From mboxrd@z Thu Jan 1 00:00:00 1970 From: bpringlemeir@nbsps.com (Bill Pringlemeir) Date: Mon, 29 Sep 2014 11:39:11 -0400 Subject: [PATCH 0/9] ARM: vf610: Suspend/resume support In-Reply-To: <64d37b2cfb24737db93afbb6cf47978a@agner.ch> (Stefan Agner's message of "Mon, 29 Sep 2014 14:47:13 +0200") References: <87zjdqmjr3.fsf@nbsps.com> <4ee0621b0bf5f4d4e10eae7673bdf1cd@agner.ch> <87y4t9kmgz.fsf@nbsps.com> <20140928030857.GA12999@dragon> <64d37b2cfb24737db93afbb6cf47978a@agner.ch> Message-ID: <87oatye8rk.fsf@nbsps.com> To: linux-arm-kernel@lists.infradead.org List-Id: linux-arm-kernel.lists.infradead.org >> On Wed, Sep 24, 2014 at 12:33:00PM -0400, Bill Pringlemeir wrote: >>>>> I think that Shawn Guo already did a patchset to remove stuff from >>>>> the vf610.dtsi to the machine/configuration DT files. > Am 2014-09-28 05:08, schrieb Shawn Guo: >> Do you mean the pin configuration? That's really more >> board-specific. But IP block/device is not really the case. Ok. But if a DT does not use a IP block/device, then it is extra stuff in the DT. I see you only meant the pin configuration which can be quite large. So it makes sense to only apply this for the pins. >> On Wed, Sep 24, 2014 at 12:33:00PM -0400, Bill Pringlemeir wrote: >>> In this patchset, I suggested that we could include the notation in >>> the headers which are included in the 'DT' files. So instead of >>> 'dtsi', we could use, >>> >>> #define VF610_GPC_SUPPORT \ >>> gpc: gpc at 4006c000 { \ >>> compatible = "fsl,vf610-gpc"; \ >>> reg = <0x4006c000 0x1000>; \ >>> }; >>> >>> Or even, >>> >>> #define VF610_SUSPEND_DT_SUPPORT \ ... > Am 2014-09-28 05:08, schrieb Shawn Guo: >> I don't think this will be accepted by DT maintainers. We have >> already got some objections when we define pin function ID as >> multiple integers. They expect the DT macro is used for single >> integer case. Supporting status="disabled" and/or requiring enabled is also good. Is that the 'DT' approved mechanism? [snip] >> On Wed, Sep 24, 2014 at 12:33:00PM -0400, Bill Pringlemeir wrote: >>> We have, >>> >>> arch/arm/mach-imx/gpc.c >>> >>> void __init vf610_gpc_init(void) >>> { >>> struct device_node *np; >>> >>> np = of_find_compatible_node(NULL, NULL, "fsl,vf610-gpc"); >>> gpc_base = of_iomap(np, 0); >>> gpc_imr_base = gpc_base + VF610_GPC_IMR1; >>> ... >>> >>> arch/arm/mach-imx/mach-vf610.c >>> >>> static void __init vf610_init_irq(void) >>> { >>> vf610_gpc_init(); >>> irqchip_init(); >>> } >>> I don't think this will work. > Am 2014-09-28 05:08, schrieb Shawn Guo: >> Yes, you're right. But I guess this can be fixed by an additional >> of_device_is_available() check after of_find_compatible_node() call. >> On Wed, Sep 24, 2014 at 12:33:00PM -0400, Bill Pringlemeir wrote: >>> Not to mention that 'mach-vf610.c' will not build if HAVE_IMX_GPC is >>> not defined. > Am 2014-09-28 05:08, schrieb Shawn Guo: >> HAVE_IMX_GPC cannot be configured out, because it's selected by >> SOC_VF610. On 29 Sep 2014, stefan at agner.ch wrote: > Maybe we can make that optional when CONFIG_PM is on, e.g. > select HAVE_IMX_GPC if PM > The GPC module is all about power management, hence in case we want > the M4 taken care of that, we just have to configure a Linux kernel > without CONFIG_PM support. Yes as per the paragraph below, I am concerned about adding too many modules (or making inter-modules dependencies) in the drivers. I guess in a Quad-core iMx6, a system might want one CPU dedicated to some feature (non-Linux). For the Vybrid, VF6xx, this is pretty much a sure thing. There is an M4, and it does have I/D TCM, but to do anything useful it has to have some AIPS modules. >>> Also, I don't really see a use of the GPC module unless >>> suspend/resume is active? Even some wall powered designs may wish >>> to exclude this functionality? >>> I think that the SRC maybe needed for secure parts. I think that >>> some designs might wish to restrict Linux's access to these >>> registers as well. I don't actually see why we need this module? I >>> think the imx6 needs it due to multi-CPU bring-up, but in the Vybrid >>> case, this does not exist. Can you check to see why we need the >>> SRC? I don't see where we actually use it? In patch9/9, we record >>> it but do we actually access the registers? Is it just for the >>> vf610_src_init() code? Even that seems the whole 'src.c' file is >>> only needed for the 'src_base' reference, which we don't use? >> +1. I'm also wondering how SRC is used by suspend routine in this >> series. > Well I tried to bring LPStop modes to work, which would need GPR0/GPR1 > registers of SRC. If I have it working in the next series, I will need > that, but if not, I will drop those changes. Most people would want to use at least some IRAM for the M4 code space as the TCM is rather limited. Carving off the whole thing for Linux is not great either. The imx series had an iram allocator by Sacha Hauer? However, I don't see this in the mainline anymore. Is the IRAM statically allocated by 'DT'? For the 'memory' node, we keep this in the machine 'DT', so someone could carve off DDR3 for the M4. I don't think the initial series used the IRAM either? But it is more clear to me why you wanted that. I think we might exclude the IRAM and SRC from the initial series? At least this patchset might hang around before the LPStop is accepted and merged? Or you could roll them together so at least we could see what was next? I guess you might not have polished the LPSTOP as well and aren't ready for review. But I think the changes to remove the SRC and IRAM aren't too big to get this applied and then later merge them? Fwiw, Bill Pringlemeir.