Re: [ath9k-devel] Patching ath9k/mac80211 to change MAC layer features

On Fri, Jun 17, 2011 at 2:09 AM, Fred Matthews [off-list ref] wrote:

Hello all,


I have a query regarding patching ath9k/mac80211 to change MAC layer
features and I would be most grateful if someone can assist.


I need to patch ath9k so that I can be able to modify/fix the A-MPDU
aggregation frame length and other parameters such that they can be
configured at runtime via debugfs. I made a newbie attempt at trying to
change the aggregate length (attached), but I am unsure of the
correctness/efficiency  - when applied throughput varies with different
frame lengths entered but I can't be sure if its actually changing the
aggregation length (passes the framelen/aggr_limit). Hence I would be very
grateful if someone can assist in the correct way of doing this (possibly
directly via mac80211).



Also if the ampdu aggregate length is changed I believe the number/length of
the subframes of the aggregate stays constant (at the default which is
32). I try to change this in my patch as well as the maximum software
retries by just changing the default definition in ath9k.h, thus is there a
(more) correct/better way of changing these. Finally, how can I patch the
driver to be able to disable Block-Acks/aggregation via debugfs or
otherwise.

to disable aggregation in the driver, please try this

diff --git a/drivers/net/wireless/ath/ath9k/init.c

b/drivers/net/wireless/ath/ath9k/init.c
index d4b166c..e0175d0 100644

--- a/drivers/net/wireless/ath/ath9k/init.c
+++ b/drivers/net/wireless/ath/ath9k/init.c

@@ -524,8 +524,6 @@ static void ath9k_init_misc(struct ath_softc *sc)
        sc->config.txpowlimit = ATH_TXPOWER_MAX;

        if (sc->sc_ah->caps.hw_caps & ATH9K_HW_CAP_HT) {
-               sc->sc_flags |= SC_OP_TXAGGR;
-               sc->sc_flags |= SC_OP_RXAGGR;
        }

        common->tx_chainmask = sc->sc_ah->caps.tx_chainmask;

if you want only tx-aggregation to be disabled , please comment out only that.
also you can completely disable HT by

diff --git a/drivers/net/wireless/ath/ath9k/hw.c

b/drivers/net/wireless/ath/ath9k/hw.c
index 6de2655..696b4cc 100644

--- a/drivers/net/wireless/ath/ath9k/hw.c
+++ b/drivers/net/wireless/ath/ath9k/hw.c

@@ -1915,10 +1915,7 @@ int ath9k_hw_fill_cap_info(struct ath_hw *ah)

        common->crypt_caps |= ATH_CRYPT_CAP_CIPHER_AESCCM;

-       if (ah->hw_version.devid != AR2427_DEVID_PCIE)
-               pCap->hw_caps |= ATH9K_HW_CAP_HT;
-       else
-               pCap->hw_caps &= ~ATH9K_HW_CAP_HT;
+       pCap->hw_caps &= ~ATH9K_HW_CAP_HT;

        if (AR_SREV_9271(ah))
                pCap->num_gpio_pins = AR9271_NUM_GPIO;

I highly appreciate, help in any of the above, as well as other advice,
Many thanks

_______________________________________________
ath9k-devel mailing list
ath9k-devel@lists.ath9k.org
https://lists.ath9k.org/mailman/listinfo/ath9k-devel

-- 
shafi

Hi all,

Much appreciated Mohammed Shafi, disabling aggregation works when
replacing those lines with
 sc->sc_flags &= ~SC_OP_TXAGGR;
 sc->sc_flags &= ~SC_OP_RXAGGR;
only issue is I must always recompile compat-wireless to
re-enable/disable it again. I tried to port it using debugfs to be
able to enable and disable at runtime but ath9k kept rebooting.

Also I would be very grateful if someone can help regarding my
attached patch and previous query:

------------
 need to patch ath9k so that I can be able to modify/fix the A-MPDU
aggregation frame length and other parameters such that they can be
configured at runtime via debugfs. I made a newbie attempt at trying
to change the aggregate length (attached), but I am unsure of the
correctness/efficiency  - when applied throughput varies with
different frame lengths entered but I can't be sure if its actually
changing the aggregation length (passes the framelen/aggr_limit).
Hence I would be very grateful if someone can assist in the correct
way of doing this (possibly directly via mac80211).


Also if the ampdu aggregate length is changed I believe the
number/length of the subframes of the aggregate stays constant (at the
default which is 32). I try to change this in my patch as well as the
maximum software retries by just changing the default definition in
ath9k.h, thus is there a (more) correct/better way of changing these.
Finally, how can I patch the driver to be able to disable Block-Acks
via debugfs or otherwise.
---------------

Thank you very much



On Mon, Jun 20, 2011 at 9:15 AM, Mohammed Shafi
[off-list ref] wrote:

On Mon, Jun 27, 2011 at 11:07 PM, Fred Matthews [off-list ref] wrote:

Hi all,

Much appreciated Mohammed Shafi, disabling aggregation works when
replacing those lines with
 sc->sc_flags &= ~SC_OP_TXAGGR;
 sc->sc_flags &= ~SC_OP_RXAGGR;
only issue is I must always recompile compat-wireless to
re-enable/disable it again. I tried to port it using debugfs to be
able to enable and disable at runtime but ath9k kept rebooting.

I am not sure we can enable/disable aggregation via debugfs, these
flags seem are set at initialization and based on this lot of things
happen in tx-path, initializing TID structure, ampdu action, rate
control  etc.

