skb functions

[GoogleCodeExporter]

i've made a change in code.

i'm trying to use the skbbuf functions from linux

please check iwi2200.h. if whe can port the linux functions it will help

the mbuf_len is the siize of mbuf data

in linux the skb->len is the size of the entire skb

there's also the skb->data_len wich is the same as mbuf->len

the firmware error will go away if the skb functions work correctly

do you have any idea on how to bring the en1 interface up? this is a problem

Original issue reported on code.google.com by jalav...@gmail.com on 27 Jan 2007 at 5:57

Attachments:

• iwi2200_source.zip

[GoogleCodeExporter]

jalavoui. 

if anonymous, you can checkout from svn 
I must merge by hand at every time.
and plz attached patch file (not full source ) . sorry.

I am checking your new sources.
i will say about other issue (sk_buff,skb function, interface up issue ) after 
checking.

Original comment by kazuh...@gmail.com on 27 Jan 2007 at 6:16

• Changed state: Accepted

[GoogleCodeExporter]

i re-send invitation mail of gmail . plz retry.

Original comment by kazuh...@gmail.com on 27 Jan 2007 at 6:38

[GoogleCodeExporter]

jalavoui.

I think that mbuf_datastart -> mbuf_data will work correctly . 

reading wrapper of skb_function.
I think that these is danger.

> the mbuf_len is the siize of mbuf data
> in linux the skb->len is the size of the entire skb

yes . but

> there's also the skb->data_len wich is the same as mbuf->len

mbuf->len is mbuf->m_len?
if so, this is wrong.

in linux system
  1 packet is 1 skb .
in bsd system
  1 packet is over 1 mbuf .

so mbuf_len() dont return entire data  size of packet. ( only 1 mbuf).

generaly, 
packet header( this is mbuf of packet header) is stored
     in first mbuf of chains( mbufs over 1 and it appear 1 packet).
so you seem that codes is no problem. becasue 
  only headers(IP/TCP/80211...) is  edited in current code.

however if data exclude its headers(IP... s ) is store next mbuf( 
mbuf_next(mbuf) ),
the data( part of mbuf_next(mbuf) )dont send.

about rx .

I seem that allocated packet(via allocatePacket) data always  have MBUF_PKTHDR 
and
MBUF_EXT.

my knowledge may be  wrong. 
jalavoui, plz replay about this.

Original comment by kazuh...@gmail.com on 27 Jan 2007 at 7:45

[GoogleCodeExporter]

did you see the comments in skbbuf.h from linux?. the comments help to 
understand
what the skb functions should do

we sould test the ported skb functions very well. i think that they still have 
bugs.

> so mbuf_len() dont return entire data  size of packet. ( only 1 mbuf).

are you shure? it looks like it returns the full size of mbuf_data

i'll check the code and try to fix the bugs

Original comment by jalav...@gmail.com on 29 Jan 2007 at 7:20

[GoogleCodeExporter]

> did you see the comments in skbbuf.h from linux?. the comments help to 
understand
> what the skb functions should do

I know good better than bsd system ( like a bsd )
because my major is linux system.

skb  system is better than mbuf . 
128 kb  is limit of normal mbuf ( without MBUF_EXT ) .....

>> so mbuf_len() dont return entire data  size of packet. ( only 1 mbuf).
> are you shure? it looks like it returns the full size of mbuf_data

yes . if the entire mbuf  is 2 mbuf , 
  size of total data is m_pkthdr.len ( can get mbuf_pkthdr_len ).

if following supposed,
  first mbuf1.
  second mbuf2.
  mbuf_len(mbuf1) == 30;
  mbuf_len(mbuf2) == 430;  
  mbuf_next(mbuf1) == mbuf2;
  mbuf_next(mbuf2) == NULL;

so 
  mbuf_len(mbuf1) == 30
  mbuf_pkthdr_len(mbuf1)  == 430.

if you have TCP/IP Illustrated volume 2. plz see Chapter2 2.1 .

finally , about skb_functions in current code.
I checked and chenge pkthdr_len and commited.

Original comment by kazuh...@gmail.com on 29 Jan 2007 at 8:04

[GoogleCodeExporter]

Original comment by kazuh...@gmail.com on 1 Feb 2007 at 7:20

• Changed state: Verified