@northbound no , different kernel. You have to use trunk packages.

I am back to trunk and copied the new 88W8897.bin into /lib/firmware/mwlwifi and rebooted
But I am still at 10.3.0.3 Oh well guess I will wait for the next trunk build
Or do I just need /lib/modules/3.18.20/mwlwifi.ko for 10.3.0.8?

northbound wrote:

I am back to trunk and copied the new 88W8897.bin into /lib/firmware/mwlwifi and rebooted
But I am still at 10.3.0.3 Oh well guess I will wait for the next trunk build
Or do I just need /lib/modules/3.18.20/mwlwifi.ko for 10.3.0.8?

These routers don't use 88W8897.  That's for future devices. 88w8864 is the firmware for this platform.

nyt wrote:

northbound wrote:

I am back to trunk and copied the new 88W8897.bin into /lib/firmware/mwlwifi and rebooted
But I am still at 10.3.0.3 Oh well guess I will wait for the next trunk build
Or do I just need /lib/modules/3.18.20/mwlwifi.ko for 10.3.0.8?

These routers don't use 88W8897.  That's for future devices. 88w8864 is the firmware for this platform.

Since both were in Kaloz's build I put both back into the trunk build which already had both. I thought that it was for a different router and wondered why both were there.

Golden wrote:

it means this issue is resolved now in latest trunk ?

Hopefully Never hit me, but hit nyt

So it seems some chipsets are having issues. Because CC is almost done, I would like to ask everyone (specially with older devices) to give a shot to these images and report back within a day, so we cna decide which driver to go with.

slg wrote:

McWRT 2.0 has great wireless performance. 20-30% better than CC RC3.

@Chadster766 , could you add mwan3 package to your package repository ?

Since McWRT uses an older kernel (3.2.xx), it benefits from the hardware acceleration implemented by Marvell when the code was initially released BUT it lacks the security improvements / patches from current state of OpenWRT.

nitroshift

nyt wrote:

New driver definitely has issues.  2008 macbook on 5ghz or 2.4ghz was having very poor performance

and of course, bunches of this
[  409.130865] ieee80211 phy1: check ba result error
[  410.164653] ieee80211 phy1: check ba result error
[  411.199304] ieee80211 phy1: check ba result error
[  412.234463] ieee80211 phy1: check ba result error
[  413.263482] ieee80211 phy1: error code: -22
[  445.911823] ieee80211 phy1: check ba result error
[  447.935847] ieee80211 phy1: check ba result error
[  449.958488] ieee80211 phy1: check ba result error

So it seems some chipsets are having issues. Because CC is almost done, I would like to ask everyone (specially with older devices) to give a shot to these images and report back within a day, so we cna decide which driver to go with.

new driver update:

1. Correct wording for log messages.
2. Remove unnecessary code.
3. Change title to "Marvell 88W8864/88W8897 PCIe driver with AP support".
4. Remove "select OF".
5. Let the driver can still work even though CONFIG_OF is not defined.
6. Fix typo error.
7. Remove unused field.
8. Change type of cdd from u32 to bool.
9. Move request_irq from mwl_mac80211_start() to mwl_wl_init().
10. Move free_irq from mwl_mac80211_stop() to mwl_wl_deinit().
11. Modify "struct mwl_priv *priv; priv = hw->priv" to "struct mwl_priv *priv = \
hw->priv;" to reduce the length of code. 12. Change type of jiffies related variables \
from u32 to unsigned long. 13. Change wiphy_info() to wiphy_debug() for some debug \
messages. 14. Remove "#if SYSADPT_NUM_OF_DESC_DATA > 3 ...#endif".
15. Use correct PCI constant DMA_MASK_BIT(32).
16. Delete mac80211.h.
17. Move exported variable in mac80211.h to dev.h.
18. Check if macid is zero before decreasing it.
19. Modify some code segment to use switch instead of if..else.
20. Remove fields vif and is_sta from structure mwl_vif.
21. Remove field sta from structure mwl_sta.
22. Change MAX_WAIT_FW_COMPLETE_ITERATIONS from 10000 to 500.
23. Rewrite mwl_tx_init() and mwl_rx_init() according to suggestion from Linux \
wireless community. 24. Fix ccmp header endian problem.
25. Use pci_dma_mapping_error() to check mapping result of pci_map_single().
26. Correct PCI mapping constant.
27. Let code can pass the checking of checkpatch.pl.
28. Remove unnecessary code.
29. Change title of Kconfig from "Marvell 88W8864/88W8897 PCIe driver with AP \
support" to "Marvell Avastar 88W8864/88W8897 PCIe driver (mac80211 compatible)".

