-const char parsers_rcs[] = "$Id: parsers.c,v 1.239 2011/12/31 14:49:07 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) 2001-2011 the
- * Privoxy team. http://www.privoxy.org/
+ * Copyright : Written by and Copyright (C) 2001-2020 the
+ * Privoxy team. https://www.privoxy.org/
*
* Based on the Internet Junkbuster originally written
* by and Copyright (C) 1997 Anonymous Coders and
#include "strptime.h"
#endif
-const char parsers_h_rcs[] = PARSERS_H_VERSION;
-
-/* Fix a problem with Solaris. There should be no effect on other
- * platforms.
- * Solaris's isspace() is a macro which uses its argument directly
- * as an array index. Therefore we need to make sure that high-bit
- * characters generate +ve values, and ideally we also want to make
- * the argument match the declared parameter type of "int".
- *
- * Why did they write a character function that can't take a simple
- * "char" argument? Doh!
- */
-#define ijb_isupper(__X) isupper((int)(unsigned char)(__X))
-#define ijb_tolower(__X) tolower((int)(unsigned char)(__X))
-
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);
static jb_err server_proxy_connection(struct client_state *csp, char **header);
static jb_err client_keep_alive(struct client_state *csp, char **header);
static jb_err client_save_content_length(struct client_state *csp, char **header);
+static jb_err client_proxy_connection(struct client_state *csp, char **header);
#endif /* def FEATURE_CONNECTION_KEEP_ALIVE */
static jb_err client_host_adder (struct client_state *csp);
#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);
#ifdef FEATURE_CONNECTION_KEEP_ALIVE
{ "Keep-Alive:", 11, client_keep_alive },
{ "Content-Length:", 15, client_save_content_length },
+ { "Proxy-Connection:", 17, client_proxy_connection },
#else
{ "Keep-Alive:", 11, crumble },
+ { "Proxy-Connection:", 17, crumble },
#endif
{ "connection:", 11, client_connection },
- { "proxy-connection:", 17, crumble },
{ "max-forwards:", 13, client_max_forwards },
{ "Accept-Language:", 16, client_accept_language },
{ "if-none-match:", 14, client_if_none_match },
{ "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 }
/*********************************************************************
*
- * Function : flush_socket
+ * Function : flush_iob
*
* Description : Write any pending "buffered" content.
*
* Parameters :
* 1 : fd = file descriptor of the socket to read
* 2 : iob = The I/O buffer to flush, usually csp->iob.
+ * 3 : delay = Number of milliseconds to delay the writes
*
* Returns : On success, the number of bytes written are returned (zero
* indicates nothing was written). On error, -1 is returned,
* file, the results are not portable.
*
*********************************************************************/
-long flush_socket(jb_socket fd, struct iob *iob)
+long flush_iob(jb_socket fd, struct iob *iob, unsigned int delay)
{
long len = iob->eod - iob->cur;
return(0);
}
- if (write_socket(fd, iob->cur, (size_t)len))
+ if (write_socket_delayed(fd, iob->cur, (size_t)len, delay))
{
return(-1);
}
*
* Function : add_to_iob
*
- * Description : Add content to the buffered page, expanding the
+ * Description : Add content to the buffer, expanding the
* buffer if necessary.
*
* Parameters :
- * 1 : csp = Current client state (buffers, headers, etc...)
- * 2 : buf = holds the content to be added to the page
- * 3 : n = number of bytes to be added
+ * 1 : iob = Destination buffer.
+ * 2 : buffer_limit = Limit to which the destination may grow
+ * 3 : src = holds the content to be added
+ * 4 : n = number of bytes to be added
*
* Returns : JB_ERR_OK on success, JB_ERR_MEMORY if out-of-memory
* or buffer limit reached.
*
*********************************************************************/
-jb_err add_to_iob(struct client_state *csp, char *buf, long n)
+jb_err add_to_iob(struct iob *iob, const size_t buffer_limit, char *src, long n)
{
- struct iob *iob = csp->iob;
size_t used, offset, need;
char *p;
* If the buffer can't hold the new data, extend it first.
* Use the next power of two if possible, else use the actual need.
