October 2012 - rtmpdump - lists.mplayerhq.hu

[PATCH] Better URL decoding support
by Steven Penny 25 Oct '12

25 Oct '12

Currently RtmpDump has limited URL decoding support. If you use Bastien’s patch as shown here http://lists.mplayerhq.hu/pipermail/rtmpdump/2012-September/002093.html You can pass encoded characters like this rtmpdump -o a.flv -i "rtmp://server.com/app playpath=foobar\0a" Where "\0a" will be decoded as a newline. Also the Playpath can be decoded using an unpatched RtmpDump, as long as it is part of the "-r" parameter and properly parsed. Example rtmpdump -o a.flv -r rtmp://server.com/app/foobar%0a Where "%0a" will be decoded as a newline. This presents two problems. One, many times RtmpDump is unable to correctly parse the Playpath, mixing it with the "app". Two, if you explicitly state the Playpath with "-y", you lose the URL decoding support. I have simply moved the URL decoding function that Howard wrote to parseurl.c and declared it in rtmp.h, so that it can be used across the project. --- librtmp/parseurl.c | 73 +++++++++++++++++++++++++++++++++++++--------------- librtmp/rtmp.h | 1 + rtmpdump.c | 1 + rtmpgw.c | 47 --------------------------------- 4 files changed, 54 insertions(+), 68 deletions(-) diff --git a/librtmp/parseurl.c b/librtmp/parseurl.c index 57abfe7..07fbbb9 100644 --- a/librtmp/parseurl.c +++ b/librtmp/parseurl.c @@ -206,7 +206,7 @@ void RTMP_ParsePlaypath(AVal *in, AVal *out) { const char *playpath = in->av_val; const char *temp, *q, *ext = NULL; const char *ppstart = playpath; - char *streamname, *destptr, *p; + char *streamname, *destptr; int pplen = in->av_len; @@ -265,26 +265,57 @@ void RTMP_ParsePlaypath(AVal *in, AVal *out) { } } - for (p=(char *)ppstart; pplen >0;) { - /* skip extension */ - if (subExt && p == ext) { - p += 4; - pplen -= 4; - continue; - } - if (*p == '%') { - unsigned int c; - sscanf(p+1, "%02x", &c); - *destptr++ = c; - pplen -= 3; - p += 3; - } else { - *destptr++ = *p++; - pplen--; - } - } - *destptr = '\0'; + memcpy(destptr, ppstart, pplen * sizeof(char)); + destptr[pplen] = '\0'; + http_unescape(streamname); out->av_val = streamname; - out->av_len = destptr - streamname; + out->av_len = strlen(streamname); +} + +/* inplace http unescape. This is possible .. strlen(unescaped_string) <= strlen(esacped_string) */ +void +http_unescape(char *data) +{ + char hex[3]; + char *stp; + int src_x = 0; + int dst_x = 0; + + int length = (int) strlen(data); + hex[2] = 0; + + while (src_x < length) + { + if (strncmp(data + src_x, "%", 1) == 0 && src_x + 2 < length) + { + // + // Since we encountered a '%' we know this is an escaped character + // + hex[0] = data[src_x + 1]; + hex[1] = data[src_x + 2]; + data[dst_x] = (char) strtol(hex, &stp, 16); + dst_x += 1; + src_x += 3; + } + else if (src_x != dst_x) + { + // + // This doesn't need to be unescaped. If we didn't unescape anything previously + // there is no need to copy the string either + // + data[dst_x] = data[src_x]; + src_x += 1; + dst_x += 1; + } + else + { + // + // This doesn't need to be unescaped, however we need to copy the string + // + src_x += 1; + dst_x += 1; + } + } + data[dst_x] = '\0'; } diff --git a/librtmp/rtmp.h b/librtmp/rtmp.h index bd85a17..242a536 100644 --- a/librtmp/rtmp.h +++ b/librtmp/rtmp.h @@ -285,6 +285,7 @@ extern "C" unsigned int *port, AVal *playpath, AVal *app); void RTMP_ParsePlaypath(AVal *in, AVal *out); + void http_unescape(char *data); void RTMP_SetBufferMS(RTMP *r, int size); void RTMP_UpdateBufferMS(RTMP *r); diff --git a/rtmpdump.c b/rtmpdump.c index e52f7d4..a6cbdfd 100644 --- a/rtmpdump.c +++ b/rtmpdump.c @@ -963,6 +963,7 @@ main(int argc, char **argv) } break; case 'y': + http_unescape(optarg); STR2AVAL(playpath, optarg); break; case 'Y': diff --git a/rtmpgw.c b/rtmpgw.c index 0cf56bb..50b8c2a 100644 --- a/rtmpgw.c +++ b/rtmpgw.c @@ -220,53 +220,6 @@ FILE *netstackdump = NULL; FILE *netstackdump_read = NULL; #endif -/* inplace http unescape. This is possible .. strlen(unescaped_string) <= strlen(esacped_string) */ -void -http_unescape(char *data) -{ - char hex[3]; - char *stp; - int src_x = 0; - int dst_x = 0; - - int length = (int) strlen(data); - hex[2] = 0; - - while (src_x < length) - { - if (strncmp(data + src_x, "%", 1) == 0 && src_x + 2 < length) - { - // - // Since we encountered a '%' we know this is an escaped character - // - hex[0] = data[src_x + 1]; - hex[1] = data[src_x + 2]; - data[dst_x] = (char) strtol(hex, &stp, 16); - dst_x += 1; - src_x += 3; - } - else if (src_x != dst_x) - { - // - // This doesn't need to be unescaped. If we didn't unescape anything previously - // there is no need to copy the string either - // - data[dst_x] = data[src_x]; - src_x += 1; - dst_x += 1; - } - else - { - // - // This doesn't need to be unescaped, however we need to copy the string - // - src_x += 1; - dst_x += 1; - } - } - data[dst_x] = '\0'; -} - TFTYPE controlServerThread(void *unused) { -- 1.7.9