Kaloz wrote:

nyt wrote:

New driver definitely has issues.  2008 macbook on 5ghz or 2.4ghz was having very poor performance

and of course, bunches of this
[  409.130865] ieee80211 phy1: check ba result error
[  410.164653] ieee80211 phy1: check ba result error
[  411.199304] ieee80211 phy1: check ba result error
[  412.234463] ieee80211 phy1: check ba result error
[  413.263482] ieee80211 phy1: error code: -22
[  445.911823] ieee80211 phy1: check ba result error
[  447.935847] ieee80211 phy1: check ba result error
[  449.958488] ieee80211 phy1: check ba result error

So it seems some chipsets are having issues. Because CC is almost done, I would like to ask everyone (specially with older devices) to give a shot to these images and report back within a day, so we cna decide which driver to go with.

I don't have a MBP or any iPhone(s) on iOS 8.4.1, or iPad(s) on iOS 9 Public

Clients here

WRT54GL dd-wrt r21676
DIR-615-E  dd-wrt r14896
Realtek RTL8188CE Wireless LAN 802.11n PCI-E NIC
Asus USB-AC53 Dual-Band Wireless-AC1200 USB Adapter (5ghz) very good performance

RC3 wifi f/w

root@AC1900M:~# ls -l /lib/firmware/mwlwifi/88W8864.bin
-rwxr-xr-x    1 root     root        116208 Jun 22 05:23 /lib/firmware/mwlwifi/88W8864.bin
root@AC1900M:~#

https://github.com/kaloz/mwlwifi/commit … 0eff2b27c9

Work here

nyt wrote:

new driver update:

1. Correct wording for log messages.
2. Remove unnecessary code.
3. Change title to "Marvell 88W8864/88W8897 PCIe driver with AP support".
4. Remove "select OF".
5. Let the driver can still work even though CONFIG_OF is not defined.
6. Fix typo error.
7. Remove unused field.
8. Change type of cdd from u32 to bool.
9. Move request_irq from mwl_mac80211_start() to mwl_wl_init().
10. Move free_irq from mwl_mac80211_stop() to mwl_wl_deinit().
11. Modify "struct mwl_priv *priv; priv = hw->priv" to "struct mwl_priv *priv = \
hw->priv;" to reduce the length of code. 12. Change type of jiffies related variables \
from u32 to unsigned long. 13. Change wiphy_info() to wiphy_debug() for some debug \
messages. 14. Remove "#if SYSADPT_NUM_OF_DESC_DATA > 3 ...#endif".
15. Use correct PCI constant DMA_MASK_BIT(32).
16. Delete mac80211.h.
17. Move exported variable in mac80211.h to dev.h.
18. Check if macid is zero before decreasing it.
19. Modify some code segment to use switch instead of if..else.
20. Remove fields vif and is_sta from structure mwl_vif.
21. Remove field sta from structure mwl_sta.
22. Change MAX_WAIT_FW_COMPLETE_ITERATIONS from 10000 to 500.
23. Rewrite mwl_tx_init() and mwl_rx_init() according to suggestion from Linux \
wireless community. 24. Fix ccmp header endian problem.
25. Use pci_dma_mapping_error() to check mapping result of pci_map_single().
26. Correct PCI mapping constant.
27. Let code can pass the checking of checkpatch.pl.
28. Remove unnecessary code.
29. Change title of Kconfig from "Marvell 88W8864/88W8897 PCIe driver with AP \
support" to "Marvell Avastar 88W8864/88W8897 PCIe driver (mac80211 compatible)".

Let's not get too cute here.

Line 29 SPECIFIES the latest released driver is for the "Marvell Avastar 88W8864/88W8897 chipset.

If people are going to post here; it would be great IF they read ALL of the posts for at least 5 pages; BACK.

IF I read this correctly; the new driver from several versions back WILL work with the 88W8864 AND, AND, AND the 88W8897.

Did Marvel screw up; did I misread this OR are PEOPLE only reading from the date they logged in; NOT FROM when they last logged in to get UP TO DATE; repeat NOT FROM when they last logged in to get UP TO DATE.

Can WE stop running a mile a minute AND read OR READ 10 OR 20 back pages to get the context of where WE ARE!