*/
- if (need > csp->config->buffer_limit)
+ if (need > buffer_limit)
{
log_error(LOG_LEVEL_INFO,
"Buffer limit reached while extending the buffer (iob). Needed: %d. Limit: %d",
- need, csp->config->buffer_limit);
+ need, buffer_limit);
return JB_ERR_MEMORY;
}
if (need > iob->size)
{
- size_t want = csp->iob->size ? csp->iob->size : 512;
+ size_t want = iob->size ? iob->size : 512;
while (want <= need)
{
want *= 2;
}
- if (want <= csp->config->buffer_limit && NULL != (p = (char *)realloc(iob->buf, want)))
+ if (want <= buffer_limit && NULL != (p = (char *)realloc(iob->buf, want)))
{
iob->size = want;
}
}
/* copy the new data into the iob buffer */
- memcpy(iob->eod, buf, (size_t)n);
+ memcpy(iob->eod, src, (size_t)n);
/* point to the end of the data */
iob->eod += n;
}
+/*********************************************************************
+ *
+ * Function : clear_iob
+ *
+ * Description : Frees the memory allocated for an I/O buffer and
+ * resets the structure.
+ *
+ * Parameters :
+ * 1 : iob = I/O buffer to clear.
+ *
+ * Returns : JB_ERR_OK on success, JB_ERR_MEMORY if out-of-memory
+ * or buffer limit reached.
+ *
+ *********************************************************************/
+void clear_iob(struct iob *iob)
+{
+ free(iob->buf);
+ memset(iob, '\0', sizeof(*iob));
+}
+
+
#ifdef FEATURE_ZLIB
/*********************************************************************
*
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;
}
}
else if (csp->content_type & CT_DEFLATE)
{
- /*
- * XXX: The debug level should be lowered
- * before the next stable release.
- */
- log_error(LOG_LEVEL_INFO, "Decompressing deflated iob: %d", *cur);
/*
* In theory (that is, according to RFC 1950), deflate-compressed
* data should begin with a two-byte zlib header and have an
if (bufsize >= csp->config->buffer_limit)
{
log_error(LOG_LEVEL_ERROR, "Buffer limit reached while decompressing iob");
+ freez(buf);
return JB_ERR_MEMORY;
}
* 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))
{
#endif /* defined(FEATURE_ZLIB) */
-/*********************************************************************
- *
- * Function : string_move
- *
- * Description : memmove wrapper to move the last part of a string
- * towards the beginning, overwriting the part in
- * the middle. strlcpy() can't be used here as the
- * strings overlap.
- *
- * Parameters :
- * 1 : dst = Destination to overwrite
- * 2 : src = Source to move.
- *
- * Returns : N/A
- *
- *********************************************************************/
-static void string_move(char *dst, char *src)
-{
- assert(dst < src);
-
- /* +1 to copy the terminating nul as well. */
- memmove(dst, src, strlen(src)+1);
-}
-
-
/*********************************************************************
*
* 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
*
while (*p != '\0')
{
- if (ijb_isspace(*p) && ijb_isspace(*(p+1)))
+ if (privoxy_isspace(*p) && privoxy_isspace(*(p+1)))
{
char *q = p+1;
- while (ijb_isspace(*q))
+ while (privoxy_isspace(*q))
{
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);
}
}
p = strchr(header, ':');
- if ((p != NULL) && (p != header) && ijb_isspace(*(p-1)))
+ if ((p != NULL) && (p != header) && privoxy_isspace(*(p-1)))
{
/*
* There's still space before the colon.
* Found: return pointer to start of value
*/
ret = cur_entry->str + length;
- while (*ret && ijb_isspace(*ret)) ret++;
+ while (*ret && privoxy_isspace(*ret)) ret++;
return ret;
}
}
}
+/*********************************************************************
+ *
+ * Function : enforce_header_order
+ *
+ * Description : Enforces a given header order.
+ *
+ * Parameters :
+ * 1 : headers = List of headers to order.
+ * 2 : ordered_headers = List of ordered header names.
