PACKET_MMAP data offset
I am writing the code to send an UDP packet using PACKET_TX_RING ring, and I don't understand the example code described here: http://wiki.ipxwarzone.com/index.php5?title=Linux_packet_mmap#Kernel_Patch
The /usr/src/linux/Documentation/networking/packet_mmap.txt says that the frame structure is this:
/* Frame structure:
- Start. Frame must be aligned to TPACKET_ALIGNMENT=16
- struct tpacket_hdr
- pad to TPACKET_ALIGNMENT=16
- struct sockaddr_ll
- Ga开发者_开发百科p, chosen so that packet data (Start+tp_net) aligns to TPACKET_ALIGNMENT=16
- Start+tp_mac: [ Optional MAC header ]
- Start+tp_net: Packet data, aligned to TPACKET_ALIGNMENT=16.
Pad to align to TPACKET_ALIGNMENT=16
*/
If data starts at Start + tp_net, then why the code in the example substracts sizeof(struct sockaddr_ll) instead of summing it, this is the line:
/* get data offset */
data_offset = TPACKET_HDRLEN - sizeof(struct sockaddr_ll);
printf("data offset = %d bytes\n", data_offset);
After calculating the pointer to packet data, the data is copied:
// fill data
off = ((void *) header) + (TPACKET_HDRLEN - sizeof(struct sockaddr_ll));
memcpy(off, pkt, pktlen);
Tis looks like data corruption to me, the socket address and optional mac address will be overwriten by the data to transmit. In my case, if I use this code, the UDP packet header will be overwritten, is this correct?
Thanks in advance
Sorry I'm a little late. I have had the very same problem and unfortunately using PACKET_TX_RING is not very well documented. But fortunately it's still quite easy to figure out using the linked example program.
First you will need tshark(or wireshark) and the packet_mm source from the link you've posted. The example source fill the buffer with 0 to 150 and then sends it directly of to the given device. Using tshark we will read out what is send.
Run tshark in one shell (tapping the loopback device):
$ tshark -V -i lo
and run packet_mm in another shell:
$ packet_mm lo
Check out one frame:
Frame 1 (150 bytes on wire, 150 bytes captured)
Arrival Time: Nov 12, 2011 13:07:02.636424000
[Time delta from previous captured frame: 0.000005000 seconds] [Time delta from previous displayed frame: 0.000005000 seconds]
[Time since reference or first frame: 337.280499000 seconds]
Frame Number: 1001
Frame Length: 150 bytes
Capture Length: 150 bytes
[Frame is marked: False]
[Protocols in frame: eth:data]
Ethernet II, Src: 06:07:08:09:0a:0b (06:07:08:09:0a:0b), Dst: 3com_03:04:05 (00:01:02:03:04:05)
Destination: 3com_03:04:05 (00:01:02:03:04:05)
Address: 3com_03:04:05 (00:01:02:03:04:05)
.... ...0 .... .... .... .... = IG bit: Individual address (unicast)
.... ..0. .... .... .... .... = LG bit: Globally unique address (factory default)
Source: 06:07:08:09:0a:0b (06:07:08:09:0a:0b)
Address: 06:07:08:09:0a:0b (06:07:08:09:0a:0b)
.... ...0 .... .... .... .... = IG bit: Individual address (unicast)
.... ..1. .... .... .... .... = LG bit: Locally administered address (this is NOT the factory default)
Type: Unknown (0x0c0d)
Data (136 bytes)
0000 0e 0f 10 11 12 13 14 15 16 17 18 19 1a 1b 1c 1d ................
0010 1e 1f 20 21 22 23 24 25 26 27 28 29 2a 2b 2c 2d .. !"#$%&'()*+,-
0020 2e 2f 30 31 32 33 34 35 36 37 38 39 3a 3b 3c 3d ./0123456789:;<=
0030 3e 3f 40 41 42 43 44 45 46 47 48 49 4a 4b 4c 4d >?@ABCDEFGHIJKLM
0040 4e 4f 50 51 52 53 54 55 56 57 58 59 5a 5b 5c 5d NOPQRSTUVWXYZ[\]
0050 5e 5f 60 61 62 63 64 65 66 67 68 69 6a 6b 6c 6d ^_`abcdefghijklm
0060 6e 6f 70 71 72 73 74 75 76 77 78 79 7a 7b 7c 7d nopqrstuvwxyz{|}
0070 7e 7f 80 81 82 83 84 85 86 87 88 89 8a 8b 8c 8d ~...............
0080 8e 8f 90 91 92 93 94 95 ........
Data: 0E0F101112131415161718191A1B1C1D1E1F202122232425...
[Length: 136]
The destination MAC is 00:01:02:03:04:05 and source MAC 06:07:08:09:0a:0b. And data start (right after Ethernet header) with 0x0e (14).
So the example program's data offset is the start of the (ethernet) packet to send.
So this is indeed the correct way to fill the buffer for sending (make sure pktlen is not larger than your frame size):
// fill data
off = ((uint8_t *) header) + (TPACKET_HDRLEN - sizeof(struct sockaddr_ll));
memcpy(off, pkt, pktlen);
And no data will be overwritten. But you have to provide ethernet, IP and UDP header yourself.
EDIT: For the TX_Ring, the frame struct seems to be:
- Start. Frame must be aligned to TPACKET_ALIGNMENT=16
- struct tpacket_hdr (size = 32 bytes)
- pad to TPACKET_ALIGNMENT=16 (so effectively padding size is 0)
- packet data
- Pad to align to TPACKET_ALIGNMENT=16 (so next frame is 16Bit aligned)
精彩评论