Rick

nyt wrote:

new driver update:

1. Correct wording for log messages.
2. Remove unnecessary code.
3. Change title to "Marvell 88W8864/88W8897 PCIe driver with AP support".
4. Remove "select OF".
5. Let the driver can still work even though CONFIG_OF is not defined.
6. Fix typo error.
7. Remove unused field.
8. Change type of cdd from u32 to bool.
9. Move request_irq from mwl_mac80211_start() to mwl_wl_init().
10. Move free_irq from mwl_mac80211_stop() to mwl_wl_deinit().
11. Modify "struct mwl_priv *priv; priv = hw->priv" to "struct mwl_priv *priv = \
hw->priv;" to reduce the length of code. 12. Change type of jiffies related variables \
from u32 to unsigned long. 13. Change wiphy_info() to wiphy_debug() for some debug \
messages. 14. Remove "#if SYSADPT_NUM_OF_DESC_DATA > 3 ...#endif".
15. Use correct PCI constant DMA_MASK_BIT(32).
16. Delete mac80211.h.
17. Move exported variable in mac80211.h to dev.h.
18. Check if macid is zero before decreasing it.
19. Modify some code segment to use switch instead of if..else.
20. Remove fields vif and is_sta from structure mwl_vif.
21. Remove field sta from structure mwl_sta.
22. Change MAX_WAIT_FW_COMPLETE_ITERATIONS from 10000 to 500.
23. Rewrite mwl_tx_init() and mwl_rx_init() according to suggestion from Linux \
wireless community. 24. Fix ccmp header endian problem.
25. Use pci_dma_mapping_error() to check mapping result of pci_map_single().
26. Correct PCI mapping constant.
27. Let code can pass the checking of checkpatch.pl.
28. Remove unnecessary code.
29. Change title of Kconfig from "Marvell 88W8864/88W8897 PCIe driver with AP \
support" to "Marvell Avastar 88W8864/88W8897 PCIe driver (mac80211 compatible)".

Let's not get too cute here.

Line 29 SPECIFIES the latest released driver is for the "Marvell Avastar 88W8864/88W8897 chipset.

If people are going to post here; it would be great IF they read ALL of the posts for at least 5 pages; BACK.

IF I read this correctly; the new driver from several versions back WILL work with the 88W8864 AND, AND, AND the 88W8897.

Did Marvel screw up; did I misread this OR are PEOPLE only reading from the date they logged in; NOT FROM when they last logged in to get UP TO DATE; repeat NOT FROM when they last logged in to get UP TO DATE.

Can WE stop running a mile a minute AND read OR READ 10 OR 20 back pages to get the context of where WE ARE!

Rick

nyt wrote:

new driver update:

1. Correct wording for log messages.
2. Remove unnecessary code.
3. Change title to "Marvell 88W8864/88W8897 PCIe driver with AP support".
4. Remove "select OF".
5. Let the driver can still work even though CONFIG_OF is not defined.
6. Fix typo error.
7. Remove unused field.
8. Change type of cdd from u32 to bool.
9. Move request_irq from mwl_mac80211_start() to mwl_wl_init().
10. Move free_irq from mwl_mac80211_stop() to mwl_wl_deinit().
11. Modify "struct mwl_priv *priv; priv = hw->priv" to "struct mwl_priv *priv = \
hw->priv;" to reduce the length of code. 12. Change type of jiffies related variables \
from u32 to unsigned long. 13. Change wiphy_info() to wiphy_debug() for some debug \
messages. 14. Remove "#if SYSADPT_NUM_OF_DESC_DATA > 3 ...#endif".
15. Use correct PCI constant DMA_MASK_BIT(32).
16. Delete mac80211.h.
17. Move exported variable in mac80211.h to dev.h.
18. Check if macid is zero before decreasing it.
19. Modify some code segment to use switch instead of if..else.
20. Remove fields vif and is_sta from structure mwl_vif.
21. Remove field sta from structure mwl_sta.
22. Change MAX_WAIT_FW_COMPLETE_ITERATIONS from 10000 to 500.
23. Rewrite mwl_tx_init() and mwl_rx_init() according to suggestion from Linux \
wireless community. 24. Fix ccmp header endian problem.
25. Use pci_dma_mapping_error() to check mapping result of pci_map_single().
26. Correct PCI mapping constant.
27. Let code can pass the checking of checkpatch.pl.
28. Remove unnecessary code.
29. Change title of Kconfig from "Marvell 88W8864/88W8897 PCIe driver with AP \
support" to "Marvell Avastar 88W8864/88W8897 PCIe driver (mac80211 compatible)".

Let's not get too cute here.

Line 29 SPECIFIES the latest released driver is for the "Marvell Avastar 88W8864/88W8897 chipset.

If people are going to post here; it would be great IF they read ALL of the posts for at least 5 pages; BACK.

IF I read this correctly; the new driver from several versions back WILL work with the 88W8864 AND, AND, AND the 88W8897.