+ *
+ * Returns : N/A
+ *
+ *********************************************************************/
+static void enforce_header_order(struct list *headers, const struct list *ordered_headers)
+{
+ struct list_entry *sorted_header;
+ struct list new_headers[1];
+ struct list_entry *header;
+
+ init_list(new_headers);
+
+ /* The request line is always the first "header" */
+
+ assert(NULL != headers->first->str);
+ enlist(new_headers, headers->first->str);
+ freez(headers->first->str)
+
+ /* Enlist the specified headers in the given order */
+
+ for (sorted_header = ordered_headers->first; sorted_header != NULL;
+ sorted_header = sorted_header->next)
+ {
+ const size_t sorted_header_length = strlen(sorted_header->str);
+ for (header = headers->first; header != NULL; header = header->next)
+ {
+ /* Header enlisted in previous run? -> ignore */
+ if (header->str == NULL) continue;
+
+ if (0 == strncmpic(sorted_header->str, header->str, sorted_header_length)
+ && (header->str[sorted_header_length] == ':'))
+ {
+ log_error(LOG_LEVEL_HEADER, "Enlisting sorted header %s", header->str);
+ if (JB_ERR_OK != enlist(new_headers, header->str))
+ {
+ log_error(LOG_LEVEL_HEADER, "Failed to enlist %s", header->str);
+ }
+ freez(header->str);
+ }
+ }
+ }
+
+ /* Enlist the rest of the headers behind the ordered ones */
+ for (header = headers->first; header != NULL; header = header->next)
+ {
+ /* Header enlisted in previous run? -> ignore */
+ if (header->str == NULL) continue;
+
+ log_error(LOG_LEVEL_HEADER,
+ "Enlisting left-over header %s", header->str);
+ if (JB_ERR_OK != enlist(new_headers, header->str))
+ {
+ log_error(LOG_LEVEL_HEADER, "Failed to enlist %s", header->str);
+ }
+ freez(header->str);
+ }
+
+ list_remove_all(headers);
+ list_duplicate(headers, new_headers);
+ list_remove_all(new_headers);
+
+ 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++;
f++;
}
+ if (!filter_server_headers && !list_is_empty(csp->config->ordered_client_headers))
+ {
+ enforce_header_order(csp->headers, csp->config->ordered_client_headers);
+ }
+
return err;
}
+#ifdef FEATURE_HTTPS_INSPECTION
+/*********************************************************************
+ *
+ * Function : sed_https
+ *
+ * Description : add, delete or modify lines in the HTTPS client
+ * header streams. Wrapper around sed().
+ *
+ * Parameters :
+ * 1 : csp = Current client state (buffers, headers, etc...)
+ *
+ * Returns : JB_ERR_OK in case off success, or
+ * JB_ERR_MEMORY on some out-of-memory errors, or
+ * JB_ERR_PARSE in case of fatal parse errors.
+ *
+ *********************************************************************/
+jb_err sed_https(struct client_state *csp)
+{
+ jb_err err;
+ struct list headers;
+
+ /*
+ * Temporarly replace csp->headers with csp->https_headers
+ * to trick sed() into filtering the https headers.
+ */
+ headers.first = csp->headers->first;
+ headers.last = csp->headers->last;
+ csp->headers->first = csp->https_headers->first;
+ csp->headers->last = csp->https_headers->last;
+
+ /*
+ * Start with fresh tags. Already exising tags may
+ * be set again. This is necessary to overrule
+ * URL-based patterns.
+ */
+ destroy_list(csp->tags);
+
+ /*
+ * We want client header filters and taggers
+ * so temporarly remove the flag.
+ */
+ csp->flags &= ~CSP_FLAG_CLIENT_HEADER_PARSING_DONE;
+ err = sed(csp, FILTER_CLIENT_HEADERS);
+ csp->flags |= CSP_FLAG_CLIENT_HEADER_PARSING_DONE;
+
+ csp->headers->first = headers.first;
+ csp->headers->last = headers.last;
+
+ return err;
+}
+#endif /* def FEATURE_HTTPS_INSPECTION */
+
+
/*********************************************************************
*
* Function : update_server_headers
log_error(LOG_LEVEL_HEADER,
"Content modified with no Content-Length header set. "
"Created: %s.", header);
+ csp->flags |= CSP_FLAG_SERVER_CONTENT_LENGTH_SET;
}
}
#endif /* def FEATURE_CONNECTION_KEEP_ALIVE */
*********************************************************************/
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) joblist = compile_dynamic_pcrs_job_list(csp, b);
+ if (b->dynamic) pcrs_free_joblist(joblist);
- if (NULL == joblist)
+ /* 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 (!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;
+ joblist = b->joblist;
- if (b->dynamic) joblist = compile_dynamic_pcrs_job_list(csp, b);
+ 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;
- }
-
- log_error(LOG_LEVEL_RE_FILTER, "filtering \'%s\' (size %d) with \'%s\' ...",
- *header, size, b->name);
-
- /* 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;
}
/*
* to remove the header. This function frees the
* original string if necessary.
*
- * Returns : JB_ERR_OK on success, or
- * JB_ERR_MEMORY on out-of-memory error.
+ * Returns : JB_ERR_OK on success.
*
*********************************************************************/
static jb_err server_connection(struct client_state *csp, char **header)
#ifdef FEATURE_CONNECTION_KEEP_ALIVE
&& !(csp->flags & CSP_FLAG_SERVER_SOCKET_TAINTED)
#endif
- )
+ )
{
#ifdef FEATURE_CONNECTION_KEEP_ALIVE
if ((csp->config->feature_flags & RUNTIME_FEATURE_CONNECTION_KEEP_ALIVE))
{
char *old_header = *header;
- *header = strdup("Connection: close");
- if (header == NULL)
- {
- return JB_ERR_MEMORY;
- }
+ *header = strdup_or_die("Connection: close");
log_error(LOG_LEVEL_HEADER, "Replaced: \'%s\' with \'%s\'", old_header, *header);
freez(old_header);
}
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;
}
* Content-Length header.
