[opensuse-kernel] USB flash don't work with MSI 760GM-P21(FX) (MS-7641) & kernel ≥ 3.16
Hello! Sorry if off-topic. It isn't a bugreport. It is a request for help to write a bug report. Recently I've upgraded my PC from Biostar TF560 A2+ to MSI 760GM-P21(FX) (MS-7641) and found USB flash drives not working. `lsusb` output is just fine, `dmesg` repeats "usb 2-5: reset high-speed USB device number 3 using ehci-pci". The issue reproduces at openSUSE 13.2 (desktop-3.16), Leap 42.1 (vanilla- & default-4.1) & TW (default-4.7) but it can't be reproduced at openSUSE 13.1 (desktop-3.12) & Kubuntu 16.04.01 (4.4). STR: openSUSE, kernel ≥ 3.16, MSI 760GM-P21(FX) (MS-7641), any USB flash drive. Expected result: /dev/sdxN, mounting etc. Actual result: sometimes there are no /dev/sdxN only /dev/sdx is present, sometimes /dev/sdxN is present but can't be mounted with timeout error, sometimes it can be mounted but file operations (`ls`) take minutes. `dmidecode` output is attached. Anu help is welcomed! -- Best regards, Dmitriy Perlow R&D chemical engineer Lekpharm JLLC
On Fri, 2016-08-26 at 17:15 +0300, Dmitriy Perlow wrote:
Hello!
Sorry if off-topic. It isn't a bugreport. It is a request for help to write a bug report.
Recently I've upgraded my PC from Biostar TF560 A2+ to MSI 760GM-P21(FX) (MS-7641) and found USB flash drives not working. `lsusb` output is just fine, `dmesg` repeats "usb 2-5: reset high-speed USB device number 3 using ehci-pci". The issue reproduces at openSUSE 13.2 (desktop-3.16), Leap 42.1 (vanilla- & default-4.1) & TW (default-4.7) but it can't be reproduced at openSUSE 13.1 (desktop-3.12) & Kubuntu 16.04.01 (4.4).
STR: openSUSE, kernel ≥ 3.16, MSI 760GM-P21(FX) (MS-7641), any USB flash drive.
Expected result: /dev/sdxN, mounting etc.
Actual result: sometimes there are no /dev/sdxN only /dev/sdx is present, sometimes /dev/sdxN is present but can't be mounted with timeout error, sometimes it can be mounted but file operations (`ls`) take minutes.
`dmidecode` output is attached. Anu help is welcomed!
Hi, this needs to be bisected. It is a kernel regression. Are you familiar with the bisection process? Regards Oliver -- To unsubscribe, e-mail: opensuse-kernel+unsubscribe@opensuse.org To contact the owner, e-mail: opensuse-kernel+owner@opensuse.org
this needs to be bisected. It is a kernel regression. Are you familiar with the bisection process?
Yep, I've used `git bisect` several times for desktop applications. Are there any challenges with kernel? Could OBS be used somehow (kernel build at my PC takes several hours)? -- Best regards, Dmitriy Perlow R&D chemical engineer Lekpharm JLLC -- To unsubscribe, e-mail: opensuse-kernel+unsubscribe@opensuse.org To contact the owner, e-mail: opensuse-kernel+owner@opensuse.org
On Mon, 2016-08-29 at 19:42 +0300, Dmitriy Perlow wrote:
this needs to be bisected. It is a kernel regression. Are you familiar with the bisection process?
Yep, I've used `git bisect` several times for desktop applications. Are there any challenges with kernel? Could OBS be used somehow (kernel build at my PC takes several hours)?
I don't know. Please ask on the list. The question is of general interest. Sorry Oliver -- To unsubscribe, e-mail: opensuse-kernel+unsubscribe@opensuse.org To contact the owner, e-mail: opensuse-kernel+owner@opensuse.org
On 29.08.2016 18:42, Dmitriy Perlow wrote:
this needs to be bisected. It is a kernel regression. Are you familiar with the bisection process?
Yep, I've used `git bisect` several times for desktop applications. Are there any challenges with kernel?
Not really. Maybe the changing kernel config options betweeen kernel versions that might cause questions to be answered before compilation. Kernel bisection is quite straightforward, especially if you have a problem that reliably reproduces. I have done it a few times for straightforward suspend failures or "display stays blank at boot after loading drm drivers" or such. Pro tip: change CONFIG_LOCALVERSION to something like "-mytestkernel"
Could OBS be used somehow (kernel build at my PC takes several hours)?
