Re: [nand] BUG: unable to handle kernel NULL pointer dereference at 000005f4

[Fengguang Wu <fengguang.wu@intel.com>]

[-- Attachment #1: Type: text/plain, Size: 6751 bytes --]

On Mon, Dec 14, 2015 at 02:22:59PM +0100, Boris Brezillon wrote:
> On Mon, 14 Dec 2015 21:05:59 +0800
> Fengguang Wu <fengguang.wu@intel.com> wrote:
> 
> > Hi Boris,
> > 
> > On Mon, Dec 14, 2015 at 01:54:43PM +0100, Boris Brezillon wrote:
> > > On Mon, 14 Dec 2015 14:45:37 +0800
> > > Fengguang Wu <fengguang.wu@intel.com> wrote:
> > > 
> > > > Hi Boris,
> > > > 
> > > > On Sun, Dec 13, 2015 at 09:50:43PM +0100, Boris Brezillon wrote:
> > > > > Hi Fengguang,
> > > > > 
> > > > > On Sun, 13 Dec 2015 09:49:17 +0800
> > > > > Fengguang Wu <fengguang.wu@intel.com> wrote:
> > > > > 
> > > > > > Hi Boris,
> > > > > > 
> > > > > > FYI, this happened when merging your branch, which indicates possible
> > > > > > conflicts with other developer's patches or the new RC release.
> > > > > 
> > > > > Is this boot test also run on my bbrezillon-0day/nand/layering-rework
> > > > > branch (before merging it into devel-spot), and if it is, do you see the
> > > > > same error?
> > > > 
> > > > The bisect tests bbrezillon-0day/nand/layering-rework, however it's
> > > > fine there.
> > > > 
> > > > > Is the devel-spot-201512102358 tag publicly accessible? If it's not,
> > > > 
> > > > Nope.
> > > > 
> > > > > could you push it somewhere (or give me a way to generate it by
> > > > > pulling the relevant branches), so that I can investigate on this bug? 
> > > > 
> > > > Or it may be more convenient to check linux-next first?
> > > > 
> > > > FYI, the merged branches in devel-spot-201512102358 are:
> > > > 
> > > > af8fc33 Merge 'bbrezillon-0day/nand/layering-rework' into devel-spot-201512102358
> > > > acfc1de Merge 'linux-review/Maury-Markowitz/renaming-NTSC-file-because-it-doesnt-actually-use-8VSV-updating-channel-listings-add-channel-numbers/20151210-210449' into devel-spot-201512102358
> > > > fb9daac Merge 'fdmanana/integration-4.4' into devel-spot-201512102358
> > > > 1b30725 Merge 'fdmanana/for-chris-4.4' into devel-spot-201512102358
> > > > 83f2246 Merge 'arm-integrator/realview-defconfig' into devel-spot-201512102358
> > > > 03df745 Merge 'arm-integrator/realview-multiplatform-base' into devel-spot-201512102358
> > > > d2aea29 Merge 'linux-review/Arnd-Bergmann/i2c-allow-building-emev2-without-slave-mode-again/20151210-211642' into devel-spot-201512102358
> > > > e2f041d Merge 'arm-integrator/realview-base-armsoc' into devel-spot-201512102358
> > > > 5c46550 Merge 'linux-review/Bert-Kenward/sfc-only-use-RSS-filters-if-we-re-using-RSS/20151210-213247' into devel-spot-201512102358
> > > > d07f80f Merge 'sound/test/hdmi-jack' into devel-spot-201512102358
> > > > 540f8b7 Merge 'sound/topic/hdmi-jack' into devel-spot-201512102358
> > > > 9111fa0 Merge 'linux-review/matrandg-cisco-com/tc358743-Use-local-array-with-fixed-size-in-i2c-write/20151210-214154' into devel-spot-201512102358
> > > > 3d81ef9 Merge 'linux-review/matrandg-cisco-com/tc358743-Print-timings-only-when-debug-level-is-set/20151210-220321' into devel-spot-201512102358
> > > > 122e43c Merge 'bbrezillon-0day/nand/ecclayout-squashed' into devel-spot-201512102358
> > > > e0572f8 Merge 'wireless-drivers-next/pending' into devel-spot-201512102358
> > > > dc5d9c9 Merge 'linux-review/Geliang-Tang/list-introduce-list_is_first/20151210-222552' into devel-spot-201512102358
