5 messages, 3 authors, 2011-05-16 · open the first message on its own page
From: Vadim Bendebury <hidden>
Date: 2011-05-13 19:58:45
This patch won't necessarily apply as it has not been verified on a real hardware, so this is more of an RFC. Vadim Bendebury (1): ARM: thumb: Fix thumb mode compilation problems. arch/arm/kernel/fiq.c | 34 ++++++++++++++++++++++++++-------- arch/arm/mach-tegra/headsmp.S | 3 ++- 2 files changed, 28 insertions(+), 9 deletions(-) -- 1.7.3.1
From: Dave Martin <hidden>
Date: 2011-05-16 10:38:35
On Fri, May 13, 2011 at 12:58:45PM -0700, Vadim Bendebury wrote:
This patch won't necessarily apply as it has not been verified on a real hardware, so this is more of an RFC. Vadim Bendebury (1): ARM: thumb: Fix thumb mode compilation problems. arch/arm/kernel/fiq.c | 34 ++++++++++++++++++++++++++-------- arch/arm/mach-tegra/headsmp.S | 3 ++- 2 files changed, 28 insertions(+), 9 deletions(-) -- 1.7.3.1 _______________________________________________ linux-arm-kernel mailing list linux-arm-kernel at lists.infradead.org http://lists.infradead.org/mailman/listinfo/linux-arm-kernel
A different patch for this has already been discussed: http://thread.gmane.org/gmane.linux.ports.arm.kernel/112864 It's a bit stalled on my side because I don't have a suitable platform to test it on. As some have suggested, in my case (imx51) the inclusion of code was inherited from from an older subarch tree and is no longer appropriate -- so my problem is actually solved just by not building this code. If the code is really used in your kernel (not just built) it would be really useful if you could test this patch. However you should definitely check whether the code is really used at all on tegra. There's a fair chance that it's not used: FIQ is not much used by linux on armv7 and above: unless you make use of cache+TLB lockdown / SRAM etc., FIQ is unlikely to be significantly faster than IRQ on recent CPUs. FIQ may also get reserved for use by TrustZone on some platforms, complicating the picture further. Cheers ---Dave
From: Dave Martin <hidden>
Date: 2011-05-16 15:08:44
On Fri, May 13, 2011 at 12:58:45PM -0700, Vadim Bendebury wrote:
This patch won't necessarily apply as it has not been verified on a real hardware, so this is more of an RFC. Vadim Bendebury (1): ARM: thumb: Fix thumb mode compilation problems. arch/arm/kernel/fiq.c | 34 ++++++++++++++++++++++++++-------- arch/arm/mach-tegra/headsmp.S | 3 ++- 2 files changed, 28 insertions(+), 9 deletions(-)
As a side note, I'm happy to be CC'd if you have Thumb-2 kernel problems since I've been working on this recently, and would like to get it working across the broadest range of platforms possible. I'm definitely interested if you hit any non-platform-specific issues, since I'm not too familiar with tegra. So far, omap was the most "interesting" target: if you have problems, the following commits for omap may contain examples of things you need to fix: git log -i --grep=thumb -- arch/arm/mach-omap2 arch/arm/plat-omap dd31394 ARM: omap3: Thumb-2 compatibility for sleep34xx.S ef7a87d ARM: omap3: Thumb-2 compatibility for sram34xx.S 76d5001 ARM: omap3: Remove hand-encoded SMC instructions f96bdfa ARM: omap4: Convert END() to ENDPROC() for correct linkage with CONFIG_THUMB2_KERNEL 85243a7 ARM: omap4: Provide do_wfi() for Thumb-2 b6338bd ARM: 6649/1: omap: use fncpy to copy the PM code functions to SRAM
From: Colin Cross <hidden>
Date: 2011-05-16 16:06:26
On Mon, May 16, 2011 at 3:38 AM, Dave Martin [off-list ref] wrote:
On Fri, May 13, 2011 at 12:58:45PM -0700, Vadim Bendebury wrote:quoted