Also I would be very grateful if someone can help regarding my
attached patch and previous query:

------------
 need to patch ath9k so that I can be able to modify/fix the A-MPDU
aggregation frame length and other parameters such that they can be
configured at runtime via debugfs. I made a newbie attempt at trying
to change the aggregate length (attached), but I am unsure of the
correctness/efficiency  - when applied throughput varies with
different frame lengths entered but I can't be sure if its actually
changing the aggregation length (passes the framelen/aggr_limit).
Hence I would be very grateful if someone can assist in the correct
way of doing this (possibly directly via mac80211).


Also if the ampdu aggregate length is changed I believe the
number/length of the subframes of the aggregate stays constant (at the
default which is 32). I try to change this in my patch as well as the
maximum software retries by just changing the default definition in
ath9k.h, thus is there a (more) correct/better way of changing these.
Finally, how can I patch the driver to be able to disable Block-Acks
via debugfs or otherwise.
---------------

Thank you very much



On Mon, Jun 20, 2011 at 9:15 AM, Mohammed Shafi
[off-list ref] wrote:

quoted

On Fri, Jun 17, 2011 at 2:09 AM, Fred Matthews [off-list ref] wrote:

quoted

Hello all,


I have a query regarding patching ath9k/mac80211 to change MAC layer
features and I would be most grateful if someone can assist.


I need to patch ath9k so that I can be able to modify/fix the A-MPDU
aggregation frame length and other parameters such that they can be
configured at runtime via debugfs. I made a newbie attempt at trying to
change the aggregate length (attached), but I am unsure of the
correctness/efficiency  - when applied throughput varies with different
frame lengths entered but I can't be sure if its actually changing the
aggregation length (passes the framelen/aggr_limit). Hence I would be very
grateful if someone can assist in the correct way of doing this (possibly
directly via mac80211).



Also if the ampdu aggregate length is changed I believe the number/length of
the subframes of the aggregate stays constant (at the default which is
32). I try to change this in my patch as well as the maximum software
retries by just changing the default definition in ath9k.h, thus is there a
(more) correct/better way of changing these. Finally, how can I patch the
driver to be able to disable Block-Acks/aggregation via debugfs or
otherwise.

to disable aggregation in the driver, please try this

diff --git a/drivers/net/wireless/ath/ath9k/init.c

b/drivers/net/wireless/ath/ath9k/init.c
index d4b166c..e0175d0 100644

--- a/drivers/net/wireless/ath/ath9k/init.c
+++ b/drivers/net/wireless/ath/ath9k/init.c

@@ -524,8 +524,6 @@ static void ath9k_init_misc(struct ath_softc *sc)

       sc->config.txpowlimit = ATH_TXPOWER_MAX;

       if (sc->sc_ah->caps.hw_caps & ATH9K_HW_CAP_HT) {
-               sc->sc_flags |= SC_OP_TXAGGR;
-               sc->sc_flags |= SC_OP_RXAGGR;
       }

       common->tx_chainmask = sc->sc_ah->caps.tx_chainmask;

if you want only tx-aggregation to be disabled , please comment out only that.
also you can completely disable HT by

diff --git a/drivers/net/wireless/ath/ath9k/hw.c

b/drivers/net/wireless/ath/ath9k/hw.c
index 6de2655..696b4cc 100644

--- a/drivers/net/wireless/ath/ath9k/hw.c
+++ b/drivers/net/wireless/ath/ath9k/hw.c

@@ -1915,10 +1915,7 @@ int ath9k_hw_fill_cap_info(struct ath_hw *ah)

       common->crypt_caps |= ATH_CRYPT_CAP_CIPHER_AESCCM;

-       if (ah->hw_version.devid != AR2427_DEVID_PCIE)
-               pCap->hw_caps |= ATH9K_HW_CAP_HT;
-       else
-               pCap->hw_caps &= ~ATH9K_HW_CAP_HT;
+       pCap->hw_caps &= ~ATH9K_HW_CAP_HT;

       if (AR_SREV_9271(ah))
               pCap->num_gpio_pins = AR9271_NUM_GPIO;

quoted

I highly appreciate, help in any of the above, as well as other advice,
Many thanks

_______________________________________________
ath9k-devel mailing list
ath9k-devel@lists.ath9k.org
https://lists.ath9k.org/mailman/listinfo/ath9k-devel

--
shafi
--
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

-- 
shafi

On Tue, Jun 28, 2011 at 3:00 AM, Mohammed Shafi
[off-list ref] wrote:

On Mon, Jun 27, 2011 at 11:07 PM, Fred Matthews [off-list ref] wrote:

quoted

Hi all,

Much appreciated Mohammed Shafi, disabling aggregation works when
replacing those lines with
 sc->sc_flags &= ~SC_OP_TXAGGR;
 sc->sc_flags &= ~SC_OP_RXAGGR;
only issue is I must always recompile compat-wireless to
re-enable/disable it again. I tried to port it using debugfs to be
able to enable and disable at runtime but ath9k kept rebooting.

I am not sure we can enable/disable aggregation via debugfs, these
flags seem are set at initialization and based on this lot of things
happen in tx-path, initializing TID structure, ampdu action, rate
control  etc.

The project I am on we have to deal with a situation where we change
the hardware environment and thus have to change some of the
initialization parameters. Obviously the environment is a little
unique, but I believe this should work for you. Before the changes are
made we do a

                if (sc->nvifs > 0) {
                        printk(KERN_INFO "ath9k:card has an open interface\n");
                        return -EBUSY;
                }

                pdev = to_pci_dev(sc->dev);
                pdev->driver->remove(pdev);

then we make our hardware change and  then do a

                pci_id = pci_match_id(pdev->driver->id_table, pdev);
                err = pdev->driver->probe(pdev, pci_id);

which gives us a re-initialized instance of ath_hw and ath_softc for the radio

Hope this helps!

V/r,
Daniel P. Smith