*
* Parameters :
- * 1 : header = The Content-Length header.
+ * 1 : header_value = The Content-Length header value.
* 2 : length = Storage to return the value.
*
* Returns : JB_ERR_OK on success, or
* JB_ERR_PARSE if no value is recognized.
*
*********************************************************************/
-static jb_err get_content_length(const char *header, unsigned long long *length)
+static jb_err get_content_length(const char *header_value, unsigned long long *length)
{
- assert(header[14] == ':');
-
#ifdef _WIN32
- assert(sizeof(unsigned long long) > 4);
- if (1 != sscanf(header+14, ": %I64u", length))
+#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+14, ": %llu", length))
+ if (1 != sscanf(header_value, "%llu", length))
#endif
{
return JB_ERR_PARSE;
static jb_err client_save_content_length(struct client_state *csp, char **header)
{
unsigned long long content_length = 0;
+ const char *header_value;
assert(*(*header+14) == ':');
- if (JB_ERR_OK != get_content_length(*header, &content_length))
+ header_value = *header + 15;
+ if (JB_ERR_OK != get_content_length(header_value, &content_length))
{
log_error(LOG_LEVEL_ERROR, "Crunching invalid header: %s", *header);
freez(*header);
* to remove the header. This function frees the
* original string if necessary.
*
- * Returns : JB_ERR_OK on success, or
- * JB_ERR_MEMORY on out-of-memory error.
+ * Returns : JB_ERR_OK on success.
*
*********************************************************************/
static jb_err client_connection(struct client_state *csp, char **header)
if (!strcmpic(*header, connection_close))
{
#ifdef FEATURE_CONNECTION_KEEP_ALIVE
- if ((csp->config->feature_flags & RUNTIME_FEATURE_CONNECTION_SHARING))
+ if ((csp->config->feature_flags & RUNTIME_FEATURE_CONNECTION_SHARING)
+ && !(csp->flags & CSP_FLAG_SERVER_SOCKET_TAINTED))
{
if (!strcmpic(csp->http->ver, "HTTP/1.1"))
{
{
char *old_header = *header;
- *header = strdup("Connection: keep-alive");
- if (header == NULL)
- {
- return JB_ERR_MEMORY;
- }
+ *header = strdup_or_die("Connection: keep-alive");
log_error(LOG_LEVEL_HEADER,
"Replaced: \'%s\' with \'%s\'", old_header, *header);
freez(old_header);
csp->flags &= ~CSP_FLAG_CLIENT_CONNECTION_KEEP_ALIVE;
}
}
- else if ((csp->config->feature_flags & RUNTIME_FEATURE_CONNECTION_KEEP_ALIVE))
+ else if ((csp->config->feature_flags & RUNTIME_FEATURE_CONNECTION_KEEP_ALIVE)
+ && !(csp->flags & CSP_FLAG_SERVER_SOCKET_TAINTED))
{
log_error(LOG_LEVEL_HEADER,
"Keeping the client header '%s' around. "
{
char *old_header = *header;
- *header = strdup(connection_close);
- if (header == NULL)
- {
- return JB_ERR_MEMORY;
- }
+ *header = strdup_or_die(connection_close);
log_error(LOG_LEVEL_HEADER,
"Replaced: \'%s\' with \'%s\'", old_header, *header);
freez(old_header);
}
- /* Signal client_connection_adder() to return early. */
+ /* Signal client_connection_header_adder() to return early. */
csp->flags |= CSP_FLAG_CLIENT_CONNECTION_HEADER_SET;
return JB_ERR_OK;
}
+#ifdef FEATURE_CONNECTION_KEEP_ALIVE
+/*********************************************************************
+ *
+ * Function : client_proxy_connection
+ *
+ * Description : Sets the CLIENT_CONNECTION_KEEP_ALIVE flag when
+ * appropriate and removes the Proxy-Connection
+ * header.
+ *
+ * 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 client_proxy_connection(struct client_state *csp, char **header)
+{
+ if (0 == (csp->flags & CSP_FLAG_CLIENT_CONNECTION_KEEP_ALIVE)
+ && (csp->http->ssl == 0)
+ && (NULL == strstr(*header, "close")))
+ {
+ log_error(LOG_LEVEL_HEADER,
+ "The client connection can be kept alive due to: %s", *header);
+ csp->flags |= CSP_FLAG_CLIENT_CONNECTION_KEEP_ALIVE;
+ }
+ crumble(csp, header);
+
+ return JB_ERR_OK;
+}
+#endif /* def FEATURE_CONNECTION_KEEP_ALIVE */
+
+
+/*********************************************************************
+ *
+ * Function : client_transfer_encoding
+ *
+ * Description : Raise the CSP_FLAG_CHUNKED_CLIENT_BODY flag if
+ * the request body is "chunked"
+ *
+ * XXX: Currently not called through sed() as we
+ * need the flag earlier on. Should be fixed.
