Re: [LIBERTAS-SDIO] Support for single transfer blocks?

Re: [LIBERTAS-SDIO] Support for single transfer blocks?

[Andrew Morton]

Let's Cc the wireless development list.

On Mon, 11 May 2009 16:41:51 +0200 Luis Galdos <luis.galdos@digi.com> wrote:

> Hi all,
> 
> I have one question concerning to the Libertas-driver: Does this driver works with 
> SDIO-hosts that only support single transfer blocks? I ask cause I have seen two problems 
> with a SDIO-port that doesn't support multiple blocks:
> 
> * The firmware installation successes only with a modification of the block size (see 
> below patch)
> 
> * The transfer of Ethernet-frames works only if the "complete" frame is smaller than the 
> block size of the SDIO-host. By larger packages, the SD8686 doesn't generate the expected 
> IRQ (cause it expected a multiple transfer) and the driver detects a timeout.
> 
> Do you know something about this? Thanks in advance,
> 
> 
> PS: Sorry for the possible wrong format of this email (my first one)
> 
> 
> diff --git a/drivers/net/wireless/libertas/if_sdio.c b/drivers/net/wireless/libertas/if_sdio.c
> index b54e2ea..f88a4da 100644
> --- a/drivers/net/wireless/libertas/if_sdio.c
> +++ b/drivers/net/wireless/libertas/if_sdio.c
> @@ -33,6 +33,7 @@
>   #include <linux/mmc/card.h>
>   #include <linux/mmc/sdio_func.h>
>   #include <linux/mmc/sdio_ids.h>
> +#include <linux/mmc/host.h>
> 
>   #include "host.h"
>   #include "decl.h"
> @@ -507,6 +508,8 @@ static int if_sdio_prog_real(struct if_sdio_card *card)
>          u32 chunk_size;
>          const u8 *firmware;
>          size_t size, req_size;
> +       struct mmc_host *host;
> +       int max_blksize = 0;
> 
>          lbs_deb_enter(LBS_DEB_SDIO);
> 
> @@ -524,7 +527,19 @@ static int if_sdio_prog_real(struct if_sdio_card *card)
> 
>          sdio_claim_host(card->func);
> 
> -       ret = sdio_set_block_size(card->func, 32);
> +       /*
> +        * If the host doesn't support multi-blocks, then use the the maximal block
> +        * size for the transfers. Otherwise the firmware installation will fail.
> +        */
> +       host = card->func->card->host;
> +       if (host->max_blk_count == 1) {
> +               lbs_pr_info("Setting block size to %u\n", host->max_blk_size);
> +               max_blksize = card->func->max_blksize;
> +               card->func->max_blksize = host->max_blk_size;
> +               ret = sdio_set_block_size(card->func, host->max_blk_size);
> +       } else
> +               ret = sdio_set_block_size(card->func, 32);
> +
>          if (ret)
>                  goto release;
> 
> @@ -593,6 +608,10 @@ static int if_sdio_prog_real(struct if_sdio_card *card)
> 
>          ret = 0;
> 
> +        /* Restore the original block size if it was changed before */
> +        if (max_blksize)
> +                card->func->max_blksize = max_blksize;
> +
>          lbs_deb_sdio("waiting for firmware to boot...\n");
> 
>          /* wait for the firmware to boot */
> 
> 
> -- 
> Luis Galdos
> --
> To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
> the body of a message to majordomo@vger.kernel.org
> More majordomo info at  http://vger.kernel.org/majordomo-info.html
> Please read the FAQ at  http://www.tux.org/lkml/

Re: [LIBERTAS-SDIO] Support for single transfer blocks?

[Dan Williams]

On Tue, 2009-05-12 at 23:11 -0700, Andrew Morton wrote:
> Let's Cc the wireless development list.

And lets CC the MMC subsystem maintainer and original author of
libertas_sdio too, so we can get an informed opinion :)

What firmware version are you using?  What SDIO controller is this?  Has
this controller been validated by Marvell for use with the 8686? (not
that that is required or anything, just curious).

All the information I have (up to firmware V10) references a block size
of 32 bytes.  The main firmware is transferred in 32-byte blocks using
CMD 53, with a maximum number of 16 blocks transferred in a single CMD
53 write depending on how much data the helper firmware can accept in
single write.  The available vendor drivers also use a 32-byte block
size.

Data transfers are specified to use a block size of 320 bytes.

