Re: [lwip-users] netconn_delete and _recv not thread safe

[Jonathan Larmour]

Rob Duncan wrote:
> This is possibly related to an older thread:
>
> http://www.mail-archive.com/address@hidden/msg01479.html
>
> I'm seeing a similar problem (under eCos) where I'm in the middle of 
> netconn_delete and a RST for that connection arrives.  I end up with an 
> extra message in conn->mbox (the one from lwip_close(), and another that 
> err_tcp() adds for the RST).  This means that the mbox is not empty when 
> sys_mbox_free() is called, and I get an assertion failure.
>
> Does this analysis seem right?  And if so what would be the best way of 
> correcting the problem?  I'm a bit fuzzy on how thread safety is 
> supposed to be done here, so any hints or pointers to more information 
> would be very useful.

I suggest doing what I said in that message, since that worked for me at 
the time: change sys_arch_mbox_fetch so that a timeout of 1 causes a poll 
of the mbox, not a timed wait. This will be fixed in a later stable 
release of lwIP.

Jifl
--
eCosCentric Limited      http://www.eCosCentric.com/     The eCos experts
Barnwell House, Barnwell Drive, Cambridge, UK.       Tel: +44 1223 245571
Registered in England and Wales: Reg No 4422071.
  **  Visit us at ESC Silicon Valley <http://www.embedded.com/esc/sv> **
  **  April 3-5 2007, Booth 1922, San Jose McEnery Convention Center  **
------["The best things in life aren't things."]------      Opinions==mine