On Fri, Sep 23, 2011 at 06:21:07PM +0800, 李朝明 wrote:
> Please set ips =0 and try again..
What does that mean?
I can trigger the grinding-to-a-halt reliably with "ips=0" - it only
takes a couple of hours of network traffic. Also, I don't want to try
the driver you sent me because the version in the kernel needs fixing
not some out-of-tree codebase.
So please clarify your request.
Thanks.
--
Regards/Gruss,
Boris.
On 09/28/2011 08:18 AM, Borislav Petkov wrote:
> On Fri, Sep 23, 2011 at 11:34:12AM -0500, Larry Finger wrote:
>> On 09/23/2011 05:33 AM, Borislav Petkov wrote:
>>> On Fri, Sep 23, 2011 at 06:21:07PM +0800, 李朝明 wrote:
>>>> Please set ips =0 and try again..
>>>
>>> What does that mean?
>>>
>>> I can trigger the grinding-to-a-halt reliably with "ips=0" - it only
>>> takes a couple of hours of network traffic. Also, I don't want to try
>>> the driver you sent me because the version in the kernel needs fixing
>>> not some out-of-tree codebase.
>>
>> I got a chance to review the rtl8192se part of the changes in that
>> 08/16/2011 version. Attached is a patch to update the kernel
>> version.
>>
>> A prerequisite is:
>>
>> commit da3ba88a9996cd64c6768bed5727e02da81e2c8d
>> Author: Larry Finger<[email protected]>
>> Date: Mon Sep 19 14:34:10 2011 -0500
>>
>> rtlwifi: Combine instances of RTL_HAL_IS_CCK_RATE macros.
>>
>> Three drivers, rtl8192ce, rtl8192cu and rtl8192de, use the same macro
>> to check if a particular rate is in the CCK set. This common code is
>> relocated to a common header file. A distinct macro used by rtl8192se
>> with the same name is renamed.
>>
>> Signed-off-by: Larry Finger<[email protected]>
>> Signed-off-by: John W. Linville<[email protected]>
>
> Ok, here's what I did.
>
> * merge 'master' branch of
> git://git.infradead.org/users/linville/wireless-next.git with -rc8 in
> order to get da3ba88a9996.
>
> * apply your attached patch with the Realsil facelift:
>
> patching file drivers/net/wireless/rtlwifi/rtl8192se/hw.c
> patching file drivers/net/wireless/rtlwifi/rtl8192se/reg.h
> patching file drivers/net/wireless/rtlwifi/rtl8192se/sw.c
> patching file drivers/net/wireless/rtlwifi/rtl8192se/trx.c
> Hunk #6 FAILED at 540.
> 1 out of 6 hunks FAILED -- saving rejects to file drivers/net/wireless/rtlwifi/rtl8192se/trx.c.rej
> patching file drivers/net/wireless/rtlwifi/wifi.h
> Hunk #1 succeeded at 1328 (offset 3 lines).
>
> (had to apply hunk #6 by hand though)
>
> * build
>
> drivers/net/wireless/rtlwifi/rtl8192se/sw.c:307:8: error: `EFUSE_OOB_PROTECT_BYTES_LEN' undeclared here (not in a function)
> drivers/net/wireless/rtlwifi/rtl8192se/sw.c:307:2: error: array index in initializer not of integer type
> drivers/net/wireless/rtlwifi/rtl8192se/sw.c:307:2: error: (near initialization for `rtl92se_hal_cfg.maps')
> make[5]: *** [drivers/net/wireless/rtlwifi/rtl8192se/sw.o] Error 1
> make[4]: *** [drivers/net/wireless/rtlwifi/rtl8192se] Error 2
> make[3]: *** [drivers/net/wireless/rtlwifi] Error 2
> make[2]: *** [drivers/net/wireless] Error 2
> make[1]: *** [drivers/net] Error 2
> make: *** [drivers] Error 2
> make: *** Waiting for unfinished jobs....
>
> fix with
>
> diff --git a/drivers/net/wireless/rtlwifi/wifi.h b/drivers/net/wireless/rtlwifi/wifi.h
> index 615f6b4..55428f2 100644
> --- a/drivers/net/wireless/rtlwifi/wifi.h
> +++ b/drivers/net/wireless/rtlwifi/wifi.h
> @@ -450,6 +450,7 @@ enum rtl_var_map {
> EFUSE_HWSET_MAX_SIZE,
> EFUSE_MAX_SECTION_MAP,
> EFUSE_REAL_CONTENT_SIZE,
> + EFUSE_OOB_PROTECT_BYTES_LEN,
>
> /*CAM map */
> RWCAM,
>
>> I have run the new version for a couple of hours without problems.
>> Perhaps it will cure your difficulty, but I am not optimistic.
>
> Yeah, I'm sorry to confirm that your pessimism turned into realism :-).
> It froze in under 20 mins. Ran the driver with default module parameters
> though.