Did Marvel screw up; did I misread this OR are PEOPLE only reading from the date they logged in; NOT FROM when they last logged in to get UP TO DATE; repeat NOT FROM when they last logged in to get UP TO DATE.

Can WE stop running a mile a minute AND read OR READ 10 OR 20 back pages to get the context of where WE ARE!

Rick

gufus wrote:

Kaloz wrote:

So it seems some chipsets are having issues. Because CC is almost done, I would like to ask everyone (specially with older devices) to give a shot to these images and report back within a day, so we cna decide which driver to go with.

I don't have a MBP or any iPhone(s) on iOS 8.4.1, or iPad(s) on iOS 9 Public

Clients here

WRT54GL dd-wrt r21676
DIR-615-E  dd-wrt r14896
Realtek RTL8188CE Wireless LAN 802.11n PCI-E NIC
Asus USB-AC53 Dual-Band Wireless-AC1200 USB Adapter (5ghz) very good performance

RC3 wifi f/w

root@AC1900M:~# ls -l /lib/firmware/mwlwifi/88W8864.bin
-rwxr-xr-x    1 root     root        116208 Jun 22 05:23 /lib/firmware/mwlwifi/88W8864.bin
root@AC1900M:~#

https://github.com/kaloz/mwlwifi/commit … 0eff2b27c9

Work here

https://github.com/kaloz/mwlwifi/commit … 0653b5ae58 is the trouble version.  need to test that.  Kaloz posted builds using it.

Simply; IF wrt1900ac firmware is loaded on an "1200" IT will become a 1900!

Rick

RickStep wrote:

Simply; IF wrt1900ac firmware is loaded on an "1200" IT will become a 1900!

Rick

Only one way to find out

RickStep wrote:

nyt wrote:

new driver update:

Let's not get too cute here.

Line 29 SPECIFIES the latest released driver is for the "Marvell Avastar 88W8864/88W8897 chipset.

If people are going to post here; it would be great IF they read ALL of the posts for at least 5 pages; BACK.

IF I read this correctly; the new driver from several versions back WILL work with the 88W8864 AND, AND, AND the 88W8897.

Did Marvel screw up; did I misread this OR are PEOPLE only reading from the date they logged in; NOT FROM when they last logged in to get UP TO DATE; repeat NOT FROM when they last logged in to get UP TO DATE.

Can WE stop running a mile a minute AND read OR READ 10 OR 20 back pages to get the context of where WE ARE!

Rick

This brings up an interesting point.

ALL devices using  "Marvell Avastar 88W8864/88W8897 chipset" should be able to use the same firmware.

ALL memory addresses should be compatible since the  "Marvell Avastar 88W8864/88W8897 chipset" CAN'T wander TOO far from their design requirements.

Am I missing something here?

The only thing that comes to mind is that Marvell is lying OR open wrt IS NOT trying to pair 3 devices; 1900v1, 1900v2 & 1200v***.

It seems to me that Belkin is still being ABSURD and does NOT WANT to explain the difference among the 3 above.

I am beginning to think that the difference between the 1900 and 1200 IS ONLT FIRMWARE.

Simply; IF wrt1900ac firmware is loaded on an "1200" IT will become a 1900!

Rick

RickStep wrote:

RickStep wrote:

Let's not get too cute here.

Line 29 SPECIFIES the latest released driver is for the "Marvell Avastar 88W8864/88W8897 chipset.

If people are going to post here; it would be great IF they read ALL of the posts for at least 5 pages; BACK.

IF I read this correctly; the new driver from several versions back WILL work with the 88W8864 AND, AND, AND the 88W8897.

Did Marvel screw up; did I misread this OR are PEOPLE only reading from the date they logged in; NOT FROM when they last logged in to get UP TO DATE; repeat NOT FROM when they last logged in to get UP TO DATE.

Can WE stop running a mile a minute AND read OR READ 10 OR 20 back pages to get the context of where WE ARE!

Rick

This brings up an interesting point.

ALL devices using  "Marvell Avastar 88W8864/88W8897 chipset" should be able to use the same firmware.

ALL memory addresses should be compatible since the  "Marvell Avastar 88W8864/88W8897 chipset" CAN'T wander TOO far from their design requirements.

Am I missing something here?

The only thing that comes to mind is that Marvell is lying OR open wrt IS NOT trying to pair 3 devices; 1900v1, 1900v2 & 1200v***.

It seems to me that Belkin is still being ABSURD and does NOT WANT to explain the difference among the 3 above.

I am beginning to think that the difference between the 1900 and 1200 IS ONLT FIRMWARE.

Simply; IF wrt1900ac firmware is loaded on an "1200" IT will become a 1900!

Rick

There is one other possibility.

