Great news!
Today I applied ath9k patches on top of linux-wireless git tree on a
MacBookPro2,2 and I found that it works! (after a little bit of
hacking...)
First of all, my hardware is the following:
# lspci -v
03:00.0 Network controller: Atheros Communications Inc. AR5418
802.11abgn Wireless PCI Express Adapter (rev 01)
Flags: bus master, fast devsel, latency 0, IRQ 17
Memory at 98100000 (64-bit, non-prefetchable) [size=64K]
Capabilities: [40] Power Management version 2
Capabilities: [50] Message Signalled Interrupts: Mask- 64bit-
Queue=0/0 Enable-
Capabilities: [60] Express Legacy Endpoint, MSI 00
Capabilities: [90] MSI-X: Enable- Mask- TabSize=1
Kernel driver in use: ath9k
Kernel modules: ath9k
# dmesg
[ 7.760727] ath9k: 0.1
[ 7.760727] ACPI: PCI Interrupt 0000:03:00.0[A] -> GSI 17 (level,
low) -> IRQ 17
[ 7.760727] PCI: Setting latency timer of device 0000:03:00.0 to 64
[ 7.892753] Bad EEPROM checksum 0x866 or revision 0x000e
[ 7.894551] phy0: Selected rate control algorithm 'ath9k_rate_control'
[ 8.167891] phy0: Atheros 5416: mem=0xffffc20000500000, irq=17
The "bad eeprom checksum" message is printed (but the failure is
ignored) because I made a small change to ath9k/hw.c:
--------------------
diff --git a/drivers/net/wireless/ath9k/hw.c b/drivers/net/wireless/ath9k/hw.c
index 5ebf35e..aa0a3d9 100644
--- a/drivers/net/wireless/ath9k/hw.c
+++ b/drivers/net/wireless/ath9k/hw.c
@@ -911,10 +911,10 @@ static inline enum hal_status
ath9k_hw_check_eeprom(struct ath_hal *ah)
if (sum != 0xffff || ar5416_get_eep_ver(ahp) != AR5416_EEP_VER ||
ar5416_get_eep_rev(ahp) < AR5416_EEP_NO_BACK_VER) {
- HDPRINTF(ah, HAL_DBG_EEPROM,
+ printk(KERN_DEBUG
"Bad EEPROM checksum 0x%x or revision 0x%04x\n",
sum, ar5416_get_eep_ver(ahp));
- return HAL_EEBADSUM;
+ //return HAL_EEBADSUM;
}
return HAL_OK;
--------------------
Without the above patch, the module would fail with the following error:
"ath_init: unable to attach hardware; HAL status 7"
Status code 7 is HAL_EEBADSUM indeed.
Maybe MacBook Pro's card has different eeprom's contents? Anyway the
card now works fine, and it also appears quite stable! ;-)
Thank you very much for your work!
Regards,
Davide
P.S.: I haven't been able to compile a kernel from wireless-testing
git tree with debug info + frame pointers, because they all failed to
boot very early (before console initialization). Is this a known bug?
On Tue, Jul 22, 2008 at 2:53 PM, Davide Pesavento <[email protected]> wrote:
> Great news!
>
> Today I applied ath9k patches on top of linux-wireless git tree on a
> MacBookPro2,2 and I found that it works! (after a little bit of
> hacking...)