Dan

> On Mon, 11 May 2009 16:41:51 +0200 Luis Galdos <luis.galdos@digi.com> wrote:
> 
> > Hi all,
> > 
> > I have one question concerning to the Libertas-driver: Does this driver works with 
> > SDIO-hosts that only support single transfer blocks? I ask cause I have seen two problems 
> > with a SDIO-port that doesn't support multiple blocks:
> > 
> > * The firmware installation successes only with a modification of the block size (see 
> > below patch)
> > 
> > * The transfer of Ethernet-frames works only if the "complete" frame is smaller than the 
> > block size of the SDIO-host. By larger packages, the SD8686 doesn't generate the expected 
> > IRQ (cause it expected a multiple transfer) and the driver detects a timeout.
> > 
> > Do you know something about this? Thanks in advance,
> > 
> > 
> > PS: Sorry for the possible wrong format of this email (my first one)
> > 
> > 
> > diff --git a/drivers/net/wireless/libertas/if_sdio.c b/drivers/net/wireless/libertas/if_sdio.c
> > index b54e2ea..f88a4da 100644
> > --- a/drivers/net/wireless/libertas/if_sdio.c
> > +++ b/drivers/net/wireless/libertas/if_sdio.c
> > @@ -33,6 +33,7 @@
> >   #include <linux/mmc/card.h>
> >   #include <linux/mmc/sdio_func.h>
> >   #include <linux/mmc/sdio_ids.h>
> > +#include <linux/mmc/host.h>
> > 
> >   #include "host.h"
> >   #include "decl.h"
> > @@ -507,6 +508,8 @@ static int if_sdio_prog_real(struct if_sdio_card *card)
> >          u32 chunk_size;
> >          const u8 *firmware;
> >          size_t size, req_size;
> > +       struct mmc_host *host;
> > +       int max_blksize = 0;
> > 
> >          lbs_deb_enter(LBS_DEB_SDIO);
> > 
> > @@ -524,7 +527,19 @@ static int if_sdio_prog_real(struct if_sdio_card *card)
> > 
> >          sdio_claim_host(card->func);
> > 
> > -       ret = sdio_set_block_size(card->func, 32);
> > +       /*
> > +        * If the host doesn't support multi-blocks, then use the the maximal block
> > +        * size for the transfers. Otherwise the firmware installation will fail.
> > +        */
> > +       host = card->func->card->host;
> > +       if (host->max_blk_count == 1) {
> > +               lbs_pr_info("Setting block size to %u\n", host->max_blk_size);
> > +               max_blksize = card->func->max_blksize;
> > +               card->func->max_blksize = host->max_blk_size;
> > +               ret = sdio_set_block_size(card->func, host->max_blk_size);
> > +       } else
> > +               ret = sdio_set_block_size(card->func, 32);
> > +
> >          if (ret)
> >                  goto release;
> > 
> > @@ -593,6 +608,10 @@ static int if_sdio_prog_real(struct if_sdio_card *card)
> > 
> >          ret = 0;
> > 
> > +        /* Restore the original block size if it was changed before */
> > +        if (max_blksize)
> > +                card->func->max_blksize = max_blksize;
> > +
> >          lbs_deb_sdio("waiting for firmware to boot...\n");
> > 
> >          /* wait for the firmware to boot */
> > 
> > 
> > -- 
> > Luis Galdos
> > --
> > To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
> > the body of a message to majordomo@vger.kernel.org
> > More majordomo info at  http://vger.kernel.org/majordomo-info.html
> > Please read the FAQ at  http://www.tux.org/lkml/
> --
> To unsubscribe from this list: send the line "unsubscribe linux-wireless" in
> the body of a message to majordomo@vger.kernel.org
> More majordomo info at  http://vger.kernel.org/majordomo-info.html

Re: [LIBERTAS-SDIO] Support for single transfer blocks?

[Pierre Ossman]

[-- Attachment #1: Type: text/plain, Size: 816 bytes --]

> > On Mon, 11 May 2009 16:41:51 +0200 Luis Galdos <luis.galdos@digi.com> wrote:
> > > 
> > > * The transfer of Ethernet-frames works only if the "complete" frame is smaller than the 
> > > block size of the SDIO-host. By larger packages, the SD8686 doesn't generate the expected 
> > > IRQ (cause it expected a multiple transfer) and the driver detects a timeout.
> > > 

Unfortunately, the libertas chip cannot support your host as it
requires everything to be done in a single transaction. So I'm afraid
you'll have to look at another host controller or another wifi chip.

Rgds
-- 
     -- Pierre Ossman

  WARNING: This correspondence is being monitored by the
  Swedish government. Make sure your server uses encryption
  for SMTP traffic and consider using PGP for end-to-end
  encryption.

[-- Attachment #2: signature.asc --]
[-- Type: application/pgp-signature, Size: 198 bytes --]

Re: [LIBERTAS-SDIO] Support for single transfer blocks?

[Luis Galdos]

Pierre Ossman wrote:
>>> On Mon, 11 May 2009 16:41:51 +0200 Luis Galdos <luis.galdos@digi.com> wrote:
>>>> * The transfer of Ethernet-frames works only if the "complete" frame is smaller than the 
>>>> block size of the SDIO-host. By larger packages, the SD8686 doesn't generate the expected 
>>>> IRQ (cause it expected a multiple transfer) and the driver detects a timeout.
>>>>
> 
> Unfortunately, the libertas chip cannot support your host as it
> requires everything to be done in a single transaction. So I'm afraid
> you'll have to look at another host controller or another wifi chip.
Thanks a lot for this info, this confirms my expectation. And sorry for the delayed 
feedback (I was on vacations).


> Rgds
Regards,


-- 
Luis Galdos

Re: [LIBERTAS-SDIO] Support for single transfer blocks?

[Luis Galdos]

Dan Williams wrote:
> On Tue, 2009-05-12 at 23:11 -0700, Andrew Morton wrote:
>> Let's Cc the wireless development list.
> 
> And lets CC the MMC subsystem maintainer and original author of libertas_sdio too, so
> we can get an informed opinion :)
> 
> What firmware version are you using?
The installed firmware version is 9.70.3p24 (downloaded from the Marvell's website).

> What SDIO controller is this?  Has this controller been validated by Marvell for use
> with the 8686? (not that that is required or anything, just curious).
The uC is from Digi (NS9215) and its SDIO-port hasn't been validated by Marvell.

> All the information I have (up to firmware V10) references a block size of 32 bytes.
> The main firmware is transferred in 32-byte blocks using CMD 53, with a maximum number
> of 16 blocks transferred in a single CMD 53 write depending on how much data the helper
> firmware can accept in single write.  The available vendor drivers also use a 32-byte
> block size.
That's right. I only was surprised that the SD8686 accepts data blocks of 512 bytes in a 
single transfer too. This is just nice for SDIO-ports that only support single blocks :) 
(like the NS9215). Anyway, Pierre has informed me that the SD8686 has need of multiple 
blocks support, and the NS9215 doesn't. So, the problem is the controller.


> Dan

Thanks and sorry for the delayed answer (I was on vacations),


-- 
Luis Galdos