This patch won't necessarily apply as it has not been verified on a real hardware, so this is more of an RFC. Vadim Bendebury (1): ? ARM: thumb: Fix thumb mode compilation problems. ?arch/arm/kernel/fiq.c ? ? ? ? | ? 34 ++++++++++++++++++++++++++-------- ?arch/arm/mach-tegra/headsmp.S | ? ?3 ++- ?2 files changed, 28 insertions(+), 9 deletions(-) -- 1.7.3.1 _______________________________________________ linux-arm-kernel mailing list linux-arm-kernel at lists.infradead.org http://lists.infradead.org/mailman/listinfo/linux-arm-kernelA different patch for this has already been discussed: http://thread.gmane.org/gmane.linux.ports.arm.kernel/112864 It's a bit stalled on my side because I don't have a suitable platform to test it on. ?As some have suggested, in my case (imx51) the inclusion of code was inherited from from an older subarch tree and is no longer appropriate -- so my problem is actually solved just by not building this code. If the code is really used in your kernel (not just built) it would be really useful if you could test this patch. However you should definitely check whether the code is really used at all on tegra. There's a fair chance that it's not used: FIQ is not much used by linux on armv7 and above: unless you make use of cache+TLB lockdown / SRAM etc., FIQ is unlikely to be significantly faster than IRQ on recent CPUs. ?FIQ may also get reserved for use by TrustZone on some platforms, complicating the picture further.
Most Android kernels, including on Tegra, use the FIQ handler for a simple debugger on the serial console. Sending a break will enter debugger mode, which can dump backtraces from FIQ context, even if the CPU is stuck in an IRQ, or with IRQs off. We've never used a Thumb2 kernel though.
From: Dave Martin <hidden>
Date: 2011-05-16 16:31:26
On Mon, May 16, 2011 at 09:06:26AM -0700, Colin Cross wrote:
On Mon, May 16, 2011 at 3:38 AM, Dave Martin [off-list ref] wrote:quoted
On Fri, May 13, 2011 at 12:58:45PM -0700, Vadim Bendebury wrote:quoted
This patch won't necessarily apply as it has not been verified on a real hardware, so this is more of an RFC. Vadim Bendebury (1): ? ARM: thumb: Fix thumb mode compilation problems. ?arch/arm/kernel/fiq.c ? ? ? ? | ? 34 ++++++++++++++++++++++++++-------- ?arch/arm/mach-tegra/headsmp.S | ? ?3 ++- ?2 files changed, 28 insertions(+), 9 deletions(-) -- 1.7.3.1 _______________________________________________ linux-arm-kernel mailing list linux-arm-kernel at lists.infradead.org http://lists.infradead.org/mailman/listinfo/linux-arm-kernelA different patch for this has already been discussed: http://thread.gmane.org/gmane.linux.ports.arm.kernel/112864 It's a bit stalled on my side because I don't have a suitable platform to test it on. ?As some have suggested, in my case (imx51) the inclusion of code was inherited from from an older subarch tree and is no longer appropriate -- so my problem is actually solved just by not building this code. If the code is really used in your kernel (not just built) it would be really useful if you could test this patch. However you should definitely check whether the code is really used at all on tegra. There's a fair chance that it's not used: FIQ is not much used by linux on armv7 and above: unless you make use of cache+TLB lockdown / SRAM etc., FIQ is unlikely to be significantly faster than IRQ on recent CPUs. ?FIQ may also get reserved for use by TrustZone on some platforms, complicating the picture further.Most Android kernels, including on Tegra, use the FIQ handler for a simple debugger on the serial console. Sending a break will enter debugger mode, which can dump backtraces from FIQ context, even if the CPU is stuck in an IRQ, or with IRQs off. We've never used a Thumb2 kernel though.
Fair enough -- in that case I guess it may be needed. Sounds like entering the debugger will help in testing this; I don't have the board myself though. ---Dave