From mboxrd@z Thu Jan 1 00:00:00 1970 From: Daniel Lezcano Subject: Re: [PATCH v5 4/6] acpi: Add GTDT table parse driver into ACPI driver Date: Tue, 31 May 2016 00:56:36 +0200 Message-ID: <574CC524.5090601@linaro.org> References: <1464096633-9309-1-git-send-email-fu.wei@linaro.org> <1464096633-9309-5-git-send-email-fu.wei@linaro.org> Mime-Version: 1.0 Content-Type: text/plain; charset=utf-8; format=flowed Content-Transfer-Encoding: QUOTED-PRINTABLE Return-path: In-Reply-To: <1464096633-9309-5-git-send-email-fu.wei@linaro.org> Sender: linux-kernel-owner@vger.kernel.org To: fu.wei@linaro.org, rjw@rjwysocki.net, lenb@kernel.org, tglx@linutronix.de, marc.zyngier@arm.com, hanjun.guo@linaro.org Cc: linux-arm-kernel@lists.infradead.org, linaro-acpi@lists.linaro.org, linux-kernel@vger.kernel.org, linux-acpi@vger.kernel.org, rruigrok@codeaurora.org, harba@codeaurora.org, cov@codeaurora.org, timur@codeaurora.org, graeme.gregory@linaro.org, al.stone@linaro.org, jcm@redhat.com, wei@redhat.com, arnd@arndb.de, wim@iguana.be, catalin.marinas@arm.com, will.deacon@arm.com, Suravee.Suthikulpanit@amd.com, leo.duran@amd.com List-Id: linux-acpi@vger.kernel.org On 05/24/2016 03:30 PM, fu.wei@linaro.org wrote: > From: Fu Wei > > This driver adds support for parsing all kinds of timer in GTDT: > (1)arch timer: provide a kernel API to parse all the PPIs and > always-on info in GTDT and export them by filling the structs > which provided by parameters(pointer of them). > > (2)memory-mapped timer: provide a kernel APIs to parse > GT Block Structure in GTDT, export all the timer info by filling > the struct which provided by parameter(pointer of the struct). > > (3)SBSA Generic Watchdog: parse all info in SBSA Generic Watchdog > Structure in GTDT, and creating a platform device with that > information. This allows the operating system to obtain device > data from the resource of platform device. > The platform device named "sbsa-gwdt" can be used by the ARM SBSA > Generic Watchdog driver. > > By this driver, we can simplify all the relevant drivers, and > separate all the ACPI GTDT knowledge from them. > > Signed-off-by: Fu Wei > Signed-off-by: Hanjun Guo Hi Fu Wei, the code is better than the previous version but it is still confusing.= =20 There are global variables accessed several times by different function= s=20 and modified. This is an encapsulation issue. This code is a bit hard to follow. Please split it out into 4 patches: 1. Create a clean base with helper functions and proper encapsulation=20 for gtdt. 2. Add gtdt timer based code 3. Add gtdt mem timer based code 4. Add gtdt watchdog based code Comments below. Thanks ! -- Daniel > --- > drivers/acpi/Kconfig | 9 + > drivers/acpi/Makefile | 1 + > drivers/acpi/acpi_gtdt.c | 309 ++++++++++++++++++++++++= +++++++++++ > include/clocksource/arm_arch_timer.h | 16 ++ > include/linux/acpi.h | 6 + > 5 files changed, 341 insertions(+) > > diff --git a/drivers/acpi/Kconfig b/drivers/acpi/Kconfig > index b7e2e77..27a5cf9 100644 > --- a/drivers/acpi/Kconfig > +++ b/drivers/acpi/Kconfig > @@ -521,4 +521,13 @@ config XPOWER_PMIC_OPREGION > > endif > > +config ACPI_GTDT > + bool "ACPI GTDT Support" > + depends on ARM64 > + help > + GTDT (Generic Timer Description Table) provides information > + for per-processor timers and Platform (memory-mapped) timers > + for ARM platforms. Select this option to provide information > + needed for the timers init. Why not ARM64's Kconfig select ACPI_GTDT ? This config option assumes an user will manually select this option=20 which I believe it makes sense to have always on when ARM64=3Dy. So why= =20 not create a silent option and select it directly from the ARM64=20 platform Kconfig ? > endif # ACPI > diff --git a/drivers/acpi/Makefile b/drivers/acpi/Makefile > index 251ce85..848aa8a 100644 > --- a/drivers/acpi/Makefile > +++ b/drivers/acpi/Makefile > @@ -99,5 +99,6 @@ obj-$(CONFIG_ACPI_EXTLOG) +=3D acpi_extlog.o > obj-$(CONFIG_PMIC_OPREGION) +=3D