This has been used in the past; not used much in todays environment.

The issue is this.

Belkin has a number of wrt1900ac devices that WON'T pass hardware certification for SPEED etc.

The solution to sell these degraded boxes as "1200" devices; rejig the firmware for the "1200" series and go forward.

It seems to me that ALL  "Marvell Avastar 88W8864/88W8897 chipset SHOULD use the same FIRMWARE.

Rick

RickStep wrote:

RickStep wrote:

This brings up an interesting point.

ALL devices using  "Marvell Avastar 88W8864/88W8897 chipset" should be able to use the same firmware.

ALL memory addresses should be compatible since the  "Marvell Avastar 88W8864/88W8897 chipset" CAN'T wander TOO far from their design requirements.

Am I missing something here?

The only thing that comes to mind is that Marvell is lying OR open wrt IS NOT trying to pair 3 devices; 1900v1, 1900v2 & 1200v***.

It seems to me that Belkin is still being ABSURD and does NOT WANT to explain the difference among the 3 above.

I am beginning to think that the difference between the 1900 and 1200 IS ONLT FIRMWARE.

Simply; IF wrt1900ac firmware is loaded on an "1200" IT will become a 1900!

Rick

There is one other possibility.

This has been used in the past; not used much in todays environment.

The issue is this.

Belkin has a number of wrt1900ac devices that WON'T pass hardware certification for SPEED etc.

The solution to sell these degraded boxes as "1200" devices; rejig the firmware for the "1200" series and go forward.

It seems to me that ALL  "Marvell Avastar 88W8864/88W8897 chipset SHOULD use the same FIRMWARE.

Rick

One last point.

This board has been dragged through the mud by Belkin.

This post as of a few minutes ago has 7,118 replies AND 1,028,134 views.

The world has been monitoring the fact that this problem started 1 and a half years ago and the probability that a stable firmware will be available is ZERO, ZERO, ZERO, ...

Reliable firmware is ZERO YEARS away!!!!!!!!!!!!!!

Rick

RickStep wrote:

RickStep wrote:

There is one other possibility.

This has been used in the past; not used much in todays environment.

The issue is this.

Belkin has a number of wrt1900ac devices that WON'T pass hardware certification for SPEED etc.

The solution to sell these degraded boxes as "1200" devices; rejig the firmware for the "1200" series and go forward.

It seems to me that ALL  "Marvell Avastar 88W8864/88W8897 chipset SHOULD use the same FIRMWARE.

Rick

One last point.

This board has been dragged through the mud by Belkin.

This post as of a few minutes ago has 7,118 replies AND 1,028,134 views.

The world has been monitoring the fact that this problem started 1 and a half years ago and the probability that a stable firmware will be available is ZERO, ZERO, ZERO, ...

Reliable firmware is ZERO YEARS away!!!!!!!!!!!!!!

Rick

Myself...

I'm sticking with RC3

RC3 wifi f/w

root@AC1900M:~# ls -l /lib/firmware/mwlwifi/88W8864.bin
-rwxr-xr-x    1 root     root        116208 Jun 22 05:23 /lib/firmware/mwlwifi/88W8864.bin
root@AC1900M:~#

Until this gets sorted out

nyt wrote:

New driver definitely has issues.  2008 macbook on 5ghz or 2.4ghz was having very poor performance

and of course, bunches of this
[  409.130865] ieee80211 phy1: check ba result error
[  410.164653] ieee80211 phy1: check ba result error
[  411.199304] ieee80211 phy1: check ba result error
[  412.234463] ieee80211 phy1: check ba result error
[  413.263482] ieee80211 phy1: error code: -22
[  445.911823] ieee80211 phy1: check ba result error
[  447.935847] ieee80211 phy1: check ba result error
[  449.958488] ieee80211 phy1: check ba result error

So it seems some chipsets are having issues. Because CC is almost done, I would like to ask everyone (specially with older devices) to give a shot to these images and report back within a day, so we cna decide which driver to go with.

One last point.

This board has been dragged through the mud by Belkin.

This post as of a few minutes ago has 7,118 replies AND 1,028,134 views.

The world has been monitoring the fact that this problem started 1 and a half years ago and the probability that a stable firmware will be available is ZERO, ZERO, ZERO, ...

Reliable firmware is ZERO YEARS away!!!!!!!!!!!!!!

Rick

most people have had great success with the RC3 driver coupled with the Jul 27 88W8864.bin update, and given that it's so late in the process and some have definitely had problems with this very latest driver, we should stick with the RC3+Jul 27 88W8864.bin - plus the single commit that removes the excessive debug code output (assuming that commit doesn't cause problems!).