5 16

[PATCH v2] Better URL decoding support
by Steven Penny 25 Oct '12

25 Oct '12

Currently RtmpDump has limited URL decoding support. If you use Bastien’s patch as shown here http://lists.mplayerhq.hu/pipermail/rtmpdump/2012-September/002093.html You can pass encoded characters like this rtmpdump -o a.flv -i "rtmp://server.com/app playpath=foobar\0a" Where "\0a" will be decoded as a newline. Also the Playpath can be decoded using an unpatched RtmpDump, as long as it is part of the "-r" parameter and properly parsed. Example rtmpdump -o a.flv -r rtmp://server.com/app/foobar%0a Where "%0a" will be decoded as a newline. This presents two problems. One, many times RtmpDump is unable to correctly parse the Playpath, mixing it with the "app". Two, if you explicitly state the Playpath with "-y", you lose the URL decoding support. I have simply moved the URL decoding function that Howard wrote to parseurl.c and declared it in rtmp.h, so that it can be used across the project. This patch also address concerns here http://lists.mplayerhq.hu/pipermail/rtmpdump/2012-October/002112.html --- librtmp/parseurl.c | 62 ++++++++++++++++++++++++++++++++++++++++++--------- librtmp/rtmp.h | 1 + rtmpdump.c | 1 + rtmpgw.c | 47 --------------------------------------- 4 files changed, 53 insertions(+), 58 deletions(-) diff --git a/librtmp/parseurl.c b/librtmp/parseurl.c index 57abfe7..aef1f3a 100644 --- a/librtmp/parseurl.c +++ b/librtmp/parseurl.c @@ -272,19 +272,59 @@ void RTMP_ParsePlaypath(AVal *in, AVal *out) { pplen -= 4; continue; } - if (*p == '%') { - unsigned int c; - sscanf(p+1, "%02x", &c); - *destptr++ = c; - pplen -= 3; - p += 3; - } else { - *destptr++ = *p++; - pplen--; - } + *destptr++ = *p++; + pplen--; } *destptr = '\0'; + http_unescape(streamname); out->av_val = streamname; - out->av_len = destptr - streamname; + out->av_len = strlen(streamname); +} + +/* inplace http unescape. This is possible .. strlen(unescaped_string) <= strlen(esacped_string) */ +void +http_unescape(char *data) +{ + char hex[3]; + char *stp; + int src_x = 0; + int dst_x = 0; + + int length = (int) strlen(data); + hex[2] = 0; + + while (src_x < length) + { + if (strncmp(data + src_x, "%", 1) == 0 && src_x + 2 < length) + { + // + // Since we encountered a '%' we know this is an escaped character + // + hex[0] = data[src_x + 1]; + hex[1] = data[src_x + 2]; + data[dst_x] = (char) strtol(hex, &stp, 16); + dst_x += 1; + src_x += 3; + } + else if (src_x != dst_x) + { + // + // This doesn't need to be unescaped. If we didn't unescape anything previously + // there is no need to copy the string either + // + data[dst_x] = data[src_x]; + src_x += 1; + dst_x += 1; + } + else + { + // + // This doesn't need to be unescaped, however we need to copy the string + // + src_x += 1; + dst_x += 1; + } + } + data[dst_x] = '\0'; } diff --git a/librtmp/rtmp.h b/librtmp/rtmp.h index bd85a17..242a536 100644 --- a/librtmp/rtmp.h +++ b/librtmp/rtmp.h @@ -285,6 +285,7 @@ extern "C" unsigned int *port, AVal *playpath, AVal *app); void RTMP_ParsePlaypath(AVal *in, AVal *out); + void http_unescape(char *data); void RTMP_SetBufferMS(RTMP *r, int size); void RTMP_UpdateBufferMS(RTMP *r); diff --git a/rtmpdump.c b/rtmpdump.c index e52f7d4..a6cbdfd 100644 --- a/rtmpdump.c +++ b/rtmpdump.c @@ -963,6 +963,7 @@ main(int argc, char **argv) } break; case 'y': + http_unescape(optarg); STR2AVAL(playpath, optarg); break; case 'Y': diff --git a/rtmpgw.c b/rtmpgw.c index 0cf56bb..50b8c2a 100644 --- a/rtmpgw.c +++ b/rtmpgw.c @@ -220,53 +220,6 @@ FILE *netstackdump = NULL; FILE *netstackdump_read = NULL; #endif -/* inplace http unescape. This is possible .. strlen(unescaped_string) <= strlen(esacped_string) */ -void -http_unescape(char *data) -{ - char hex[3]; - char *stp; - int src_x = 0; - int dst_x = 0; - - int length = (int) strlen(data); - hex[2] = 0; - - while (src_x < length) - { - if (strncmp(data + src_x, "%", 1) == 0 && src_x + 2 < length) - { - // - // Since we encountered a '%' we know this is an escaped character - // - hex[0] = data[src_x + 1]; - hex[1] = data[src_x + 2]; - data[dst_x] = (char) strtol(hex, &stp, 16); - dst_x += 1; - src_x += 3; - } - else if (src_x != dst_x) - { - // - // This doesn't need to be unescaped. If we didn't unescape anything previously - // there is no need to copy the string either - // - data[dst_x] = data[src_x]; - src_x += 1; - dst_x += 1; - } - else - { - // - // This doesn't need to be unescaped, however we need to copy the string - // - src_x += 1; - dst_x += 1; - } - } - data[dst_x] = '\0'; -} - TFTYPE controlServerThread(void *unused) { -- 1.7.9