OBS can not really be used for this IMHO (or it would take considerable effort to implement this). And OBS also takes hours to build a kernel RPM, so you won't gain much after all. I use ccache for kernel bisection and in general, the nearer you come to the end of bisection, the faster it compiles (less changed files). -- Stefan Seyfried "For a successful technology, reality must take precedence over public relations, for nature cannot be fooled." -- Richard Feynman -- To unsubscribe, e-mail: opensuse-kernel+unsubscribe@opensuse.org To contact the owner, e-mail: opensuse-kernel+owner@opensuse.org
On Tue 2016-08-30 17:35:47, Stefan Seyfried wrote:
On 29.08.2016 18:42, Dmitriy Perlow wrote:
this needs to be bisected. It is a kernel regression. Are you familiar with the bisection process?
Yep, I've used `git bisect` several times for desktop applications. Are there any challenges with kernel?
Not really. Maybe the changing kernel config options betweeen kernel versions that might cause questions to be answered before compilation.
Kernel bisection is quite straightforward, especially if you have a problem that reliably reproduces. I have done it a few times for straightforward suspend failures or "display stays blank at boot after loading drm drivers" or such.
Pro tip: change CONFIG_LOCALVERSION to something like "-mytestkernel"
Could OBS be used somehow (kernel build at my PC takes several hours)?
OBS can not really be used for this IMHO (or it would take considerable effort to implement this). And OBS also takes hours to build a kernel RPM, so you won't gain much after all. I use ccache for kernel bisection and in general, the nearer you come to the end of bisection, the faster it compiles (less changed files).
Just for completeness. There is the friendly "binrpm-pkg" makefile target that will produce installable RPMs without using the SUSE's spec file. This way, you could simply get RPMs from an incremental build. Also if you have more CPUs, make them busy. I personally use: make -j15 binrpm-pkg Best Regards, Petr -- To unsubscribe, e-mail: opensuse-kernel+unsubscribe@opensuse.org To contact the owner, e-mail: opensuse-kernel+owner@opensuse.org
On mar., 2016-08-30 at 17:35 +0200, Stefan Seyfried wrote:
On 29.08.2016 18:42, Dmitriy Perlow wrote:
Could OBS be used somehow (kernel build at my PC takes several hours)?
OBS can not really be used for this IMHO (or it would take considerable effort to implement this). And OBS also takes hours to build a kernel RPM, so you won't gain much after all. I use ccache for kernel bisection and in general, the nearer you come to the end of bisection, the faster it compiles (less changed files).
For a kernel bisection, it can also be beneficial to first disable all the device drivers you don't need. It will take you a few minutes before starting the first build, but will save you a lot more than that for each subsequent build. Everything that can be disabled at the menu level is very profitable to disable: MTD, Parallel Port, NVMe, RAID/LVM, SPI, Hardware Monitoring, Watchdog Timer, Multimedia, Sound, Ultra Wideband, MMC, LED, EDAC, Userspace I/O, Virtualization, Staging Drivers, PWM, Thunderbolt, NVDIMM... As long as you know your system boots without them and it's not what you are bisecting, of course. -- Jean Delvare SUSE L3 Support -- To unsubscribe, e-mail: opensuse-kernel+unsubscribe@opensuse.org To contact the owner, e-mail: opensuse-kernel+owner@opensuse.org
On 2016-09-07 11:11, Jean Delvare wrote:
For a kernel bisection, it can also be beneficial to first disable all the device drivers you don't need. It will take you a few minutes before starting the first build, but will save you a lot more than that for each subsequent build. Everything that can be disabled at the menu level is very profitable to disable: MTD, Parallel Port, NVMe, RAID/LVM, SPI, Hardware Monitoring, Watchdog Timer, Multimedia, Sound, Ultra Wideband, MMC, LED, EDAC, Userspace I/O, Virtualization, Staging Drivers, PWM, Thunderbolt, NVDIMM... As long as you know your system boots without them and it's not what you are bisecting, of course.
You can also try make localmodconfig, which does this for you. Michal -- To unsubscribe, e-mail: opensuse-kernel+unsubscribe@opensuse.org To contact the owner, e-mail: opensuse-kernel+owner@opensuse.org
On Wed, 7 Sep 2016 11:39:06 +0200, Michal Marek wrote:
On 2016-09-07 11:11, Jean Delvare wrote:
For a kernel bisection, it can also be beneficial to first disable all the device drivers you don't need. It will take you a few minutes before starting the first build, but will save you a lot more than that for each subsequent build. Everything that can be disabled at the menu level is very profitable to disable: MTD, Parallel Port, NVMe, RAID/LVM, SPI, Hardware Monitoring, Watchdog Timer, Multimedia, Sound, Ultra Wideband, MMC, LED, EDAC, Userspace I/O, Virtualization, Staging Drivers, PWM, Thunderbolt, NVDIMM... As long as you know your system boots without them and it's not what you are bisecting, of course.
