Current tftp/tftpd implementation violates RFC2347 ("TFTP Option
Extension") on write request. The first DATA packet should have block
number 1 but tftp send DATA packet with block number 0 and tftpd
expects this wrong behavior.
>From RFC2347:
client server
-------------------------------------------------------
|2|barfile|0|octet|0|blksize|0|2048|0| --> WRQ
<-- |6|blksize|0|2048|0| OACK
|3|1| 2048 octets of data | --> DATA
<-- |4|1| ACK
Current behavior:
client server
-------------------------------------------------------
|2|barfile|0|octet|0|blksize|0|2048|0| --> WRQ
<-- |6|blksize|0|2048|0| OACK
|3|0| 2048 octets of data | --> DATA
<-- |4|0| ACK
The first part of this patch is for tftpd and the second part is for tftp.
Note that this fix breaks compatibility with old version.
Signed-off-by: Atsushi Nemoto <[email protected]>
---
networking/tftp.c | 6 +++++-
1 files changed, 5 insertions(+), 1 deletions(-)
diff --git a/networking/tftp.c b/networking/tftp.c
index 6cd3f69..f6a9813 100644
--- a/networking/tftp.c
+++ b/networking/tftp.c
@@ -262,6 +262,9 @@ static int tftp_protocol(
* we expect 1st ACK from peer to be for (block_nr-1),
* that is, for "block 0" which is our OACK pkt */
opcode = TFTP_OACK;
+ /* for upload case, we expect "block 1" after OACK */
+ if (CMD_GET(option_mask32))
+ block_nr = 1;
goto add_blksize_opt;
}
#endif
@@ -472,7 +475,8 @@ static int tftp_protocol(
}
io_bufsize = blksize + 4;
/* Send ACK for OACK ("block" no: 0) */
- block_nr = 0;
+ if (CMD_GET(option_mask32))
+ block_nr = 0;
continue;
}
/* rfc2347:
--
1.5.6.5
_______________________________________________
busybox mailing list
[email protected]
http://lists.busybox.net/mailman/listinfo/busybox