pmic/intel_pmic.o > obj-$(CONFIG_CRC_PMIC_OPREGION) +=3D pmic/intel_pmic_crc.o > obj-$(CONFIG_XPOWER_PMIC_OPREGION) +=3D pmic/intel_pmic_xpower.o > +obj-$(CONFIG_ACPI_GTDT) +=3D acpi_gtdt.o > > video-objs +=3D acpi_video.o video_detect.o > diff --git a/drivers/acpi/acpi_gtdt.c b/drivers/acpi/acpi_gtdt.c > new file mode 100644 > index 0000000..e54fadf > --- /dev/null > +++ b/drivers/acpi/acpi_gtdt.c > @@ -0,0 +1,309 @@ > +/* > + * ARM Specific GTDT table Support > + * > + * Copyright (C) 2015, Linaro Ltd. > + * Author: Fu Wei > + * Hanjun Guo > + * > + * This program is free software; you can redistribute it and/or mod= ify > + * it under the terms of the GNU General Public License version 2 as > + * published by the Free Software Foundation. > + */ > + > +#include > +#include > +#include > +#include > +#include > +#include > + > +#include > + > +#undef pr_fmt > +#define pr_fmt(fmt) "GTDT: " fmt > + > +typedef int (*platform_timer_handler)(void *platform_timer, int inde= x, > + void *data); > + > +static void *platform_timer_struct __initdata; > +static u32 platform_timer_count __initdata; > +static void *gtdt_end __initdata; I am pretty sure you can get ride of these global variables with a=20 little effort. > +static int __init for_platform_timer(enum acpi_gtdt_type type, > + platform_timer_handler handler, void *data) =46or the clarity of the code, I suggest to use a macro with a name=20 similar to what we find in the kernel: #define gtdt_for_each(type, ...) ... #define gtdt_for_each_timer gtdt_for_each(ACPI_GTDT_TYPE_TIMER_BLOCK) #define gtdt_for_each_wd gtdt_for_each(ACPI_GTDT_TYPE_WATCHDOG) =2E.. and rework this function to clarify its purpose. What is for ? Count the number of 'type' which succeed to call the hand= ler ? > +{ > + struct acpi_gtdt_header *header; > + int i, index, ret; > + void *platform_timer =3D platform_timer_struct; > + > + for (i =3D 0, index =3D 0; i < platform_timer_count; i++) { > + if (platform_timer > gtdt_end) { > + pr_err(FW_BUG "subtable pointer overflows.\n"); > + platform_timer_count =3D i; =46ix firmware bug in the kernel ? No. Up to the firmware to fix it, "n= o=20 handy, no candy". > + break; > + } > + header =3D (struct acpi_gtdt_header *)platform_timer; > + if (header->type =3D=3D type) { > + ret =3D handler(platform_timer, index, data); > + if (ret) > + pr_err("failed to handler subtable %d.\n", i); > + else > + index++; > + } > + platform_timer +=3D header->length; That is not correct. This function is setting the platform_timer pointe= r=20 to the end. Even if that works, it violates the encapsulation paradigm.= =20 Regarding how are used those global variables, please kill them and use= =20 proper parameter and structure encapsulation. > + } > + > + return index; > +} > + > +/* > + * Get some basic info from GTDT table, and init the global variable= s above > + * for all timers initialization of Generic Timer. > + * This function does some validation on GTDT table, and will be run= only once. > + */ > +static void __init platform_timer_init(struct acpi_table_header *tab= le, > + struct acpi_table_gtdt *gtdt) > +{ > + gtdt_end =3D (void *)table + table->length; > + > + if (table->revision < 2) { > + pr_info("Revision:%d doesn't support Platform Timers.\n", > + table->revision); > + return; > + } This check should be called much sooner, before entering the gtdt=20 subsystems. It is too late here. > + platform_timer_count =3D gtdt->platform_timer_count; > + if (!platform_timer_count) { > + pr_info("No Platform Timer structures.\n"); > + return; > + } > + > + platform_timer_struct =3D (void *)gtdt + gtdt->platform_timer_offse= t; > + if (platform_timer_struct < (void *)table + > + sizeof(struct acpi_table_gtdt)) { > + pr_err(FW_BUG "Platform Timer pointer error.