You can also try make localmodconfig, which does this for you.
Doh, you'd think I would know that... Thanks Michal! -- Jean Delvare SUSE L3 Support -- To unsubscribe, e-mail: opensuse-kernel+unsubscribe@opensuse.org To contact the owner, e-mail: opensuse-kernel+owner@opensuse.org
Oliver Neukum <oneukum@suse.com> Mon, 29 Aug 2016 10:09:30 +0300:
this needs to be bisected. It is a kernel regression. Are you familiar with the bisection process?
Using home projects by Takashi Iwai (thank you very much!) I've found that issue had been introduced between 3.14.6 (good) & 3.15.8 (bad). 3.15 RC6 from https://build.opensuse.org/project/show/home:colgur:kernel-mainline is good. -- Best regards, Dmitriy Perlow R&D chemical engineer Lekpharm JLLC -- To unsubscribe, e-mail: opensuse-kernel+unsubscribe@opensuse.org To contact the owner, e-mail: opensuse-kernel+owner@opensuse.org
On Mon, 2016-08-29 at 22:55 +0300, Dmitriy Perlow wrote:
Oliver Neukum <oneukum@suse.com> Mon, 29 Aug 2016 10:09:30 +0300:
this needs to be bisected. It is a kernel regression. Are you familiar with the bisection process?
Using home projects by Takashi Iwai (thank you very much!) I've found that issue had been introduced between 3.14.6 (good) & 3.15.8 (bad).
3.15 RC6 from https://build.opensuse.org/project/show/home:colgur:kernel-mainline is good.
Hi, this is very odd, because there are no changes to the storage driver between v3.15-rc6 and v3.15.0. Could you check v3.15 vanilla? Regards Oliver -- To unsubscribe, e-mail: opensuse-kernel+unsubscribe@opensuse.org To contact the owner, e-mail: opensuse-kernel+owner@opensuse.org
Oliver Neukum <oneukum@suse.com> Tue, 30 Aug 2016 11:26:50 +0300:
On Mon, 2016-08-29 at 22:55 +0300, Dmitriy Perlow wrote:
Oliver Neukum <oneukum@suse.com> Mon, 29 Aug 2016 10:09:30 +0300:
this needs to be bisected. It is a kernel regression. Are you familiar with the bisection process?
Using home projects by Takashi Iwai (thank you very much!) I've found that issue had been introduced between 3.14.6 (good) & 3.15.8 (bad).
3.15 RC6 from https://build.opensuse.org/project/show/home:colgur:kernel-mainline is good.
Hi,
this is very odd, because there are no changes to the storage driver between v3.15-rc6 and v3.15.0. Could you check v3.15 vanilla?
The same for vanilla packages: 3.15 RC6 works fine, 3.15.8 provides the issue. -- Best regards, Dmitriy Perlow R&D chemical engineer Lekpharm JLLC -- To unsubscribe, e-mail: opensuse-kernel+unsubscribe@opensuse.org To contact the owner, e-mail: opensuse-kernel+owner@opensuse.org
On Sat, 2016-09-03 at 16:41 +0300, Dmitriy Perlow wrote:
Oliver Neukum <oneukum@suse.com> Tue, 30 Aug 2016 11:26:50 +0300:
On Mon, 2016-08-29 at 22:55 +0300, Dmitriy Perlow wrote:
Oliver Neukum <oneukum@suse.com> Mon, 29 Aug 2016 10:09:30 +0300:
this needs to be bisected. It is a kernel regression. Are you familiar with the bisection process?
Using home projects by Takashi Iwai (thank you very much!) I've found that issue had been introduced between 3.14.6 (good) & 3.15.8 (bad).
3.15 RC6 from https://build.opensuse.org/project/show/home:colgur:kernel-mainline is good.
Hi,
this is very odd, because there are no changes to the storage driver between v3.15-rc6 and v3.15.0. Could you check v3.15 vanilla?
The same for vanilla packages: 3.15 RC6 works fine, 3.15.8 provides the issue.
Sorry, I meant v3.15.0 with the vanilla kernel. Regards Oliver -- To unsubscribe, e-mail: opensuse-kernel+unsubscribe@opensuse.org To contact the owner, e-mail: opensuse-kernel+owner@opensuse.org
participants (6)
-
Dmitriy Perlow -
Jean Delvare -
Michal Marek -
Oliver Neukum -
Petr Mladek -
Stefan Seyfried