Re: Regression in Linux next again
Hi, * Maciej Purski [180604 14:02]: > Tony, please apply this patchset and test it on your Beaglebone. It'd be > great if you could try to find out, which patch causes failure. They should > be appliable on the current next. It seems beagle-x15 boots for me except with the last patch in the series with compile issue fixed. With that applied after modules get loaded it just hangs with: [ 25.925607] regulator_resolve_supply: 1608 [ 25.930620] regulator_resolve_supply: 1608 [ 26.098449] lib80211: common routines for IEEE802.11 drivers [ 26.446713] cfg80211: Loading compiled-in X.509 certificates for regulatory database [ 26.514834] cfg80211: Loaded X.509 cert 'sforshee: 00b28ddf47aef9cea7' [ 26.525311] platform regulatory.0: Direct firmware load for regulatory.db failed with error -2 [ 26.534250] cfg80211: failed to load regulatory.db [ 26.998736] regulator_set_voltage: 3381 [ 27.007662] _regulator_do_set_voltage: 2913 [ 27.012593] regulator_set_voltage: 3381 [ 27.016969] _regulator_do_set_voltage: 2913 [ 27.022847] regulator_set_voltage: 3381 Not sure what module gets loaded at that point, I'll test again with initcall_debug once you post a compiling version of your last patch. Regards, Tony
Re: Regression in Linux next again
Hi, * Maciej Purski [180604 14:02]: > Tony, please apply this patchset and test it on your Beaglebone. It'd be > great if you could try to find out, which patch causes failure. They should > be appliable on the current next. It seems beagle-x15 boots for me except with the last patch in the series with compile issue fixed. With that applied after modules get loaded it just hangs with: [ 25.925607] regulator_resolve_supply: 1608 [ 25.930620] regulator_resolve_supply: 1608 [ 26.098449] lib80211: common routines for IEEE802.11 drivers [ 26.446713] cfg80211: Loading compiled-in X.509 certificates for regulatory database [ 26.514834] cfg80211: Loaded X.509 cert 'sforshee: 00b28ddf47aef9cea7' [ 26.525311] platform regulatory.0: Direct firmware load for regulatory.db failed with error -2 [ 26.534250] cfg80211: failed to load regulatory.db [ 26.998736] regulator_set_voltage: 3381 [ 27.007662] _regulator_do_set_voltage: 2913 [ 27.012593] regulator_set_voltage: 3381 [ 27.016969] _regulator_do_set_voltage: 2913 [ 27.022847] regulator_set_voltage: 3381 Not sure what module gets loaded at that point, I'll test again with initcall_debug once you post a compiling version of your last patch. Regards, Tony
Re: Regression in Linux next again
Hi, this patchset contains previous two reverted patches. They were split in order to make it easier to bisect, where the problem lies. It adds also some simple debugs, which should help us track down the problem. Tony, please apply this patchset and test it on your Beaglebone. It'd be great if you could try to find out, which patch causes failure. They should be appliable on the current next. Thanks, Maciej Purski Maciej Purski (7): regulator: core: Add debug messages regulator: core: Add regulator_set_voltage_rdev() regulator: core: Use re-entrant locks regulator: core: Implement voltage balancing algorithm regulator: core: Lock dependent regulators regulator: core: Lock dependent regulators on regulator_enable() regulator: core: Enable voltage balancing drivers/regulator/core.c | 341 +-- 1 file changed, 300 insertions(+), 41 deletions(-) -- 2.7.4
Re: Regression in Linux next again
Hi, this patchset contains previous two reverted patches. They were split in order to make it easier to bisect, where the problem lies. It adds also some simple debugs, which should help us track down the problem. Tony, please apply this patchset and test it on your Beaglebone. It'd be great if you could try to find out, which patch causes failure. They should be appliable on the current next. Thanks, Maciej Purski Maciej Purski (7): regulator: core: Add debug messages regulator: core: Add regulator_set_voltage_rdev() regulator: core: Use re-entrant locks regulator: core: Implement voltage balancing algorithm regulator: core: Lock dependent regulators regulator: core: Lock dependent regulators on regulator_enable() regulator: core: Enable voltage balancing drivers/regulator/core.c | 341 +-- 1 file changed, 300 insertions(+), 41 deletions(-) -- 2.7.4
Re: Regression in Linux next again
On 30 May 2018 at 20:23, Naresh Kamboju wrote: > On 30 May 2018 at 20:03, Mark Brown wrote: >> On Wed, May 30, 2018 at 04:03:15PM +0200, Maciej Purski wrote: >> >>> I'm afraid, I have no idea, how to fix it quickly. You can revert it and >>> in the next version I'll fix the build error and split the last patch even >>> more, so we could perform a more precise bisection. I'd be grateful if you >>> could push it on your test coupled branch and Tony could test it again >>> before >>> merging it with next again. >> >> Yeah, if we could get testing from Tony first that'd be ideal. > > Linux next 4.17.0-rc7-next-20180529 boot failed on x15 device. > Manually reproduced boot failed problem on x15 device. > The qemu_arm32 boots successfully. > > Ref bug: > LKFT: linux-next: Boot failed on beagle board x15 > https://bugs.linaro.org/show_bug.cgi?id=3863 This bug still happening on 4.17.0-rc7-next-20180530 on beagle board x15. Linux version 4.17.0-rc7-next-20180530 (buildslave@x86-64-07) (gcc version 6.2.1 20161016 (Linaro GCC 6.2-2016.11)) #1 SMP Wed May 30 12:38:23 UTC 2018 - Naresh
