| Message ID | 1549355535-22052-3-git-send-email-clabbe@baylibre.com |
|---|---|
| State | New |
| Headers | show |
| Series | arc: hsdk_defconfig: permit kernelci jobs | expand |
Hi Corentin, In case of devboards (like HSDK) we really often disable bootloader and load Linux image in memory via JTAG. Enabling CONFIG_ARC_UBOOT_SUPPORT by default will break it as we will try to interpret some junk in a registers as a pointers to bootargs/etc which aren't set by anyone in case of JTAG using. So it isn't a good idea to have CONFIG_ARC_UBOOT_SUPPORT enabled by default. On Tue, 2019-02-05 at 08:32 +0000, Corentin Labbe wrote: > We have now a HSDK device in our kernelci lab, but kernel builded via > the hsdk_defconfig ignore bootargs, so it cannot boot kernelci jobs yet. > Furthermore, I think the probability is that devices will be booted more via > uboot than by a debugger. > > So this patch enable CONFIG_ARC_UBOOT_SUPPORT in hsdk_defconfig. > > Signed-off-by: Corentin Labbe <clabbe@baylibre.com> > --- > arch/arc/configs/hsdk_defconfig | 1 + > 1 file changed, 1 insertion(+) > > diff --git a/arch/arc/configs/hsdk_defconfig b/arch/arc/configs/hsdk_defconfig > index 0e5fd29ed238..b80e3ac5c93c 100644 > --- a/arch/arc/configs/hsdk_defconfig > +++ b/arch/arc/configs/hsdk_defconfig > @@ -9,6 +9,7 @@ CONFIG_NAMESPACES=y > # CONFIG_PID_NS is not set > CONFIG_BLK_DEV_INITRD=y > CONFIG_BLK_DEV_RAM=y > +CONFIG_ARC_UBOOT_SUPPORT=y > CONFIG_EMBEDDED=y > CONFIG_PERF_EVENTS=y > # CONFIG_VM_EVENT_COUNTERS is not set
On 2/5/19 3:42 AM, Eugeniy Paltsev wrote: > Hi Corentin, > > In case of devboards (like HSDK) we really often disable bootloader and load > Linux image in memory via JTAG. Enabling CONFIG_ARC_UBOOT_SUPPORT by > default will break it as we will try to interpret some junk in a registers > as a pointers to bootargs/etc which aren't set by anyone in case of JTAG using. > > So it isn't a good idea to have CONFIG_ARC_UBOOT_SUPPORT enabled by default. Right. It is difficult to accommodate everyone's needs (often conflicting) in a single defconfig. Can you folks create an out-of-tree defconfig or some such. -Vineet
Hi Vineet, Corentin, > -----Original Message----- > From: Vineet Gupta <vgupta@synopsys.com> > Sent: Tuesday, February 5, 2019 7:42 PM > To: Eugeniy Paltsev <paltsev@synopsys.com>; clabbe@baylibre.com > Cc: linux-kernel@vger.kernel.org; alexey.brodkin@synopsys.com; khilman@baylibre.com; linux-snps- > arc@lists.infradead.org > Subject: Re: [PATCH 2/2] arc: hsdk_defconfig: enable CONFIG_ARC_UBOOT_SUPPORT > > On 2/5/19 3:42 AM, Eugeniy Paltsev wrote: > > Hi Corentin, > > > > In case of devboards (like HSDK) we really often disable bootloader and load > > Linux image in memory via JTAG. Enabling CONFIG_ARC_UBOOT_SUPPORT by > > default will break it as we will try to interpret some junk in a registers > > as a pointers to bootargs/etc which aren't set by anyone in case of JTAG using. > > > > So it isn't a good idea to have CONFIG_ARC_UBOOT_SUPPORT enabled by default. > > Right. > > It is difficult to accommodate everyone's needs (often conflicting) in a single > defconfig. > Can you folks create an out-of-tree defconfig or some such. I do think there's a proper solution which [hopefully] makes both parties happy. Eugeniy is about to send-out a patch which allows us to not care about garbage in R0/R2 when running after U-Boot. -Alexey
Alexey Brodkin <alexey.brodkin@synopsys.com> writes: > Hi Vineet, Corentin, > >> -----Original Message----- >> From: Vineet Gupta <vgupta@synopsys.com> >> Sent: Tuesday, February 5, 2019 7:42 PM >> To: Eugeniy Paltsev <paltsev@synopsys.com>; clabbe@baylibre.com >> Cc: linux-kernel@vger.kernel.org; alexey.brodkin@synopsys.com; khilman@baylibre.com; linux-snps- >> arc@lists.infradead.org >> Subject: Re: [PATCH 2/2] arc: hsdk_defconfig: enable CONFIG_ARC_UBOOT_SUPPORT >> >> On 2/5/19 3:42 AM, Eugeniy Paltsev wrote: >> > Hi Corentin, >> > >> > In case of devboards (like HSDK) we really often disable bootloader and load >> > Linux image in memory via JTAG. Enabling CONFIG_ARC_UBOOT_SUPPORT by >> > default will break it as we will try to interpret some junk in a registers >> > as a pointers to bootargs/etc which aren't set by anyone in case of JTAG using. >> > >> > So it isn't a good idea to have CONFIG_ARC_UBOOT_SUPPORT enabled by default. >> >> Right. >> >> It is difficult to accommodate everyone's needs (often conflicting) in a single >> defconfig. >> Can you folks create an out-of-tree defconfig or some such. > > I do think there's a proper solution which [hopefully] makes both parties happy. > Eugeniy is about to send-out a patch which allows us to not care about > garbage in R0/R2 when running after U-Boot. OK, cool, that sounds like a good workaround for the JTAG use case. Just curious: what is the more common case for end-users? u-boot or JTAG? At least on every other arch/board we use in kernelCI, u-boot is the *much* more common load path than JTAG. IMO, I would suggest that in the case of unresolvable conflicts like, u-boot should be the normal path, and JTAG would be the special case. Anyways, hopefully the patch from Eugeniy works and gets rid of the conflict. Related: if you end up accepting this series, they'll also need to be backported to stable branches if we want to support them in kernelCI. Kevin
diff --git a/arch/arc/configs/hsdk_defconfig b/arch/arc/configs/hsdk_defconfig index 0e5fd29ed238..b80e3ac5c93c 100644 --- a/arch/arc/configs/hsdk_defconfig +++ b/arch/arc/configs/hsdk_defconfig @@ -9,6 +9,7 @@ CONFIG_NAMESPACES=y # CONFIG_PID_NS is not set CONFIG_BLK_DEV_INITRD=y CONFIG_BLK_DEV_RAM=y +CONFIG_ARC_UBOOT_SUPPORT=y CONFIG_EMBEDDED=y CONFIG_PERF_EVENTS=y # CONFIG_VM_EVENT_COUNTERS is not set
We have now a HSDK device in our kernelci lab, but kernel builded via the hsdk_defconfig ignore bootargs, so it cannot boot kernelci jobs yet. Furthermore, I think the probability is that devices will be booted more via uboot than by a debugger. So this patch enable CONFIG_ARC_UBOOT_SUPPORT in hsdk_defconfig. Signed-off-by: Corentin Labbe <clabbe@baylibre.com> --- arch/arc/configs/hsdk_defconfig | 1 + 1 file changed, 1 insertion(+)