Re: kernel BUG at drivers/net/wireless/iwlwifi/iwl3945-base.c:3127!

[reinette chatre <reinette.chatre@intel.com>]

On Thu, 2009-03-05 at 20:12 -0800, Jason Andryuk wrote:
> On Thu, 2009-03-05 at 16:24 -0800, reinette chatre wrote:
> > On Thu, 2009-03-05 at 15:50 -0800, Jason Andryuk wrote:
> > > On Wed, 2009-03-04 at 16:04 -0800, reinette chatre wrote:
> > > > On Wed, 2009-03-04 at 11:47 -0800, Jason Andryuk wrote:
> > > > > On Wed, Mar 4, 2009 at 2:19 PM, reinette chatre
> > > > > <reinette.chatre@intel.com> wrote:
> > > > > > On Tue, 2009-03-03 at 20:32 -0800, Jason Andryuk wrote:
> > > > > >> On Sun, Mar 1, 2009 at 10:37 PM, Jason Andryuk <jandryuk@gmail.com> wrote:
> > > > > >> > I tested c2a0aa3cb733452e749727680e380dca6cc10a68 "iwl3945: use
> > > > > >> > iwl_rb_status" with the additional patches to initialize rb_stts
> > > > > >> > (copied from iwl_rx_queue_alloc) and to convert the BUG_ON to a
> > > > > >> > WARN_ON.
> > > > > >> >
> > > > > >> > The driver runs without segfaulting.  So far the WARN_ON has not been
> > > > > >> > triggered, but the driver fails to finish authentication with the AP.
> > > > > >> > I receive "wlan0: authentication with AP 00:30:bd:c0:90:24 timed out"
> > > > > >>
> > > > > >> Ok, I started poking around with manually checking interesting commits instead of git bisect.
> > > > > >>
> > > > > >> 55b3e21... iwl3945: use iwl-io.h and delete iwl-3945-io.h with BUG->WARN patch works.
> > > > > >>
> > > > > >> bb64785... iwlwifi: use iwl_cmd instead of iwl3945_cmd
> > > > > >> fails to authenticate to the AP.  It generates a Microcode SW Error attached below.
> > > > > >>
> > > > > >> The previous revision:
> > > > > >> a0dedce... iwlwifi: replace IWL_ERROR with IWL_ERR
> > > > > >> Authenticates to the AP and works just fine.
> > > > > >>
> > > > > >> I keep looking over commitdiff bb64785... but I cannot see what is causing problems.  I even tried modifying iwl3945_cmd in various ways from a0dedce... but those attempts kept working.
> > > > > >>
> > > > > >> During testing, I saw a lot of UNKNOWN 0x1b
> > > > > >> [ 8892.625374] ieee80211 phy0: I iwl3945_rx_handle r = 162, i = 160, UNKNOWN, 0x1b
> > > > > >> Currently in the logs for commit a0dedce... it shows up as REPLY_3945_RX.
> > > > > >>
> > > > > >> Any ideas?
> > > > > >
> > > > > > Thank you very much for this digging.
> > > > > >
> > > > > > Are you running on 64bit?
> > > > >
> > > > > Yes, 64bit.
> > > >
> > > > could you please try this patch on top of latest wireless-testing?
> > > >
> > > > Thanks
> > > >
> > > > Reinette
> > > >
> > > >
> > > > diff --git a/drivers/net/wireless/iwlwifi/iwl-3945.c b/drivers/net/wireless/iwlwifi/iwl-3945.c
> > > > index 0a75053..00b25a8 100644
> > > > --- a/drivers/net/wireless/iwlwifi/iwl-3945.c
> > > > +++ b/drivers/net/wireless/iwlwifi/iwl-3945.c
> > > > @@ -743,6 +743,7 @@ void iwl3945_hw_txq_free_tfd(struct iwl_priv *priv, struct iwl_tx_queue *txq)
> > > >       struct iwl3945_tfd *tfd_tmp = (struct iwl3945_tfd *)txq->tfds;
> > > >       struct iwl3945_tfd *tfd = &tfd_tmp[txq->q.read_ptr];
> > > >       struct pci_dev *dev = priv->pci_dev;
> > > > +     int index = txq->q.read_ptr;
> > > >       int i;
> > > >       int counter;
> > > >
> > > > @@ -759,6 +760,13 @@ void iwl3945_hw_txq_free_tfd(struct iwl_priv *priv, struct iwl_tx_queue *txq)
> > > >               return;
> > > >       }
> > > >
> > > > +     /* Unmap tx_cmd */
> > > > +     if (counter)
> > > > +             pci_unmap_single(dev,
> > > > +                             pci_unmap_addr(&txq->cmd[index]->meta, mapping),
> > > > +                             pci_unmap_len(&txq->cmd[index]->meta, len),
> > > > +                             PCI_DMA_TODEVICE);
> > > > +
> > > >       /* unmap chunks if any */
> > > >
> > > >       for (i = 1; i < counter; i++) {
> > >
> > > Reinette,
> > >
> > > I applied the patch, but iwl3945 did not authenticate with my AP
> > >
> >
> > oh no.
> >
> > I will install a 64bit system here to try and reproduce this ...
> >
> > Reinette
> 
> I created the patch below, made of the iwl_host_cmd and iwl_meta_cmd
> portions of commit "bb64785... iwlwifi: use iwl_cmd instead of
> iwl3945_cmd".  Applied to "a0dedce... iwlwifi: replace IWL_ERROR with
> IWL_ERR", I am able to connect and transfer data successfully.  There
> are three warning during compilation (type mismatch with
> cmd.meta.u.callback).

wow - thanks!

hmmm ... your patch just blasted my theory about what is going on. I was
convinced that the problem is the presence of the DECLARE_PCI_UNMAP...
members in the iwl_cmd_meta struct. These only really count on a 64bit
system. The iwl3945_cmd_meta did not have them and I thus focused here.

Thank you very much for your digging. I'll look into this more detail
tomorrow.

About your compile warning - it is probably caused by the callback now
having a iwl_cmd function parameter and somewhere the code places a
iwl3945_cmd in there.

Reinette