>
> Looks like I'll have to test the Realsil tarball after all. Question, do
> I simply overwrite the subtree under drivers/net/wireless/rtlwifi/ with
> the files from the tarball?
Asa you probably do not want to kill your standard tree, unpack the Realtek
version normally, cd to where it unpacked, and do a make. Once it builds, do the
following:
sudo modprobe -rv rtl8192se
sudo modprobe -v mac80211
sudo insmod rtlwifi.ko
sudo insmod rtl8192se/rtl8192se.ko
>
> Also, any preferred module parameters settings you want me to test?
Try it first with the default parameters. If that fails, use "ips=0".
Larry
On Fri, Sep 23, 2011 at 11:34:12AM -0500, Larry Finger wrote:
> On 09/23/2011 05:33 AM, Borislav Petkov wrote:
> >On Fri, Sep 23, 2011 at 06:21:07PM +0800, 李朝明 wrote:
> >>Please set ips =0 and try again..
> >
> >What does that mean?
> >
> >I can trigger the grinding-to-a-halt reliably with "ips=0" - it only
> >takes a couple of hours of network traffic. Also, I don't want to try
> >the driver you sent me because the version in the kernel needs fixing
> >not some out-of-tree codebase.
>
> I got a chance to review the rtl8192se part of the changes in that
> 08/16/2011 version. Attached is a patch to update the kernel
> version.
>
> A prerequisite is:
>
> commit da3ba88a9996cd64c6768bed5727e02da81e2c8d
> Author: Larry Finger <[email protected]>
> Date: Mon Sep 19 14:34:10 2011 -0500
>
> rtlwifi: Combine instances of RTL_HAL_IS_CCK_RATE macros.
>
> Three drivers, rtl8192ce, rtl8192cu and rtl8192de, use the same macro
> to check if a particular rate is in the CCK set. This common code is
> relocated to a common header file. A distinct macro used by rtl8192se
> with the same name is renamed.
>
> Signed-off-by: Larry Finger <[email protected]>
> Signed-off-by: John W. Linville <[email protected]>
Ok, here's what I did.
* merge 'master' branch of
git://git.infradead.org/users/linville/wireless-next.git with -rc8 in
order to get da3ba88a9996.
* apply your attached patch with the Realsil facelift:
patching file drivers/net/wireless/rtlwifi/rtl8192se/hw.c
patching file drivers/net/wireless/rtlwifi/rtl8192se/reg.h
patching file drivers/net/wireless/rtlwifi/rtl8192se/sw.c
patching file drivers/net/wireless/rtlwifi/rtl8192se/trx.c
Hunk #6 FAILED at 540.
1 out of 6 hunks FAILED -- saving rejects to file drivers/net/wireless/rtlwifi/rtl8192se/trx.c.rej
patching file drivers/net/wireless/rtlwifi/wifi.h
Hunk #1 succeeded at 1328 (offset 3 lines).
(had to apply hunk #6 by hand though)
* build
drivers/net/wireless/rtlwifi/rtl8192se/sw.c:307:8: error: `EFUSE_OOB_PROTECT_BYTES_LEN' undeclared here (not in a function)
drivers/net/wireless/rtlwifi/rtl8192se/sw.c:307:2: error: array index in initializer not of integer type
drivers/net/wireless/rtlwifi/rtl8192se/sw.c:307:2: error: (near initialization for `rtl92se_hal_cfg.maps')
make[5]: *** [drivers/net/wireless/rtlwifi/rtl8192se/sw.o] Error 1
make[4]: *** [drivers/net/wireless/rtlwifi/rtl8192se] Error 2
make[3]: *** [drivers/net/wireless/rtlwifi] Error 2
make[2]: *** [drivers/net/wireless] Error 2
make[1]: *** [drivers/net] Error 2
make: *** [drivers] Error 2
make: *** Waiting for unfinished jobs....
fix with
diff --git a/drivers/net/wireless/rtlwifi/wifi.h b/drivers/net/wireless/rtlwifi/wifi.h
index 615f6b4..55428f2 100644
--- a/drivers/net/wireless/rtlwifi/wifi.h
+++ b/drivers/net/wireless/rtlwifi/wifi.h
@@ -450,6 +450,7 @@ enum rtl_var_map {
EFUSE_HWSET_MAX_SIZE,
EFUSE_MAX_SECTION_MAP,
EFUSE_REAL_CONTENT_SIZE,
+ EFUSE_OOB_PROTECT_BYTES_LEN,
/*CAM map */
RWCAM,
> I have run the new version for a couple of hours without problems.
> Perhaps it will cure your difficulty, but I am not optimistic.
Yeah, I'm sorry to confirm that your pessimism turned into realism :-).
It froze in under 20 mins. Ran the driver with default module parameters
though.
Looks like I'll have to test the Realsil tarball after all. Question, do
I simply overwrite the subtree under drivers/net/wireless/rtlwifi/ with
the files from the tarball?