+ *
+ * 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_transfer_encoding(struct client_state *csp, char **header)
+{
+ if (strstr(*header, "chunked"))
+ {
+ csp->flags |= CSP_FLAG_CHUNKED_CLIENT_BODY;
+ log_error(LOG_LEVEL_HEADER, "Expecting chunked client body");
+ }
+
+ return JB_ERR_OK;
+}
+
+
+/*********************************************************************
+ *
+ * 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("Content-Type: ");
- string_append(header, csp->action->string[ACTION_STRING_CONTENT_TYPE]);
+ *header = strdup_or_die("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. "
static jb_err server_save_content_length(struct client_state *csp, char **header)
{
unsigned long long content_length = 0;
+ const char *header_value;
assert(*(*header+14) == ':');
- if (JB_ERR_OK != get_content_length(*header, &content_length))
+ header_value = *header + 15;
+ if (JB_ERR_OK != get_content_length(header_value, &content_length))
{
log_error(LOG_LEVEL_ERROR, "Crunching invalid header: %s", *header);
freez(*header);
newval = csp->action->string[ACTION_STRING_LAST_MODIFIED];
- if (0 == strcmpic(newval, "block") )
+ if (0 == strcmpic(newval, "block"))
{
/*
* Blocking Last-Modified header. Useless but why not.
}
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
newval = csp->action->string[ACTION_STRING_LANGUAGE];
- if ((newval == NULL) || (0 == strcmpic(newval, "block")) )
+ if ((newval == NULL) || (0 == strcmpic(newval, "block")))
{
/*
* Blocking Accept-Language header
/*
* Are we blocking the e-mail address?
*/
- if ((newval == NULL) || (0 == strcmpic(newval, "block")) )
+ if ((newval == NULL) || (0 == strcmpic(newval, "block")))
{
log_error(LOG_LEVEL_HEADER, "crunched From!");
return JB_ERR_OK;
*********************************************************************/
static jb_err client_send_cookie(struct client_state *csp, char **header)
{
- if (csp->action->flags & ACTION_NO_COOKIE_READ)
+ if (csp->action->flags & ACTION_CRUNCH_OUTGOING_COOKIES)
{
log_error(LOG_LEVEL_HEADER, "Crunched outgoing cookie: %s", *header);
freez(*header);
* 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;
}
*csp->http->hostport == ' ' || *csp->http->hostport == '\0')
{
- if (NULL == (p = strdup((*header)+6)))
- {
- return JB_ERR_MEMORY;
- }
+ p = strdup_or_die((*header)+6);
chomp(p);
- if (NULL == (q = strdup(p)))
- {
- freez(p);
- return JB_ERR_MEMORY;
- }
+ q = strdup_or_die(p);
freez(csp->http->hostport);
csp->http->hostport = p;
const char *newval;
char * endptr;
- if ( 0 == strcmpic(*header, "If-Modified-Since: Wed, 08 Jun 1955 12:00:00 GMT"))
+ if (0 == strcmpic(*header, "If-Modified-Since: Wed, 08 Jun 1955 12:00:00 GMT"))
{
/*
* The client got an error message because of a temporary problem,
}
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
*********************************************************************/
jb_err client_x_filter(struct client_state *csp, char **header)
{
- if ( 0 == strcmpic(*header, "X-Filter: No"))
+ if (0 == strcmpic(*header, "X-Filter: No"))
{
if (!(csp->config->feature_flags & RUNTIME_FEATURE_HTTP_TOGGLE))
{
* Function : client_range
*
* Description : Removes Range, Request-Range and If-Range headers if
- * content filtering is enabled. If the client's version
- * of the document has been altered by Privoxy, the server
- * could interpret the range differently than the client
- * intended in which case the user could end up with
- * corrupted content.
+ * content filtering is enabled and the range doesn't
+ * start at byte 0.
+ *
+ * If the client's version of the document has been
+ * altered by Privoxy, the server could interpret the
+ * range differently than the client intended in which
+ * case the user could end up with corrupted content.
+ *
+ * If the range starts at byte 0 this isn't an issue
+ * so the header can pass. Partial requests like this
+ * are used to render preview images for videos without
+ * downloading the whole video.
