxy.
- *
* Parameters :
* 1 : csp = Current client state (buffers, headers, etc...)
* 2 : header = On input, pointer to header to modify.
{
char *p, *q;
- /*
- * If the header field name is all upper-case, chances are that it's
- * an ill-formed one from iTunes. BTW, killing innocent headers here is
- * not a problem -- they are regenerated later.
- */
- if ((*header)[1] == 'O')
+ if (strlen(*header) < 7)
{
- log_error(LOG_LEVEL_HEADER, "Killed all-caps Host header line: %s", *header);
+ log_error(LOG_LEVEL_HEADER, "Removing empty Host header");
freez(*header);
return JB_ERR_OK;
}
}
else /* add random value */
{
- const char *header_time = *header + sizeof("If-Modified-Since:");
-
- if (JB_ERR_OK != parse_header_time(header_time, &tm))
+ if (JB_ERR_OK != parse_time_header(*header, &tm))
{
- log_error(LOG_LEVEL_HEADER, "Couldn't parse: %s in %s (crunching!)", header_time, *header);
+ log_error(LOG_LEVEL_HEADER,
+ "Couldn't parse time in %s (crunching!)", *header);
freez(*header);
}
else
if (!csp->http->hostport || !*(csp->http->hostport))
{
- /* XXX: When does this happen and why is it OK? */
- log_error(LOG_LEVEL_INFO, "Weirdness in client_host_adder detected and ignored.");
- return JB_ERR_OK;
+ log_error(LOG_LEVEL_ERROR, "Destination host unknown.");
+ return JB_ERR_PARSE;
}
/*
* Function : client_connection_header_adder
*
* Description : Adds a proper "Connection:" header to csp->headers
- * unless the header was already present. Called from `sed'.
+ * unless the header was already present or it's a
+ * CONNECT request. Called from `sed'.
*
* Parameters :
* 1 : csp = Current client state (buffers, headers, etc...)
return JB_ERR_OK;
}
+ /*
+ * In case of CONNECT requests "Connection: close" is implied,
+ * but actually setting the header has been reported to cause
+ * problems with some forwarding proxies that close the
+ * connection prematurely.
+ */
+ if (csp->http->ssl != 0)
+ {
+ return JB_ERR_OK;
+ }
+
#ifdef FEATURE_CONNECTION_KEEP_ALIVE
if ((csp->config->feature_flags & RUNTIME_FEATURE_CONNECTION_KEEP_ALIVE)
&& !(csp->flags & CSP_FLAG_SERVER_SOCKET_TAINTED)
- && (csp->http->ssl == 0)
&& !strcmpic(csp->http->ver, "HTTP/1.1"))
{
csp->flags |= CSP_FLAG_CLIENT_CONNECTION_KEEP_ALIVE;
* is a partial range (HTTP status 206)
* - Rewrite HTTP/1.1 answers to HTTP/1.0 if +downgrade
* action applies.
+ * - Normalize the HTTP-version.
*
* Parameters :
* 1 : csp = Current client state (buffers, headers, etc...)
* original string if necessary.
*
* Returns : JB_ERR_OK on success, or
- * JB_ERR_MEMORY on out-of-memory error.
+ * JB_ERR_PARSE on fatal parse errors.
*
*********************************************************************/
static jb_err server_http(struct client_state *csp, char **header)
{
- sscanf(*header, "HTTP/%*d.%*d %d", &(csp->http->status));
+ char *reason_phrase = NULL;
+ char *new_response_line;
+ char *p;
+ size_t length;
+ unsigned int major_version;
+ unsigned int minor_version;
+
+ /* Get the reason phrase which start after the second whitespace */
+ p = strchr(*header, ' ');
+ if (NULL != p)
+ {
+ p++;
+ reason_phrase = strchr(p, ' ');
+ }
+
+ if (reason_phrase != NULL)
+ {
+ reason_phrase++;
+ }
+ else
+ {
+ log_error(LOG_LEVEL_ERROR,
+ "Response line lacks reason phrase: %s", *header);
+ reason_phrase="";
+ }
+
+ if (3 != sscanf(*header, "HTTP/%u.%u %d", &major_version,
+ &minor_version, &(csp->http->status)))
+ {
+ log_error(LOG_LEVEL_ERROR,
+ "Failed to parse the response line: %s", *header);
+ return JB_ERR_PARSE;
+ }
+
if (csp->http->status == 206)
{
csp->content_type = CT_TABOO;
}
- if ((csp->action->flags & ACTION_DOWNGRADE) != 0)
+ if (major_version != 1 || (minor_version != 0 && minor_version != 1))
{
- /* XXX: Should we do a real validity check here? */
- if (strlen(*header) > 8)
- {
- (*header)[7] = '0';
- log_error(LOG_LEVEL_HEADER, "Downgraded answer to HTTP/1.0");
- }
- else
- {
- /*
- * XXX: Should we block the request or
- * enlist a valid status code line here?
- */
- log_error(LOG_LEVEL_INFO, "Malformed server response detected. "
- "Downgrading to HTTP/1.0 impossible.");
- }
+ /*
+ * According to RFC 7230 2.6 intermediaries MUST send
+ * their own HTTP-version in forwarded messages.