Also, any preferred module parameters settings you want me to test?
Thanks.
--
Regards/Gruss,
Boris.
On 09/23/2011 05:33 AM, Borislav Petkov wrote:
> On Fri, Sep 23, 2011 at 06:21:07PM +0800, 李朝明 wrote:
>> Please set ips =0 and try again..
>
> What does that mean?
>
> I can trigger the grinding-to-a-halt reliably with "ips=0" - it only
> takes a couple of hours of network traffic. Also, I don't want to try
> the driver you sent me because the version in the kernel needs fixing
> not some out-of-tree codebase.
I got a chance to review the rtl8192se part of the changes in that 08/16/2011
version. Attached is a patch to update the kernel version.
A prerequisite is:
commit da3ba88a9996cd64c6768bed5727e02da81e2c8d
Author: Larry Finger <[email protected]>
Date: Mon Sep 19 14:34:10 2011 -0500
rtlwifi: Combine instances of RTL_HAL_IS_CCK_RATE macros.
Three drivers, rtl8192ce, rtl8192cu and rtl8192de, use the same macro
to check if a particular rate is in the CCK set. This common code is
relocated to a common header file. A distinct macro used by rtl8192se
with the same name is renamed.
Signed-off-by: Larry Finger <[email protected]>
Signed-off-by: John W. Linville <[email protected]>
I have run the new version for a couple of hours without problems. Perhaps it
will cure your difficulty, but I am not optimistic.
Larry
On 09/23/2011 05:33 AM, Borislav Petkov wrote:
> On Fri, Sep 23, 2011 at 06:21:07PM +0800, 李朝明 wrote:
>> Please set ips =0 and try again..
>
> What does that mean?
>
> I can trigger the grinding-to-a-halt reliably with "ips=0" - it only
> takes a couple of hours of network traffic. Also, I don't want to try
> the driver you sent me because the version in the kernel needs fixing
> not some out-of-tree codebase.
Does the likelihood of the failure change when "ips=0" is used?
The Realtek group made several changes in the driver that Chaoming sent you that
have not yet been incorporated in the kernel version. If you test that driver,
we might learn if any of them are important to your problem. As neither of us
can duplicate your results, it is not possible for us to do those tests.
I agree that we want to fix the kernel version. It is unfortunate that Realtek
does not generate their improvements as patches to that kernel version, and
publish them that way, but that is a fact of life. When they produce a new
version, I have to look at the diff file between it and the previous version and
test those differences with my devices. Thus far, there have been no changes
that have any effect on my system, but who knows on yours. Please run the test
as Chaoming asked you to do.
Thanks,
Larry
On 10/05/2011 10:15 AM, Borislav Petkov wrote:
> Ok, I can cautiosly say now that after a couple days of running the
> Realtek version that the box runs just fine, no hiccups whatsoever.
>
> Larry, you said in an earlier mail that you've gone through the
> rtl8192se changes and weren't optimistic with the attached diff you sent
> me. What about the rtlwifi changes, is there something in Realtek's
> version which is missing upstream that would cause the sluggishness?
I have been going though the differences between the 06/20/2011 and 08/16/2011
drivers and making those changes to the kernel drivers. Could you please apply
the 5 attached patches to the wireless-testing tree and see if your sluggishness
is fixed?
Thanks,
Larry
On Wed, Sep 28, 2011 at 08:57:38PM -0500, Larry Finger wrote:
> Asa you probably do not want to kill your standard tree, unpack the
> Realtek version normally, cd to where it unpacked, and do a make.
> Once it builds, do the following:
>
> sudo modprobe -rv rtl8192se
> sudo modprobe -v mac80211
> sudo insmod rtlwifi.ko
> sudo insmod rtl8192se/rtl8192se.ko
>
> >
> >Also, any preferred module parameters settings you want me to test?
>
> Try it first with the default parameters. If that fails, use "ips=0".
Ok, I can cautiosly say now that after a couple days of running the
Realtek version that the box runs just fine, no hiccups whatsoever.
Larry, you said in an earlier mail that you've gone through the
rtl8192se changes and weren't optimistic with the attached diff you sent
me. What about the rtlwifi changes, is there something in Realtek's
version which is missing upstream that would cause the sluggishness?
Thanks.
--
Regards/Gruss,
Boris.
On Wed, Oct 05, 2011 at 09:37:25PM -0500, Larry Finger wrote:
> I have been going though the differences between the 06/20/2011 and
> 08/16/2011 drivers and making those changes to the kernel drivers.
> Could you please apply the 5 attached patches to the
> wireless-testing tree and see if your sluggishness is fixed?
Just to document it here:
3.1 still has the problem, if you have something ready ported to 3.1
from the Realsil driver version, let me know and I'll give it a run.
Thanks.
--
Regards/Gruss,
Boris