+ *
+ * While HTTP doesn't require that range requests are
+ * honoured and the client could simply abort the download
+ * after receiving a sufficient amount of data, various
+ * clients don't handle complete responses to range
+ * requests gracefully and emit misleading error messages
+ * instead.
*
* Parameters :
* 1 : csp = Current client state (buffers, headers, etc...)
*********************************************************************/
static jb_err client_range(struct client_state *csp, char **header)
{
- if (content_filters_enabled(csp->action))
+ if (content_filters_enabled(csp->action)
+ && (0 != strncmpic(strstr(*header, ":"), ": bytes=0-", 10)))
{
log_error(LOG_LEVEL_HEADER, "Content filtering is enabled."
" Crunching: \'%s\' to prevent range-mismatch problems.", *header);
return JB_ERR_OK;
}
- if ( !csp->http->hostport || !*(csp->http->hostport))
+ 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;
}
/*
* remove 'user:pass@' from 'proto://user:pass@host'
*/
- if ( (p = strchr( csp->http->hostport, '@')) != NULL )
+ if ((p = strchr( csp->http->hostport, '@')) != NULL)
{
p++;
}
#ifdef FEATURE_CONNECTION_KEEP_ALIVE
&& !(csp->flags & CSP_FLAG_SERVER_SOCKET_TAINTED)
#endif
- )
+ )
{
log_error(LOG_LEVEL_HEADER, "A HTTP/1.1 response "
"without Connection header implies keep-alive.");
* Function : server_proxy_connection_adder
*
* Description : Adds a "Proxy-Connection: keep-alive" header to
- * csp->headers if the client asked for keep-alive.
- * XXX: We should reuse existent ones.
+ * csp->headers when appropriate.
*
* Parameters :
* 1 : csp = Current client state (buffers, headers, etc...)
if ((csp->flags & CSP_FLAG_CLIENT_CONNECTION_KEEP_ALIVE)
&& !(csp->flags & CSP_FLAG_SERVER_SOCKET_TAINTED)
- && !(csp->flags & CSP_FLAG_SERVER_PROXY_CONNECTION_HEADER_SET))
+ && !(csp->flags & CSP_FLAG_SERVER_PROXY_CONNECTION_HEADER_SET)
+ && ((csp->flags & CSP_FLAG_SERVER_CONTENT_LENGTH_SET)
+ || (csp->flags & CSP_FLAG_CHUNKED)))
{
log_error(LOG_LEVEL_HEADER, "Adding: %s", proxy_connection_header);
err = enlist(csp->headers, proxy_connection_header);
* 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->http->ssl == 0)
+ && !(csp->flags & CSP_FLAG_SERVER_SOCKET_TAINTED)
&& !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;
}
+/*********************************************************************
+ *
+ * Function : add_cooky_expiry_date
+ *
+ * Description : Adds a cookie expiry date to a string.
+ *
+ * Parameters :
+ * 1 : cookie = On input, pointer to cookie to modify.
+ * On output, pointer to the modified header.
+ * The original string is freed.
+ * 2 : lifetime = Seconds the cookie should be valid
+ *
+ * Returns : N/A
+ *
+ *********************************************************************/
+static void add_cookie_expiry_date(char **cookie, time_t lifetime)
+{
+ char tmp[50];
+ struct tm *timeptr = NULL;
+ time_t expiry_date = time(NULL) + lifetime;
+#ifdef HAVE_GMTIME_R
+ struct tm gmt;
+
+ timeptr = gmtime_r(&expiry_date, &gmt);
+#elif defined(MUTEX_LOCKS_AVAILABLE)
+ privoxy_mutex_lock(&gmtime_mutex);
+ timeptr = gmtime(&expiry_date);
+ privoxy_mutex_unlock(&gmtime_mutex);
+#else
+ timeptr = gmtime(&expiry_date);
+#endif
+
+ if (NULL == timeptr)
+ {
+ log_error(LOG_LEVEL_FATAL,
+ "Failed to get the time in add_cooky_expiry_date()");
+ }
+ strftime(tmp, sizeof(tmp), "; expires=%a, %d-%b-%Y %H:%M:%S GMT", timeptr);
+ if (JB_ERR_OK != string_append(cookie, tmp))
+ {
+ log_error(LOG_LEVEL_FATAL, "Out of memory in add_cooky_expiry()");
+ }
+}
+
/*********************************************************************
*
* Crunch, accept or rewrite it to a session cookie.
* Called from `sed'.
*
- * TODO: Allow the user to specify a new expiration
- * time to cause the cookie to expire even before the
- * browser is closed.