1 0

Escape encoding
by Steven Penny 20 Oct '12

20 Oct '12

This site http://kanalia.eu/player.php?id=91 provides a challenge for me. The parsed playpath end with a Carriage Return, Line Feed. This was discussed on the forums. http://stream-recorder.com/forum/help-getting-greek-channel-work-rtmpdump-t… A few solutions were offered up, including - remove CRLF and prepend "mp4:" to playpath - remove CRLF and append ".mp4" to playpath - pass a literal Carriage return using Bash and $'\r' KSV has noted that these solutions are nonstandard and might not be future proof. I noticed in the MAN this paragraph > Special characters in values may need to be escaped to prevent > misinterpretation by the option parser. The escape encoding uses a backslash > followed by two hexadecimal digits representing the ASCII value of the > character. E.g., spaces must be escaped as \20 and backslashes must be escaped > as \5c. This looks promising, but after trying to append "\0d\0a" or just "\0d", it appears "not to work".

2 2

PolarSSL patch.
by bat guano 18 Oct '12

18 Oct '12

Hi Has a patch been submitted to allow RTMPDump to be compiled with PolarSSL-1.1.4 ?

3 5

Feature request
by Steven Spartz 15 Oct '12

15 Oct '12

I was wondering if there is a way to have rtmpdump auto reconnect a live stream? In the windows world Adobe Flash Media Encoder will do this automatically. Is there something similar in rtmpdump? Thank you

1 0

RTMPE doc
by triton bamb 08 Oct '12

08 Oct '12

I do not find any document about the RTMPE protocol, can someone help me to understand how to decrypt the connect packet containing the main infos of the media: swfurl, url...? Best, Triton

2 4

ipfw for Windows 7 64-bit/32-bit
by Steven Penny 04 Oct '12

04 Oct '12

>From the README > (Based on reading the BSD ipfw manpage, this rule ought to work on BSD: > ipfw add 40 fwd 127.0.0.1,1935 tcp from any to any 1935 not uid proxy > Some confirmation from any BSD users would be nice.) > (We have a solution for Windows based on a TDI driver; this is known to > work on Win2K and WinXP but is assumed to not work on Vista or Win7 as the > TDI is no longer used on those OS versions. Also, none of the known > solutions are available as freeware.) I took a crack at ipfw for Windows a while back but I stopped because the one I found only had a 32-bit build http://sourceforge.net/projects/wipfw/files/experimental It turns out theres is another open source implementation of ipfw here http://info.iet.unipi.it/~luigi/dummynet The site notes both 64-bit and 32-bit support for Windows 7. I successfully installed it for Windows 7 64-bit, detailed here http://stackoverflow.com/a/12719202 However the next challenge is running 2 users. Apparently Cygwin "su" has never worked http://cygwin.com/faq-nochunks.html#faq.using.su and the suggested workaround is to install "sshd". I gave up at this point, but I thought I would share my findings. Currently I am using "RtmpDumpHelper" to capture streams but with "ipfw" we could have an open source solution.

1 0