Re: Regression in Linux next again
On 30 May 2018 at 20:23, Naresh Kamboju wrote: > On 30 May 2018 at 20:03, Mark Brown wrote: >> On Wed, May 30, 2018 at 04:03:15PM +0200, Maciej Purski wrote: >> >>> I'm afraid, I have no idea, how to fix it quickly. You can revert it and >>> in the next version I'll fix the build error and split the last patch even >>> more, so we could perform a more precise bisection. I'd be grateful if you >>> could push it on your test coupled branch and Tony could test it again >>> before >>> merging it with next again. >> >> Yeah, if we could get testing from Tony first that'd be ideal. > > Linux next 4.17.0-rc7-next-20180529 boot failed on x15 device. > Manually reproduced boot failed problem on x15 device. > The qemu_arm32 boots successfully. > > Ref bug: > LKFT: linux-next: Boot failed on beagle board x15 > https://bugs.linaro.org/show_bug.cgi?id=3863 This bug still happening on 4.17.0-rc7-next-20180530 on beagle board x15. Linux version 4.17.0-rc7-next-20180530 (buildslave@x86-64-07) (gcc version 6.2.1 20161016 (Linaro GCC 6.2-2016.11)) #1 SMP Wed May 30 12:38:23 UTC 2018 - Naresh
Re: Regression in Linux next again
On 30 May 2018 at 20:03, Mark Brown wrote: > On Wed, May 30, 2018 at 04:03:15PM +0200, Maciej Purski wrote: > >> I'm afraid, I have no idea, how to fix it quickly. You can revert it and >> in the next version I'll fix the build error and split the last patch even >> more, so we could perform a more precise bisection. I'd be grateful if you >> could push it on your test coupled branch and Tony could test it again before >> merging it with next again. > > Yeah, if we could get testing from Tony first that'd be ideal. Linux next 4.17.0-rc7-next-20180529 boot failed on x15 device. Manually reproduced boot failed problem on x15 device. The qemu_arm32 boots successfully. Ref bug: LKFT: linux-next: Boot failed on beagle board x15 https://bugs.linaro.org/show_bug.cgi?id=3863
Re: Regression in Linux next again
On 30 May 2018 at 20:03, Mark Brown wrote: > On Wed, May 30, 2018 at 04:03:15PM +0200, Maciej Purski wrote: > >> I'm afraid, I have no idea, how to fix it quickly. You can revert it and >> in the next version I'll fix the build error and split the last patch even >> more, so we could perform a more precise bisection. I'd be grateful if you >> could push it on your test coupled branch and Tony could test it again before >> merging it with next again. > > Yeah, if we could get testing from Tony first that'd be ideal. Linux next 4.17.0-rc7-next-20180529 boot failed on x15 device. Manually reproduced boot failed problem on x15 device. The qemu_arm32 boots successfully. Ref bug: LKFT: linux-next: Boot failed on beagle board x15 https://bugs.linaro.org/show_bug.cgi?id=3863
Re: Regression in Linux next again
* Mark Brown [180530 14:36]: > On Wed, May 30, 2018 at 04:03:15PM +0200, Maciej Purski wrote: > > > I'm afraid, I have no idea, how to fix it quickly. You can revert it and > > in the next version I'll fix the build error and split the last patch even > > more, so we could perform a more precise bisection. I'd be grateful if you > > could push it on your test coupled branch and Tony could test it again > > before > > merging it with next again. > > Yeah, if we could get testing from Tony first that'd be ideal. I can boot test the patches easily no problem. Maciej, maybe email me some debug version and I'll email you back the dmesg output. Regards, Tony
Re: Regression in Linux next again
* Mark Brown [180530 14:36]: > On Wed, May 30, 2018 at 04:03:15PM +0200, Maciej Purski wrote: > > > I'm afraid, I have no idea, how to fix it quickly. You can revert it and > > in the next version I'll fix the build error and split the last patch even > > more, so we could perform a more precise bisection. I'd be grateful if you > > could push it on your test coupled branch and Tony could test it again > > before > > merging it with next again. > > Yeah, if we could get testing from Tony first that'd be ideal. I can boot test the patches easily no problem. Maciej, maybe email me some debug version and I'll email you back the dmesg output. Regards, Tony
Re: Regression in Linux next again
On Wed, May 30, 2018 at 04:03:15PM +0200, Maciej Purski wrote: > I'm afraid, I have no idea, how to fix it quickly. You can revert it and > in the next version I'll fix the build error and split the last patch even > more, so we could perform a more precise bisection. I'd be grateful if you > could push it on your test coupled branch and Tony could test it again before > merging it with next again. Yeah, if we could get testing from Tony first that'd be ideal. signature.asc Description: PGP signature
