2011/2/9 George Kashperko <george@xxxxxxxxxxx>:
>> 2011/2/9 George Kashperko <george@xxxxxxxxxxx>:
>> > From: George Kashperko <george@xxxxxxxxxxx>
>> >
>> > Major differences between AI/SB from drivers' perespective is control
>> > and state flags handling. These long term used to be TMSLOW/TMSHIGH ones
>> > and in order to be handled transparently should be masked under some
>> > indirect access handlers. This patch forces use of ssb_core_ctl_flags
>> > and ssb_core_state_flags introduced early. TMSLOW/TMSHIGH defines outside
>> > ssb code (b43, ssb gige and ohci drivers) were renamed to reflect actual
>> > meaning, flags' definitions moved to low two bytes. Common TMSLOW/TMSHIGH
>> > defines moved to ssb_private.h in order to limit their use by internal
>> > SB-specic code only.
>> > Signed-off-by: George Kashperko <george@xxxxxxxxxxx>
>> > ---
>> > Âdrivers/net/b44.c             Â|  Â2
>> > Âdrivers/net/wireless/b43/b43.h       |  32 ++++-----
>> > Âdrivers/net/wireless/b43/dma.c       |  Â4 -
>> > Âdrivers/net/wireless/b43/main.c      Â|  61 +++++++------------
>> > Âdrivers/net/wireless/b43/phy_g.c      |  Â2
>> > Âdrivers/net/wireless/b43/phy_n.c      |  18 +----
>> > Âdrivers/net/wireless/b43legacy/b43legacy.h | Â 20 +++---
>> > Âdrivers/net/wireless/b43legacy/dma.c    |  Â4 -
>> > Âdrivers/net/wireless/b43legacy/main.c   Â|  52 +++++-----------
>> > Âdrivers/net/wireless/b43legacy/phy.c    |  Â2
>> > Âdrivers/ssb/driver_gige.c         Â|  19 ++---
>> > Âdrivers/ssb/main.c             |  20 ++++--
>> > Âdrivers/ssb/ssb_private.h         Â|  24 +++++++
>> > Âdrivers/usb/host/ohci-ssb.c        Â|  Â4 -
>> > Âinclude/linux/ssb/ssb.h          Â|  Â4 -
>> > Âinclude/linux/ssb/ssb_driver_gige.h    Â|  12 +--
>> > Âinclude/linux/ssb/ssb_regs.h        |  42 ++++++-------
>> > Â17 files changed, 160 insertions(+), 162 deletions(-)
>> > --- linux-next-20110203.orig/drivers/net/b44.c Â2011-02-01 05:05:49.000000000 +0200
>> > +++ linux-next-20110203/drivers/net/b44.c    2011-02-08 12:17:03.000000000 +0200
>> > @@ -1568,7 +1568,7 @@ static void b44_setup_wol_pci(struct b44
>> > Â Â Â Âu16 val;
>> >
>> > Â Â Â Âif (bp->sdev->bus->bustype != SSB_BUSTYPE_SSB) {
>> > - Â Â Â Â Â Â Â bw32(bp, SSB_TMSLOW, br32(bp, SSB_TMSLOW) | SSB_TMSLOW_PE);
>> > + Â Â Â Â Â Â Â ssb_core_ctl_flags(bp->sdev, 0, SSB_CORECTL_PE, NULL);
>> > Â Â Â Â Â Â Â Âpci_read_config_word(bp->sdev->bus->host_pci, SSB_PMCSR, &val);
>> > Â Â Â Â Â Â Â Âpci_write_config_word(bp->sdev->bus->host_pci, SSB_PMCSR, val | SSB_PE);
>> > Â Â Â Â}
>> > --- linux-next-20110203.orig/drivers/net/wireless/b43/b43.h   2011-02-01 05:05:49.000000000 +0200
>> > +++ linux-next-20110203/drivers/net/wireless/b43/b43.h Â2011-02-08 12:17:03.000000000 +0200
>> > @@ -414,22 +414,22 @@ enum {
>> > Â#define B43_MACCMD_CCA Â Â Â Â Â Â Â Â 0x00000008 Â Â Â/* Clear channel assessment */
>> > Â#define B43_MACCMD_BGNOISE Â Â Â Â Â Â 0x00000010 Â Â Â/* Background noise */
>> >
>> > -/* 802.11 core specific TM State Low (SSB_TMSLOW) flags */
>> > -#define B43_TMSLOW_GMODE Â Â Â Â Â Â Â 0x20000000 Â Â Â/* G Mode Enable */
>> > -#define B43_TMSLOW_PHY_BANDWIDTH Â Â Â 0x00C00000 Â Â Â/* PHY band width and clock speed mask (N-PHY only) */
>> > -#define ÂB43_TMSLOW_PHY_BANDWIDTH_10MHZ Â Â Â Â0x00000000 Â Â Â/* 10 MHz bandwidth, 40 MHz PHY */
>> > -#define ÂB43_TMSLOW_PHY_BANDWIDTH_20MHZ Â Â Â Â0x00400000 Â Â Â/* 20 MHz bandwidth, 80 MHz PHY */
>> > -#define ÂB43_TMSLOW_PHY_BANDWIDTH_40MHZ Â Â Â Â0x00800000 Â Â Â/* 40 MHz bandwidth, 160 MHz PHY */
>> > -#define B43_TMSLOW_PLLREFSEL Â Â Â Â Â 0x00200000 Â Â Â/* PLL Frequency Reference Select (rev >= 5) */