- *
* Parameters :
* 1 : csp = Current client state (buffers, headers, etc...)
* 2 : header = On input, pointer to header to modify.
*********************************************************************/
static jb_err server_set_cookie(struct client_state *csp, char **header)
{
- time_t now;
- time_t cookie_time;
-
- time(&now);
-
- if ((csp->action->flags & ACTION_NO_COOKIE_SET) != 0)
+ if ((csp->action->flags & ACTION_CRUNCH_INCOMING_COOKIES) != 0)
{
log_error(LOG_LEVEL_HEADER, "Crunching incoming cookie: %s", *header);
freez(*header);
}
- else if ((csp->action->flags & ACTION_NO_COOKIE_KEEP) != 0)
+ else if ((0 != (csp->action->flags & ACTION_SESSION_COOKIES_ONLY))
+ || (0 != (csp->action->flags & ACTION_LIMIT_COOKIE_LIFETIME)))
{
- /* Flag whether or not to log a message */
- int changed = 0;
+ time_t now;
+ time_t cookie_time;
+ long cookie_lifetime = 0;
+ enum
+ {
+ NO_EXPIRY_DATE_SPECIFIED,
+ EXPIRY_DATE_ACCEPTABLE,
+ EXPIRY_DATE_UNACCEPTABLE
+ } expiry_date_status = NO_EXPIRY_DATE_SPECIFIED;
/* A variable to store the tag we're working on */
char *cur_tag;
cur_tag = *header + 11;
/* skip whitespace between "Set-Cookie:" and value */
- while (*cur_tag && ijb_isspace(*cur_tag))
+ while (*cur_tag && privoxy_isspace(*cur_tag))
{
cur_tag++;
}
+ time(&now);
+
+ if ((csp->action->flags & ACTION_LIMIT_COOKIE_LIFETIME) != 0)
+ {
+ const char *param = csp->action->string[ACTION_STRING_LIMIT_COOKIE_LIFETIME];
+
+ cookie_lifetime = strtol(param, NULL, 0);
+ if (cookie_lifetime < 0)
+ {
+ log_error(LOG_LEVEL_FATAL, "Invalid cookie lifetime limit: %s", param);
+ }
+ cookie_lifetime *= 60;
+ }
+
/* Loop through each tag in the cookie */
while (*cur_tag)
{
next_tag++;
/* skip whitespace ";" and start of tag */
- while (*next_tag && ijb_isspace(*next_tag))
+ while (*next_tag && privoxy_isspace(*next_tag))
{
next_tag++;
}
log_error(LOG_LEVEL_ERROR,
"Can't parse \'%s\', send by %s. Unsupported time format?", cur_tag, csp->http->url);
string_move(cur_tag, next_tag);
- changed = 1;
+ expiry_date_status = EXPIRY_DATE_UNACCEPTABLE;
}
else
{
* anyway, which in many cases will be shorter
* than a browser session.
*/
- if (cookie_time - now < 0)
+ if (cookie_time < now)
{
log_error(LOG_LEVEL_HEADER,
"Cookie \'%s\' is already expired and can pass unmodified.", *header);
/* Just in case some clown sets more then one expiration date */
cur_tag = next_tag;
+ expiry_date_status = EXPIRY_DATE_ACCEPTABLE;
+ }
+ else if ((cookie_lifetime != 0) && (cookie_time < (now + cookie_lifetime)))
+ {
+ log_error(LOG_LEVEL_HEADER, "Cookie \'%s\' can pass unmodified. "
+ "Its lifetime is below the limit.", *header);
+ /* Just in case some clown sets more then one expiration date */
+ cur_tag = next_tag;
+ expiry_date_status = EXPIRY_DATE_ACCEPTABLE;
}
else
{
string_move(cur_tag, next_tag);
/* That changed the header, need to issue a log message */
- changed = 1;
+ expiry_date_status = EXPIRY_DATE_UNACCEPTABLE;
/*
* Note that the next tag has now been moved to *cur_tag,
}
}
- if (changed)
+ if (expiry_date_status != EXPIRY_DATE_ACCEPTABLE)
{
assert(NULL != *header);
- log_error(LOG_LEVEL_HEADER, "Cookie rewritten to a temporary one: %s",
- *header);
+ if (cookie_lifetime != 0)
+ {
+ add_cookie_expiry_date(header, cookie_lifetime);
+ log_error(LOG_LEVEL_HEADER, "Cookie rewritten to: %s", *header);
+ }
+ else if (expiry_date_status != NO_EXPIRY_DATE_SPECIFIED)
+ {
+ log_error(LOG_LEVEL_HEADER,
+ "Cookie rewritten to a temporary one: %s", *header);
+ }
}
}
* attempt to work around GNU libc's strptime()
* reporting negative year values when using %Y.
