From: Ivo van Doorn <ivdoorn@gmail.com>
To: users@rt2x00.serialmonkey.com
Cc: Eric <eric@lente.dyndns.org>,
linux-wireless@vger.kernel.org,
Johannes Berg <johannes@sipsolutions.net>
Subject: Re: [rt2x00-users] Alignment between rt2500pci and mac80211
Date: Sun, 24 May 2009 11:43:48 +0200 [thread overview]
Message-ID: <200905241143.48549.IvDoorn@gmail.com> (raw)
In-Reply-To: <4A190304.9050002@lente.dyndns.org>
Hi,
My knowledge about this part of mac80211 is very limited,
so perhaps Johannes or somebody else on the linux-wireless mailinglist
can answer below questions (I already added them to the CC list). :)
> As posted already in:
> http://rt2x00.serialmonkey.com/phpBB/viewtopic.php?f=5&t=5150&sid=ba55e2676ff5ff92ab36e7e16c29e7de
> Rafaello discovered an alignment problem between rt2500pci and mac80211
> while operating with hostapd in master mode.
> I did some debugging to find the cause of this problem and discovered
> that inside the net/mac80211/tx.c:__ieee80211_tx_prepare() function
> __ieee80211_parse_tx_radiotap() is being called, which in itself does an
> skb_pull with iterator.max_length (being 14 in my case).
> I think this is one of the causes of an alignment offset of 2 bytes of
> the skb->data pointer (IF !IS_ALIGNED((int)skb->data,4)), as
> dma_map_single requests require at least alignment at the 4 byte
> boundary (on x86 at least, corrrect me if I'm wrong, as
> http://www.mjmwired.net/kernel/Documentation/DMA-API.txt only talks
> about the cache line boundary).
> Another cause is the __skb_queue_tail() call inside
> __ieee80211_tx_prepare(), unfortunuatly I couldnt always reproduce the
> latter one.
>
> I have a few questions:
> Is there any agreement between mac80211 and maintaing alignment at all?
> Should anything coming back from mac80211 be rechecked for device
> specific alignment ?
>
> According to the comment inside __ieee80211_parse_tx_radiotap(), that
> function should only be called for injected packages, however I couldn't
> discover (yet) where exactly the flag IEE80211_TX_CTL_INJECTED is
> set.Forcefully skipping that part causes the frame to fail, so
> apparently it's neccesary, however I fail to see the logic behind it.
> Is this something the driver explicitly flags / needs?
>
> All the the debugging was done on the daily wireless-testing tree from
> linuxwireless.org.
>
>
> Any comments / insights would be appreciated :-)
>
>
> kind regards,
>
> Eric
>
> _______________________________________________
> users mailing list
> users@rt2x00.serialmonkey.com
> http://rt2x00.serialmonkey.com/mailman/listinfo/users_rt2x00.serialmonkey.com
>
parent reply other threads:[~2009-05-24 9:43 UTC|newest]
Thread overview: expand[flat|nested] mbox.gz Atom feed
[parent not found: <4A190304.9050002@lente.dyndns.org>]
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
Avoid top-posting and favor interleaved quoting:
https://en.wikipedia.org/wiki/Posting_style#Interleaved_style
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to=200905241143.48549.IvDoorn@gmail.com \
--to=ivdoorn@gmail.com \
--cc=eric@lente.dyndns.org \
--cc=johannes@sipsolutions.net \
--cc=linux-wireless@vger.kernel.org \
--cc=users@rt2x00.serialmonkey.com \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).