>> > -#define B43_TMSLOW_MACPHYCLKEN Â Â Â Â 0x00100000 Â Â Â/* MAC PHY Clock Control Enable (rev >= 5) */
>> > -#define B43_TMSLOW_PHYRESET Â Â Â Â Â Â0x00080000 Â Â Â/* PHY Reset */
>> > -#define B43_TMSLOW_PHYCLKEN Â Â Â Â Â Â0x00040000 Â Â Â/* PHY Clock Enable */
>> > -
>> > -/* 802.11 core specific TM State High (SSB_TMSHIGH) flags */
>> > -#define B43_TMSHIGH_DUALBAND_PHY Â Â Â 0x00080000 Â Â Â/* Dualband PHY available */
>> > -#define B43_TMSHIGH_FCLOCK Â Â Â Â Â Â 0x00040000 Â Â Â/* Fast Clock Available (rev >= 5) */
>> > -#define B43_TMSHIGH_HAVE_5GHZ_PHY Â Â Â0x00020000 Â Â Â/* 5 GHz PHY available (rev >= 5) */
>> > -#define B43_TMSHIGH_HAVE_2GHZ_PHY Â Â Â0x00010000 Â Â Â/* 2.4 GHz PHY available (rev >= 5) */
>> > +/* 802.11 core specific control flags */
>> > +#define B43_CORECTL_GMODE Â Â Â Â Â Â Â Â Â Â Â0x2000 Â/* G Mode Enable */
>> > +#define B43_CORECTL_PHY_BANDWIDTH Â Â Â Â Â Â Â0x00C0 Â/* PHY band width and clock speed mask (N-PHY only) */
>> > +#define ÂB43_CORECTL_PHY_BANDWIDTH_10MHZ Â Â Â 0x0000 Â/* 10 MHz bandwidth, 40 MHz PHY */
>> > +#define ÂB43_CORECTL_PHY_BANDWIDTH_20MHZ Â Â Â 0x0040 Â/* 20 MHz bandwidth, 80 MHz PHY */
>> > +#define ÂB43_CORECTL_PHY_BANDWIDTH_40MHZ Â Â Â 0x0080 Â/* 40 MHz bandwidth, 160 MHz PHY */
>> > +#define B43_CORECTL_PLLREFSEL Â Â Â Â Â Â Â Â Â0x0020 Â/* PLL Frequency Reference Select (rev >= 5) */
>> > +#define B43_CORECTL_MACPHYCLKEN Â Â Â Â Â Â Â Â Â Â Â Â0x0010 Â/* MAC PHY Clock Control Enable (rev >= 5) */
>> > +#define B43_CORECTL_PHYRESET Â Â Â Â Â Â Â Â Â 0x0008 Â/* PHY Reset */
>> > +#define B43_CORECTL_PHYCLKEN Â Â Â Â Â Â Â Â Â 0x0004 Â/* PHY Clock Enable */
>> > +
>> > +/* 802.11 core specific state flags */
>> > +#define B43_CORESTAT_DUALBAND_PHY Â Â Â Â Â Â Â0x0008 Â/* Dualband PHY available */
>> > +#define B43_CORESTAT_FCLOCK Â Â Â Â Â Â Â Â Â Â0x0004 Â/* Fast Clock Available (rev >= 5) */
>> > +#define B43_CORESTAT_HAVE_5GHZ_PHY Â Â Â Â Â Â 0x0002 Â/* 5 GHz PHY available (rev >= 5) */
>> > +#define B43_CORESTAT_HAVE_2GHZ_PHY Â Â Â Â Â Â 0x0001 Â/* 2.4 GHz PHY available (rev >= 5) */
>>
>> I don't know, I've doubts about it :/ That flags are quite messy, mask
>> for them is 0x3FFC0000. Ideally we should shift it by 14 bits, but
>> that would make writing code harder. Maaaaybe... Michael what do you
>> think about this?
>>
>>
>> > @@ -1213,6 +1214,9 @@ static void ssb_device_disable_sb(struct
>> > Â Â Â Âif (ssb_read32(dev, SSB_TMSLOW) & SSB_TMSLOW_RESET)
>> > Â Â Â Â Â Â Â Âreturn;
>> >
>> > + Â Â Â SSB_WARN_ON(core_specific_flags & 0xFFFF0000);
>>
>> Make it ~0x00003FFC, value 0x1 is not valid flag.
> Actual mask for _core_ flags is 0xFFFF0000 for SB and 0x0000FFFF for AI.
OK.
> 0x3FFC is mask for core dependent flags.
> SSB_CORECTL_CLOCK | SSB_CORECTL_FGC | SSB_CORECTL_PE | SSB_CORECTL_BE |
> 0x3FFC = 0xFFFF
> Therefore sanity check for SB is &0xFFFF0000, for AI is &0x0000FFFF
> Layout of core flags in these 2 bytes is the same for AI and SB cores.
You created WARN_ON against core_specific_flags, not sth like
core_flags. That's why I think test should be against 0x3FFC, not
0xFFFF. Am I wrong? Or if we ever pass not-core-specific flags to
ssb_device_disable/ssb_device_enable, we should rename variable from
core_specific_flags to core_flags.
>> Why don't you go ahead and make core flags u16?
>>
> Dont see any problems with this as other 2 bytes aint used currently in
> any way.
You do not see problems with making it u16, or problems with it being u32?
--
RafaÅ
--
To unsubscribe from this list: send the line "unsubscribe linux-wireless" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