\n"); > + platform_timer_struct =3D NULL; > + platform_timer_count =3D 0; > + } Why this check ? > +} > + > +static int __init map_generic_timer_interrupt(u32 interrupt, u32 fla= gs) > +{ > + int trigger, polarity; > + > + if (!interrupt) > + return 0; > + > + trigger =3D (flags & ACPI_GTDT_INTERRUPT_MODE) ? ACPI_EDGE_SENSITIV= E > + : ACPI_LEVEL_SENSITIVE; > + > + polarity =3D (flags & ACPI_GTDT_INTERRUPT_POLARITY) ? ACPI_ACTIVE_L= OW > + : ACPI_ACTIVE_HIGH; > + > + return acpi_register_gsi(NULL, interrupt, trigger, polarity); > +} > + > +/* > + * Get the necessary info of arch_timer from GTDT table. > + */ > +int __init gtdt_arch_timer_init(struct acpi_table_header *table, int= *ppi, > + bool *c3stop, u32 *timer_count) > +{ > + struct acpi_table_gtdt *gtdt; This API is not correctly separating both subsystems. Moving the code=20 from the arch_arm_timer init function here is not the solution. I suggest the following: 1. Change CLOCKSOURCE_ACPI_DECLARE to pass the init function as: int __init gtdt_init(struct acpi_table_gtdt *gtdt); 2. Add helpers: int gtdt_map_irq(struct acpi_table_gtdt *gtdt, int flags); int gtdt_c3stop(struct acpi_table_gtdt *gtdt); 3. Export these helpers arch_arm_timer.c should see nothing more than: struct acpi_table_gtdt; 4. Use these helpers in the arch_arm_timer.c static int __init arch_timer_acpi_init(struct acpi_table_gtdt *gtdt) { arch_timer_ppi[PHYS_SECURE_PPI] =3D gtdt_map_irq(gtdt, flags); arch_timer_ppi[PHYS_NONSECURE_PPI] =3D gtdt_map_irq(gtdt, flags); arch_timer_ppi[VIRT_PPI] =3D gtdt_map_irq(gtdt, flags); arch_timer_ppi[HYP_PPI] =3D gtdt_map_irq(gtdt, flags); arch_timer_c3stop =3D gtdt_c3stop(gtdt); arch_timer_init(); return 0; } > + if (acpi_disabled || !table || !ppi || !c3stop) > + return -EINVAL; Why acpi_disabled is an error here but ok for gtdt_sbsa_gwdt_init() ? > + gtdt =3D container_of(table, struct acpi_table_gtdt, header); > + if (!gtdt) { > + pr_err("table pointer error.\n"); > + return -EINVAL; > + } > + > + ppi[PHYS_SECURE_PPI] =3D > + map_generic_timer_interrupt(gtdt->secure_el1_interrupt, > + gtdt->secure_el1_flags); > + > + ppi[PHYS_NONSECURE_PPI] =3D > + map_generic_timer_interrupt(gtdt->non_secure_el1_interrupt, > + gtdt->non_secure_el1_flags); > + > + ppi[VIRT_PPI] =3D > + map_generic_timer_interrupt(gtdt->virtual_timer_interrupt, > + gtdt->virtual_timer_flags); > + > + ppi[HYP_PPI] =3D > + map_generic_timer_interrupt(gtdt->non_secure_el2_interrupt, > + gtdt->non_secure_el2_flags); > + > + *c3stop =3D !(gtdt->non_secure_el1_flags & ACPI_GTDT_ALWAYS_ON); > + > + platform_timer_init(table, gtdt); > + if (timer_count) > + *timer_count =3D platform_timer_count; come on ! > + > + return 0; > +} > + > +/* > + * Helper function for getting the pointer of a timer frame in GT bl= ock. > + */ > +static void __init *gtdt_gt_timer_frame(struct acpi_gtdt_timer_block= *gt_block, > + int index) > +{ > + void *timer_frame =3D (void *)gt_block + gt_block->timer_offset + > + sizeof(struct acpi_gtdt_timer_entry) * index; > + > + if (timer_frame <=3D (void *)gt_block + gt_block->header.length - > + sizeof(struct acpi_gtdt_timer_entry)) > + return timer_frame; > + > + return NULL; > +} > + > +static int __init gtdt_parse_gt_block(void *platform_timer, int inde= x, > + void *data) > +{ > + struct acpi_gtdt_timer_block *block; > + struct acpi_gtdt_timer_entry *frame; > + struct gt_block_data *block_data; > + int i, j; > + > + if (!platform_timer || !data) > + return -EINVAL; > + > + block =3D platform_timer; > + block_data =3D data + sizeof(struct gt_block_data) * index; > + > + if (!block->block_address || !block->timer_count) { > + pr_err(FW_BUG "invalid GT Block data.\n"); > + return -EINVAL; > + } > + block_data->cntctlbase_phy =3D (phys_addr_t)block->block_address; > + block_data->timer_count =3D block->timer_count; > + > + /* > + * Get the GT timer Frame data for every GT Block Timer > + */ > + for (i =3D 0, j =3D 0; i < block->timer_count; i++) { > + frame =3D gtdt_gt_timer_frame(block, i); > + if (!frame || !frame->base_address || !frame->timer_interrupt) { > + pr_err(FW_BUG "invalid GT Block Timer data, skip.