Re: Regression in Linux next again
On Wed, May 30, 2018 at 04:03:15PM +0200, Maciej Purski wrote: > I'm afraid, I have no idea, how to fix it quickly. You can revert it and > in the next version I'll fix the build error and split the last patch even > more, so we could perform a more precise bisection. I'd be grateful if you > could push it on your test coupled branch and Tony could test it again before > merging it with next again. Yeah, if we could get testing from Tony first that'd be ideal. signature.asc Description: PGP signature
Re: Regression in Linux next again
On 05/30/2018 11:13 AM, Mark Brown wrote: On Tue, May 29, 2018 at 03:15:01PM -0700, Tony Lindgren wrote: I think I bisected this same issue for the second time now but for a different merge window. What's up with that? Last time we just reverted it as Maciej was unable to reproduce your problem, he's tried again with some alterations. Reverting both patches fixes the issue for me. I could not debug it further because of the compile error(s). OK, unless this gets fixed really quickly I'll revert again. I'm afraid, I have no idea, how to fix it quickly. You can revert it and in the next version I'll fix the build error and split the last patch even more, so we could perform a more precise bisection. I'd be grateful if you could push it on your test coupled branch and Tony could test it again before merging it with next again. Best regards, Maciej Purski
Re: Regression in Linux next again
On 05/30/2018 11:13 AM, Mark Brown wrote: On Tue, May 29, 2018 at 03:15:01PM -0700, Tony Lindgren wrote: I think I bisected this same issue for the second time now but for a different merge window. What's up with that? Last time we just reverted it as Maciej was unable to reproduce your problem, he's tried again with some alterations. Reverting both patches fixes the issue for me. I could not debug it further because of the compile error(s). OK, unless this gets fixed really quickly I'll revert again. I'm afraid, I have no idea, how to fix it quickly. You can revert it and in the next version I'll fix the build error and split the last patch even more, so we could perform a more precise bisection. I'd be grateful if you could push it on your test coupled branch and Tony could test it again before merging it with next again. Best regards, Maciej Purski
Re: Regression in Linux next again
On Tue, May 29, 2018 at 03:15:01PM -0700, Tony Lindgren wrote: > I think I bisected this same issue for the second time now > but for a different merge window. What's up with that? Last time we just reverted it as Maciej was unable to reproduce your problem, he's tried again with some alterations. > Reverting both patches fixes the issue for me. I could > not debug it further because of the compile error(s). OK, unless this gets fixed really quickly I'll revert again. signature.asc Description: PGP signature
Re: Regression in Linux next again
On Tue, May 29, 2018 at 03:15:01PM -0700, Tony Lindgren wrote: > I think I bisected this same issue for the second time now > but for a different merge window. What's up with that? Last time we just reverted it as Maciej was unable to reproduce your problem, he's tried again with some alterations. > Reverting both patches fixes the issue for me. I could > not debug it further because of the compile error(s). OK, unless this gets fixed really quickly I'll revert again. signature.asc Description: PGP signature
Regression in Linux next again
Hi, Linux next has a regression at least on beaglebone-x15 where booting Linux hangs shortly after starting init when loading modules with no output. I bisected it down to commit 456e7cdf3b1a ("regulator: core: Change voltage setting path") I think I bisected this same issue for the second time now but for a different merge window. What's up with that? And then we also have commit 696861761a58 ("regulator: core: Add voltage balancing mechanism") that fails to compile breaking git bisect: drivers/regulator/core.c: In function 'regulator_balance_voltage': drivers/regulator/core.c:3284:9: error: implicit declaration of function 'regulator_set_voltage'. Reverting both patches fixes the issue for me. I could not debug it further because of the compile error(s). Regards, Tony
Regression in Linux next again
Hi, Linux next has a regression at least on beaglebone-x15 where booting Linux hangs shortly after starting init when loading modules with no output. I bisected it down to commit 456e7cdf3b1a ("regulator: core: Change voltage setting path") I think I bisected this same issue for the second time now but for a different merge window. What's up with that? And then we also have commit 696861761a58 ("regulator: core: Add voltage balancing mechanism") that fails to compile breaking git bisect: drivers/regulator/core.c: In function 'regulator_balance_voltage': drivers/regulator/core.c:3284:9: error: implicit declaration of function 'regulator_set_voltage'. Reverting both patches fixes the issue for me. I could not debug it further because of the compile error(s). Regards, Tony