>
> First of all, my hardware is the following:
>
> # lspci -v
> 03:00.0 Network controller: Atheros Communications Inc. AR5418
> 802.11abgn Wireless PCI Express Adapter (rev 01)
> Flags: bus master, fast devsel, latency 0, IRQ 17
> Memory at 98100000 (64-bit, non-prefetchable) [size=64K]
> Capabilities: [40] Power Management version 2
> Capabilities: [50] Message Signalled Interrupts: Mask- 64bit-
> Queue=0/0 Enable-
> Capabilities: [60] Express Legacy Endpoint, MSI 00
> Capabilities: [90] MSI-X: Enable- Mask- TabSize=1
> Kernel driver in use: ath9k
> Kernel modules: ath9k
>
> # dmesg
> [ 7.760727] ath9k: 0.1
> [ 7.760727] ACPI: PCI Interrupt 0000:03:00.0[A] -> GSI 17 (level,
> low) -> IRQ 17
> [ 7.760727] PCI: Setting latency timer of device 0000:03:00.0 to 64
> [ 7.892753] Bad EEPROM checksum 0x866 or revision 0x000e
> [ 7.894551] phy0: Selected rate control algorithm 'ath9k_rate_control'
> [ 8.167891] phy0: Atheros 5416: mem=0xffffc20000500000, irq=17
>
> The "bad eeprom checksum" message is printed (but the failure is
> ignored) because I made a small change to ath9k/hw.c:
>
> --------------------
> diff --git a/drivers/net/wireless/ath9k/hw.c b/drivers/net/wireless/ath9k/hw.c
> index 5ebf35e..aa0a3d9 100644
> --- a/drivers/net/wireless/ath9k/hw.c
> +++ b/drivers/net/wireless/ath9k/hw.c
> @@ -911,10 +911,10 @@ static inline enum hal_status
> ath9k_hw_check_eeprom(struct ath_hal *ah)
>
> if (sum != 0xffff || ar5416_get_eep_ver(ahp) != AR5416_EEP_VER ||
> ar5416_get_eep_rev(ahp) < AR5416_EEP_NO_BACK_VER) {
> - HDPRINTF(ah, HAL_DBG_EEPROM,
> + printk(KERN_DEBUG
> "Bad EEPROM checksum 0x%x or revision 0x%04x\n",
> sum, ar5416_get_eep_ver(ahp));
> - return HAL_EEBADSUM;
> + //return HAL_EEBADSUM;
> }
>
> return HAL_OK;
> --------------------
>
> Without the above patch, the module would fail with the following error:
> "ath_init: unable to attach hardware; HAL status 7"
> Status code 7 is HAL_EEBADSUM indeed.
>
> Maybe MacBook Pro's card has different eeprom's contents? Anyway the
> card now works fine, and it also appears quite stable! ;-)
> Thank you very much for your work!
I believe your last patch may have been "Use skb_end_pointer() ..." if
so then this is why you had this issue. Give the git tree a shot:
git://git.kernel.org/pub/scm/linux/kernel/git/mcgrof/ath9k.git
Luis
On Wed, 2008-07-23 at 05:13 -0700, Luis R. Rodriguez wrote:
> > Maybe MacBook Pro's card has different eeprom's contents? Anyway the
> > card now works fine, and it also appears quite stable! ;-)
> > Thank you very much for your work!
>
> I believe your last patch may have been "Use skb_end_pointer() ..." if
> so then this is why you had this issue. Give the git tree a shot:
>
> git://git.kernel.org/pub/scm/linux/kernel/git/mcgrof/ath9k.git
Luis, I'm afraid that code is still wrong. It used to be min(), but now
it's essentially max(). You wrote that there was an error in my patch
and I agreed, but we were talking of different errors, I'm afraid.
Since you have touched the code after me and you know more about the
hardware, I would prefer that you fix it and recheck everything once
again.
Here's the patch. I don't feel comfortable about describing things I
don't fully understand.
diff --git a/ksrc/hw.c b/ksrc/hw.c
index 5c8127e..e0a5464 100644
--- a/ksrc/hw.c
+++ b/ksrc/hw.c
@@ -852,7 +852,7 @@ static inline enum hal_status ath9k_hw_check_eeprom(struct ath_hal *ah)
else
el = ahp->ah_eeprom.baseEepHeader.length;
- if (el < sizeof(struct ar5416_eeprom))
+ if (el > sizeof(struct ar5416_eeprom))
el = sizeof(struct ar5416_eeprom) / sizeof(u_int16_t);
else
el = el / sizeof(u_int16_t);
--
Regards,
Pavel Roskin
On Wed, Jul 23, 2008 at 8:42 PM, Pavel Roskin <[email protected]> wrote:
> On Wed, 2008-07-23 at 05:13 -0700, Luis R. Rodriguez wrote:
>> > Maybe MacBook Pro's card has different eeprom's contents? Anyway the
>> > card now works fine, and it also appears quite stable! ;-)
>> > Thank you very much for your work!
>>
>> I believe your last patch may have been "Use skb_end_pointer() ..." if
>> so then this is why you had this issue. Give the git tree a shot:
>>
>> git://git.kernel.org/pub/scm/linux/kernel/git/mcgrof/ath9k.git
>
> Luis, I'm afraid that code is still wrong. It used to be min(), but now
> it's essentially max(). You wrote that there was an error in my patch
> and I agreed, but we were talking of different errors, I'm afraid.
>
> Since you have touched the code after me and you know more about the
> hardware, I would prefer that you fix it and recheck everything once
> again.
>
> Here's the patch. I don't feel comfortable about describing things I
> don't fully understand.
Pavel, you are correct, thanks for the new fix. I'll arrange so that
the patch that goes in only fixes this once on wireless-testing.
Luis