+ */
+ log_error(LOG_LEVEL_ERROR,
+ "Unsupported HTTP version. Downgrading to 1.1.");
+ major_version = 1;
+ minor_version = 1;
+ }
+
+ if (((csp->action->flags & ACTION_DOWNGRADE) != 0) && (minor_version == 1))
+ {
+ log_error(LOG_LEVEL_HEADER, "Downgrading answer to HTTP/1.0");
+ minor_version = 0;
+ }
+
+ /* Rebuild response line. */
+ length = sizeof("HTTP/1.1 200 ") + strlen(reason_phrase) + 1;
+ new_response_line = malloc_or_die(length);
+
+ snprintf(new_response_line, length, "HTTP/%u.%u %d %s",
+ major_version, minor_version, csp->http->status, reason_phrase);
+
+ if (0 != strcmp(*header, new_response_line))
+ {
+ log_error(LOG_LEVEL_HEADER, "Response line '%s' changed to '%s'",
+ *header, new_response_line);
}
+ freez(*header);
+ *header = new_response_line;
+
return JB_ERR_OK;
}
{
log_error(LOG_LEVEL_FATAL, "Invalid cookie lifetime limit: %s", param);
}
- cookie_lifetime *= 60U;
+ cookie_lifetime *= 60;
}
/* Loop through each tag in the cookie */
continue;
}
*result = timegm(&gmt);
+
+#ifdef FEATURE_STRPTIME_SANITY_CHECKS
+ /*
+ * Verify that parsing the date recreated from the first
+ * parse operation gets the previous result. If it doesn't,
+ * either strptime() or strftime() are malfunctioning.
+ *
+ * We could string-compare the recreated date with the original
+ * header date, but this leads to false positives as strptime()
+ * may let %a accept all day formats while strftime() will only
+ * create one.
+ */
+ {
+ char recreated_date[100];
+ struct tm *tm;
+ time_t result2;
+
+ tm = gmtime(result);
+ if (!strftime(recreated_date, sizeof(recreated_date),
+ time_formats[i], tm))
+ {
+ log_error(LOG_LEVEL_ERROR, "Failed to recreate date '%s' with '%s'.",
+ header_time, time_formats[i]);
+ continue;
+ }
+ memset(&gmt, 0, sizeof(gmt));
+ if (NULL == strptime(recreated_date, time_formats[i], &gmt))
+ {
+ log_error(LOG_LEVEL_ERROR,
+ "Failed to parse '%s' generated with '%s' to recreate '%s'.",
+ recreated_date, time_formats[i], header_time);
+ continue;
+ }
+ result2 = timegm(&gmt);
+ if (*result != result2)
+ {
+ log_error(LOG_LEVEL_ERROR, "strftime() and strptime() disagree. "
+ "Format: '%s'. In: '%s', out: '%s'. %d != %d. Rejecting.",
+ time_formats[i], header_time, recreated_date, *result, result2);
+ continue;
+ }
+ }
+#endif
+
return JB_ERR_OK;
}
}
}
+/*********************************************************************
+ *
+ * Function : parse_time_header
+ *
+ * Description : Parses the time in an HTTP time header to get
+ * the numerical respresentation.
+ *
+ * Parameters :
+ * 1 : header = HTTP header with a time value
+ * 2 : result = storage for header_time in seconds
+ *
+ * Returns : JB_ERR_OK if the time format was recognized, or
+ * JB_ERR_PARSE otherwise.
+ *
+ *********************************************************************/
+static jb_err parse_time_header(const char *header, time_t *result)
+{
+ const char *header_time;
+
+ header_time = strchr(header, ':');
+
+ /*
+ * Currently this can't happen as all callers are called
+ * through sed() which requires a header name followed by
+ * a colon.
+ */
+ assert(header_time != NULL);
+
+ header_time++;
+ if (*header_time == ' ')
+ {
+ header_time++;
+ }
+
+ return parse_header_time(header_time, result);
+
+}
+
/*********************************************************************
*
char *p;
char *host;
+ assert(!http->ssl);
+
host = get_header_value(headers, "Host:");
if (NULL == host)
}
else
{
- http->port = http->ssl ? 443 : 80;
+ http->port = 80;
}
/* Rebuild request URL */
freez(http->url);
- http->url = strdup(http->ssl ? "https://" : "http://");
+ http->url = strdup("http://");
string_append(&http->url, http->hostport);
string_append(&http->url, http->path);
if (http->url == NULL)
return JB_ERR_MEMORY;
}
- log_error(LOG_LEVEL_HEADER, "Destination extracted from \"Host:\" header. New request URL: %s",
+ log_error(LOG_LEVEL_HEADER,
+ "Destination extracted from \"Host\" header. New request URL: %s",
http->url);
+ /*
+ * Regenerate request line in "proxy format"
+ * to make rewrites more convenient.
+ */
+ assert(http->cmd != NULL);
+ freez(http->cmd);
+ http->cmd = strdup_or_die(http->gpc);
+ string_append(&http->cmd, " ");
+ string_append(&http->cmd, http->url);
+ string_append(&http->cmd, " ");
+ string_append(&http->cmd, http->ver);
+ if (http->cmd == NULL)
+ {
+ return JB_ERR_MEMORY;
+ }
+
return JB_ERR_OK;
}
static void create_content_length_header(unsigned long long content_length,
char *header, size_t buffer_length)
{
+#ifdef _WIN32
+#if SIZEOF_LONG_LONG < 8
+#error sizeof(unsigned long long) too small
+#endif
+ snprintf(header, buffer_length, "Content-Length: %I64u", content_length);
+#else
snprintf(header, buffer_length, "Content-Length: %llu", content_length);
+#endif
}