> > > > 289e2a8 Merge 'linux-review/Arnd-Bergmann/staging-davinci_vfpe-allow-modular-build/20151210-223348' into devel-spot-201512102358
> > > > d74c50d Merge 'linux-review/Damien-Horsley/Add-driver-for-Pistachio-internal-DAC/20151210-224304' into devel-spot-201512102358
> > > > ee3c58b Merge 'cryptodev/master' into devel-spot-201512102358
> > > > d390f7d Merge 'linux-review/John-Garry/hisi_sas-use-platform_get_irq/20151210-224851' into devel-spot-201512102358
> > > > de31f5d Merge 'linux-review/Mike-Marciniszyn/staging-rdma-hfi1-add-definitions-for-OPA-traps/20151210-230158' into devel-spot-201512102358
> > > > 5618568 Merge 'linux-review/Andrei-Emeltchenko/Bluetooth-h5-Do-not-initialize-Configuration-field/20151210-230440' into devel-spot-201512102358
> > > > 759aa32 Merge 'linux-review/Moritz-K-nig/staging-COMEDI-checkpatch-pl-cleanups-in-comedi-h/20151210-230707' into devel-spot-201512102358
> > > > 2edcbb4 Merge 'linux-review/Andy-Shevchenko/net-hns-annotate-IO-address-space-properly/20151210-232510' into devel-spot-201512102358
> > > > 0f51fdd Merge 'integrity/next-kexec' into devel-spot-201512102358
> > > > b32bcbc Merge 'dm/dm-4.5' into devel-spot-201512102358
> > > > ea297f5 Merge 'linux-review/Mathieu-Desnoyers/thread_local_abi-system-call-caching-current-CPU-number-x86/20151210-234242' into devel-spot-201512102358
> > > > 52bc163 Merge 'linux-review/yegorslists-googlemail-com/USB-serial-option-add-support-for-Quectel-UC20/20151210-234659' into devel-spot-201512102358
> > > > 797cc7c Merge 'linux-review/Andy-Shevchenko/staging-rtl8188eu-provide-an-absolute-path-to-include-folder/20151210-234752' into devel-spot-201512102358
> > > > 23d8b49 Merge 'omap-dss2/for-next' into devel-spot-201512102358
> > > > 88a768f 0day base guard for 'devel-spot-201512102358'
> > > > 527e931 Linux 4.4-rc4
> > > > 
> > > 
> > > Okay, I think I found the problem: you're pulling
> > > bbrezillon-0day/nand/ecclayout-squashed and then
> > > bbrezillon-0day/nand/layering-rework. Those branches are both based on
> > > Brian's l2-mtd/master branch, and are both changing NAND subsystem
> > > internals. IOW, the changes brought by nand/ecclayout-squashed are
> > > introducing new "mtd->priv" occurrences, while the goal of
> > > nand/layering-rework is to replace them by mtd_to_nand(mtd) and kill
> > > the mtd->priv = xxx assignments, hence the NULL pointer exception.
> > 
> > Ah got it, sorry for the noises!
> > 
> > > I guess I'm the one responsible for this false positive, because I
> > > pushed two non-independent branches in my 0day tree.
> > > I could rebase my nand/ecclayout-squashed branch and fix all the faulty
> > > places, but that implies reordering your pull commands
> > > (nand/layering-rework before nand/ecclayout-squashed). Would that work
> > > for you? Any other suggestions to gracefully handle this kind of
> > > dependencies?
> > 
> > There is no much to do about such dependencies. One possible fix is to
> > improve the 0day system to bisect the conflicting merges, which should
> > save the lots of manual efforts to locate the problem. I've put it as
> > a TODO item.
> 
> Okay, I think I'll temporary kill the nand/ecclayout-squashed branch
> until I've properly rebased it on top of the nand/layering-rework one,
> so that you can still tests changes coming from other trees.

OK, thank you very much!

Thanks,
Fengguang