\n"); > + continue; > + } > + block_data->timer[j].frame_nr =3D frame->frame_number; > + block_data->timer[j].cntbase_phy =3D frame->base_address; > + block_data->timer[j].irq =3D map_generic_timer_interrupt( > + frame->timer_interrupt, > + frame->timer_flags); > + if (frame->virtual_timer_interrupt) > + block_data->timer[j].virt_irq =3D > + map_generic_timer_interrupt( > + frame->virtual_timer_interrupt, > + frame->virtual_timer_flags); > + j++; > + } > + > + if (j) > + return 0; > + > + block_data->cntctlbase_phy =3D (phys_addr_t)NULL; > + block_data->timer_count =3D 0; > + > + return -EINVAL; > +} > + > +/* > + * Get the GT block info for memory-mapped timer from GTDT table. > + * Please make sure we have called gtdt_arch_timer_init, because it = helps to > + * init the global variables. > + */ > +int __init gtdt_arch_timer_mem_init(struct gt_block_data *data) > +{ > + int ret =3D for_platform_timer(ACPI_GTDT_TYPE_TIMER_BLOCK, > + gtdt_parse_gt_block, (void *)data); (void *) explicit cast is not needed. > + pr_info("found %d memory-mapped timer block.\n", ret); > + > + return ret; > +} > + > +/* > + * Initialize a SBSA generic Watchdog platform device info from GTDT > + */ > +static int __init gtdt_import_sbsa_gwdt(void *platform_timer, > + int index, void *data) > +{ > + struct platform_device *pdev; > + struct acpi_gtdt_watchdog *wd =3D platform_timer; > + int irq =3D map_generic_timer_interrupt(wd->timer_interrupt, > + wd->timer_flags); > + int no_irq =3D 1; > + > + /* > + * According to SBSA specification the size of refresh and control > + * frames of SBSA Generic Watchdog is SZ_4K(Offset 0x000 =E2=80=93 = 0xFFF). > + */ > + struct resource res[] =3D { > + DEFINE_RES_MEM(wd->control_frame_address, SZ_4K), > + DEFINE_RES_MEM(wd->refresh_frame_address, SZ_4K), > + DEFINE_RES_IRQ(irq), > + }; > + > + pr_debug("a Watchdog GT(0x%llx/0x%llx gsi:%u flags:0x%x).\n", > + wd->refresh_frame_address, wd->control_frame_address, > + wd->timer_interrupt, wd->timer_flags); > + > + if (!(wd->refresh_frame_address && wd->control_frame_address)) { > + pr_err(FW_BUG "failed getting the Watchdog GT frame addr.\n"); > + return -EINVAL; > + } > + > + if (!wd->timer_interrupt) > + pr_warn(FW_BUG "failed getting the Watchdog GT GSIV.\n"); > + else if (irq <=3D 0) > + pr_warn("failed to map the Watchdog GT GSIV.\n"); > + else > + no_irq =3D 0; > + > + /* > + * Add a platform device named "sbsa-gwdt" to match the platform dr= iver. > + * "sbsa-gwdt": SBSA(Server Base System Architecture) Generic Watch= dog > + * The platform driver (like drivers/watchdog/sbsa_gwdt.c)can get d= evice > + * info below by matching this name. > + */ > + pdev =3D platform_device_register_simple("sbsa-gwdt", index, res, > + ARRAY_SIZE(res) - no_irq); > + if (IS_ERR(pdev)) { > + acpi_unregister_gsi(wd->timer_interrupt); > + return PTR_ERR(pdev); > + } > + > + return 0; > +} > + > +static int __init gtdt_sbsa_gwdt_init(void) > +{ > + struct acpi_table_header *table; > + struct acpi_table_gtdt *gtdt; > + int ret; > + > + if (acpi_disabled) > + return 0; > + > + if (ACPI_FAILURE(acpi_get_table(ACPI_SIG_GTDT, 0, &table))) > + return -EINVAL; > + > + /* global variables initialization */ > + gtdt_end =3D (void *)table + table->length; That is prone to bug. The end of the table is computed here and stored=20 in the global variable and then computed again in platform_timer_init=20 and also stored in the global variable. Kill this global variable. > + gtdt =3D container_of(table, struct acpi_table_gtdt, header); > + platform_timer_struct =3D (void *)gtdt + gtdt->platform_timer_offse= t; > + > + ret =3D for_platform_timer(ACPI_GTDT_TYPE_WATCHDOG, > + gtdt_import_sbsa_gwdt, NULL); > + pr_info("found %d SBSA generic Watchdog.\n", ret); > + > + return 0; > +} > + > +device_initcall(gtdt_sbsa_gwdt_init); Do you really have to call device_initcall ? --=20 Linaro.org =E2=94=82 Open source software fo= r ARM SoCs =46ollow Linaro: Facebook | Twitter | Blog