*/
- static const char * const time_formats[] = {
+ static const char time_formats[][22] = {
/* Tue, 02-Jun-37 20:00:00 */
"%a, %d-%b-%y %H:%M:%S",
/* Tue, 02 Jun 2037 20:00:00 */
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);
+
+}
+
/*********************************************************************
*
* the "Host:" header)
* 2 : http = storage for the result (host, port and hostport).
*
- * Returns : JB_ERR_MEMORY in case of memory problems,
+ * Returns : JB_ERR_MEMORY (or terminates) in case of memory problems,
* JB_ERR_PARSE if the host header couldn't be found,
* JB_ERR_OK otherwise.
*
return JB_ERR_PARSE;
}
- p = strdup(host);
- if (NULL == p)
- {
- log_error(LOG_LEVEL_ERROR, "Out of memory while parsing \"Host:\" header");
- return JB_ERR_MEMORY;
- }
+ p = strdup_or_die(host);
chomp(p);
- if (NULL == (q = strdup(p)))
- {
- freez(p);
- log_error(LOG_LEVEL_ERROR, "Out of memory while parsing \"Host:\" header");
- return JB_ERR_MEMORY;
- }
+ q = strdup_or_die(p);
freez(http->hostport);
http->hostport = p;
}
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;
+
+}
+
+
+#ifdef FEATURE_HTTPS_INSPECTION
+/*********************************************************************
+ *
+ * Function : get_destination_from_https_headers
+ *
+ * Description : Parse the previously encrypted "Host:" header to
+ * get the request's destination.
+ *
+ * Parameters :
+ * 1 : headers = List of headers (one of them hopefully being
+ * the "Host:" header)
+ * 2 : http = storage for the result (host, port and hostport).
+ *
+ * Returns : JB_ERR_MEMORY (or terminates) in case of memory problems,
+ * JB_ERR_PARSE if the host header couldn't be found,
+ * JB_ERR_OK otherwise.
+ *
+ *********************************************************************/
+jb_err get_destination_from_https_headers(const struct list *headers, struct http_request *http)
+{
+ char *q;
+ char *p;
+ char *host;
+
+ host = get_header_value(headers, "Host:");
+
+ if (NULL == host)
+ {
+ log_error(LOG_LEVEL_ERROR, "No \"Host:\" header found.");
+ return JB_ERR_PARSE;
+ }
+
+ p = strdup_or_die(host);
+ chomp(p);
+ q = strdup_or_die(p);
+
+ freez(http->hostport);
+ http->hostport = p;
+ freez(http->host);
+ http->host = q;
+ q = strchr(http->host, ':');
+ if (q != NULL)
+ {
+ /* Terminate hostname and evaluate port string */
+ *q++ = '\0';
+ http->port = atoi(q);
+ }
+ else
+ {
+ http->port = 443;
+ }
+
+ /* Rebuild request URL */
+ freez(http->url);
+ http->url = strdup_or_die(http->path);
+
+ 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;
}
+#endif /* def FEATURE_HTTPS_INSPECTION */
/*********************************************************************
static jb_err handle_conditional_hide_referrer_parameter(char **header,
const char *host, const int parameter_conditional_block)
{
- char *referer = strdup(*header);
+ char *referer = strdup_or_die(*header);
const size_t hostlength = strlen(host);
const char *referer_url = NULL;
- if (NULL == referer)
- {
- freez(*header);
- return JB_ERR_MEMORY;
- }
-
/* referer begins with 'Referer: http[s]://' */
if ((hostlength+17) < strlen(referer))
{
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
}
+#ifdef FEATURE_CONNECTION_KEEP_ALIVE
+/*********************************************************************
+ *
+ * Function : get_expected_content_length
+ *
+ * Description : Figures out the content length from a list of headers.
+ *
+ * Parameters :
+ * 1 : headers = List of headers
+ *
+ * Returns : Number of bytes to expect
+ *
+ *********************************************************************/
+unsigned long long get_expected_content_length(struct list *headers)
+{
+ const char *content_length_header;
+ unsigned long long content_length = 0;
+
+ content_length_header = get_header_value(headers, "Content-Length:");
+ if (content_length_header != NULL)
+ {
+ if (JB_ERR_OK != get_content_length(content_length_header, &content_length))
+ {
+ log_error(LOG_LEVEL_ERROR,
+ "Failed to get the Content-Length in %s", content_length_header);
+ /* XXX: The header will be removed later on */
+ return 0;
+ }
+ }
+
+ return content_length;
+}
+#endif
+
/*
Local Variables:
tab-width: 3