From c459ce8b5a7d933a3bcf6915ab17ac1e036e2ac4 Mon Sep 17 00:00:00 2001 From: Tony Luck Date: Mon, 29 Sep 2008 16:39:19 -0700 Subject: [IA64] Put the space for cpu0 per-cpu area into .data section Initial fix for making sure that we can access percpu variables in all C code (commit: 10617bbe84628eb18ab5f723d3ba35005adde143) inadvertantly allocated the memory in the "percpu" section of the vmlinux ELF executable. This confused kexec/dump. Signed-off-by: Tony Luck --- arch/ia64/include/asm/sections.h | 3 +++ arch/ia64/kernel/head.S | 9 +++++---- arch/ia64/kernel/vmlinux.lds.S | 8 +++++--- arch/ia64/mm/contig.c | 2 +- arch/ia64/mm/discontig.c | 2 +- 5 files changed, 15 insertions(+), 9 deletions(-) (limited to 'arch') diff --git a/arch/ia64/include/asm/sections.h b/arch/ia64/include/asm/sections.h index f66799891036..1a873b36a4a1 100644 --- a/arch/ia64/include/asm/sections.h +++ b/arch/ia64/include/asm/sections.h @@ -11,6 +11,9 @@ #include extern char __per_cpu_start[], __per_cpu_end[], __phys_per_cpu_start[]; +#ifdef CONFIG_SMP +extern char __cpu0_per_cpu[]; +#endif extern char __start___vtop_patchlist[], __end___vtop_patchlist[]; extern char __start___rse_patchlist[], __end___rse_patchlist[]; extern char __start___mckinley_e9_bundles[], __end___mckinley_e9_bundles[]; diff --git a/arch/ia64/kernel/head.S b/arch/ia64/kernel/head.S index 8bdea8eb62e3..66e491d8baac 100644 --- a/arch/ia64/kernel/head.S +++ b/arch/ia64/kernel/head.S @@ -367,16 +367,17 @@ start_ap: ;; #else (isAP) br.few 2f - mov r20=r19 - sub r19=r19,r18 + movl r20=__cpu0_per_cpu ;; shr.u r18=r18,3 1: - ld8 r21=[r20],8;; - st8[r19]=r21,8 + ld8 r21=[r19],8;; + st8[r20]=r21,8 adds r18=-1,r18;; cmp4.lt p7,p6=0,r18 (p7) br.cond.dptk.few 1b + mov r19=r20 + ;; 2: #endif tpa r19=r19 diff --git a/arch/ia64/kernel/vmlinux.lds.S b/arch/ia64/kernel/vmlinux.lds.S index de71da811cd6..10a7d47e8510 100644 --- a/arch/ia64/kernel/vmlinux.lds.S +++ b/arch/ia64/kernel/vmlinux.lds.S @@ -215,9 +215,6 @@ SECTIONS /* Per-cpu data: */ percpu : { } :percpu . = ALIGN(PERCPU_PAGE_SIZE); -#ifdef CONFIG_SMP - . = . + PERCPU_PAGE_SIZE; /* cpu0 per-cpu space */ -#endif __phys_per_cpu_start = .; .data.percpu PERCPU_ADDR : AT(__phys_per_cpu_start - LOAD_OFFSET) { @@ -233,6 +230,11 @@ SECTIONS data : { } :data .data : AT(ADDR(.data) - LOAD_OFFSET) { +#ifdef CONFIG_SMP + . = ALIGN(PERCPU_PAGE_SIZE); + __cpu0_per_cpu = .; + . = . + PERCPU_PAGE_SIZE; /* cpu0 per-cpu space */ +#endif DATA_DATA *(.data1) *(.gnu.linkonce.d*) diff --git a/arch/ia64/mm/contig.c b/arch/ia64/mm/contig.c index e566ff43884a..0ee085efbe29 100644 --- a/arch/ia64/mm/contig.c +++ b/arch/ia64/mm/contig.c @@ -163,7 +163,7 @@ per_cpu_init (void) * get_zeroed_page(). */ if (first_time) { - void *cpu0_data = __phys_per_cpu_start - PERCPU_PAGE_SIZE; + void *cpu0_data = __cpu0_per_cpu; first_time=0; diff --git a/arch/ia64/mm/discontig.c b/arch/ia64/mm/discontig.c index 78026aabaa7f..d8c5fcd89e5b 100644 --- a/arch/ia64/mm/discontig.c +++ b/arch/ia64/mm/discontig.c @@ -144,7 +144,7 @@ static void *per_cpu_node_setup(void *cpu_data, int node) for_each_possible_early_cpu(cpu) { if (cpu == 0) { - void *cpu0_data = __phys_per_cpu_start - PERCPU_PAGE_SIZE; + void *cpu0_data = __cpu0_per_cpu; __per_cpu_offset[cpu] = (char*)cpu0_data - __per_cpu_start; } else if (node == node_cpuid[cpu].nid) { -- cgit v1.2.3 From ad611045ce5d059af84a9855b22ca3f7a99d47be Mon Sep 17 00:00:00 2001 From: David Gibson Date: Wed, 24 Sep 2008 16:39:04 +0000 Subject: powerpc: Fix PCI in Holly device tree The PCI bridge on the Holly board is incorrectly represented in the device tree. The current device tree node for the PCI bridge sits under the tsi-bridge node. That's not obviously wrong, but the PCI bridge translates some PCI spaces into CPU address ranges which were not translated by the "ranges" property in tsi-bridge node. We used to get away with this problem because the PCI bridge discovery code was also buggy, assuming incorrectly that PCI host bridge nodes were always directly under the root bus and treating the translated addresses as raw CPU addresses, rather than parent bus addresses. This has since been fixed, thus breaking Holly. This could be fixed by adding extra translations to the tsi-bridge node, but this patch instead moves the Holly PCI bridge out of the tsi-bridge node to the root bus. This makes the tsi-bridge node represent only the built-in IO devices in the bridge, with a more-or-less contiguous address range. This is the same convention used on Freescale SoC chips, where the "soc" node represents only the IMMR region, and the PCI and other bus bridges are separate nodes under the root bus. Signed-off-by: David Gibson Acked-by: Josh Boyer Signed-off-by: Benjamin Herrenschmidt --- arch/powerpc/boot/dts/holly.dts | 106 ++++++++++++++++++++-------------------- 1 file changed, 53 insertions(+), 53 deletions(-) (limited to 'arch') diff --git a/arch/powerpc/boot/dts/holly.dts b/arch/powerpc/boot/dts/holly.dts index f87fe7b9ced9..c6e11ebecebb 100644 --- a/arch/powerpc/boot/dts/holly.dts +++ b/arch/powerpc/boot/dts/holly.dts @@ -133,61 +133,61 @@ reg = <0x00007400 0x00000400>; big-endian; }; + }; - pci@1000 { - device_type = "pci"; - compatible = "tsi109-pci", "tsi108-pci"; - #interrupt-cells = <1>; - #size-cells = <2>; - #address-cells = <3>; - reg = <0x00001000 0x00001000>; - bus-range = <0x0 0x0>; - /*----------------------------------------------------+ - | PCI memory range. - | 01 denotes I/O space - | 02 denotes 32-bit memory space - +----------------------------------------------------*/ - ranges = <0x02000000 0x00000000 0x40000000 0x40000000 0x00000000 0x10000000 - 0x01000000 0x00000000 0x00000000 0x7e000000 0x00000000 0x00010000>; - clock-frequency = <133333332>; - interrupt-parent = <&MPIC>; + pci@c0001000 { + device_type = "pci"; + compatible = "tsi109-pci", "tsi108-pci"; + #interrupt-cells = <1>; + #size-cells = <2>; + #address-cells = <3>; + reg = <0xc0001000 0x00001000>; + bus-range = <0x0 0x0>; + /*----------------------------------------------------+ + | PCI memory range. + | 01 denotes I/O space + | 02 denotes 32-bit memory space + +----------------------------------------------------*/ + ranges = <0x02000000 0x00000000 0x40000000 0x40000000 0x00000000 0x10000000 + 0x01000000 0x00000000 0x00000000 0x7e000000 0x00000000 0x00010000>; + clock-frequency = <133333332>; + interrupt-parent = <&MPIC>; + interrupts = <0x17 0x2>; + interrupt-map-mask = <0xf800 0x0 0x0 0x7>; + /*----------------------------------------------------+ + | The INTA, INTB, INTC, INTD are shared. + +----------------------------------------------------*/ + interrupt-map = < + 0x800 0x0 0x0 0x1 &RT0 0x24 0x0 + 0x800 0x0 0x0 0x2 &RT0 0x25 0x0 + 0x800 0x0 0x0 0x3 &RT0 0x26 0x0 + 0x800 0x0 0x0 0x4 &RT0 0x27 0x0 + + 0x1000 0x0 0x0 0x1 &RT0 0x25 0x0 + 0x1000 0x0 0x0 0x2 &RT0 0x26 0x0 + 0x1000 0x0 0x0 0x3 &RT0 0x27 0x0 + 0x1000 0x0 0x0 0x4 &RT0 0x24 0x0 + + 0x1800 0x0 0x0 0x1 &RT0 0x26 0x0 + 0x1800 0x0 0x0 0x2 &RT0 0x27 0x0 + 0x1800 0x0 0x0 0x3 &RT0 0x24 0x0 + 0x1800 0x0 0x0 0x4 &RT0 0x25 0x0 + + 0x2000 0x0 0x0 0x1 &RT0 0x27 0x0 + 0x2000 0x0 0x0 0x2 &RT0 0x24 0x0 + 0x2000 0x0 0x0 0x3 &RT0 0x25 0x0 + 0x2000 0x0 0x0 0x4 &RT0 0x26 0x0 + >; + + RT0: router@1180 { + device_type = "pic-router"; + interrupt-controller; + big-endian; + clock-frequency = <0>; + #address-cells = <0>; + #interrupt-cells = <2>; interrupts = <0x17 0x2>; - interrupt-map-mask = <0xf800 0x0 0x0 0x7>; - /*----------------------------------------------------+ - | The INTA, INTB, INTC, INTD are shared. - +----------------------------------------------------*/ - interrupt-map = < - 0x800 0x0 0x0 0x1 &RT0 0x24 0x0 - 0x800 0x0 0x0 0x2 &RT0 0x25 0x0 - 0x800 0x0 0x0 0x3 &RT0 0x26 0x0 - 0x800 0x0 0x0 0x4 &RT0 0x27 0x0 - - 0x1000 0x0 0x0 0x1 &RT0 0x25 0x0 - 0x1000 0x0 0x0 0x2 &RT0 0x26 0x0 - 0x1000 0x0 0x0 0x3 &RT0 0x27 0x0 - 0x1000 0x0 0x0 0x4 &RT0 0x24 0x0 - - 0x1800 0x0 0x0 0x1 &RT0 0x26 0x0 - 0x1800 0x0 0x0 0x2 &RT0 0x27 0x0 - 0x1800 0x0 0x0 0x3 &RT0 0x24 0x0 - 0x1800 0x0 0x0 0x4 &RT0 0x25 0x0 - - 0x2000 0x0 0x0 0x1 &RT0 0x27 0x0 - 0x2000 0x0 0x0 0x2 &RT0 0x24 0x0 - 0x2000 0x0 0x0 0x3 &RT0 0x25 0x0 - 0x2000 0x0 0x0 0x4 &RT0 0x26 0x0 - >; - - RT0: router@1180 { - device_type = "pic-router"; - interrupt-controller; - big-endian; - clock-frequency = <0>; - #address-cells = <0>; - #interrupt-cells = <2>; - interrupts = <0x17 0x2>; - interrupt-parent = <&MPIC>; - }; + interrupt-parent = <&MPIC>; }; }; -- cgit v1.2.3 From 61e9916eba35dfb76d38013a5aae9a59cc50877a Mon Sep 17 00:00:00 2001 From: Johannes Berg Date: Wed, 24 Sep 2008 22:56:25 +0000 Subject: powerpc: Fix failure to shutdown with CPU hotplug I tracked down the shutdown regression to CPUs not dying when being shut down during power-off. This turns out to be due to the system_state being SYSTEM_POWER_OFF, which this code doesn't take as a valid state for shutting off CPUs in. This has never made sense to me, but when I added hotplug code to implement hibernate I only "made it work" and did not question the need to check the system_state. Thomas Gleixner helped me dig, but the only thing we found is that it was added with the original commit that added CPU hotplug support. Signed-off-by: Johannes Berg Acked-by: Joel Schopp Signed-off-by: Benjamin Herrenschmidt --- arch/powerpc/kernel/idle.c | 6 +----- 1 file changed, 1 insertion(+), 5 deletions(-) (limited to 'arch') diff --git a/arch/powerpc/kernel/idle.c b/arch/powerpc/kernel/idle.c index d308a9f70f1b..31982d05d81a 100644 --- a/arch/powerpc/kernel/idle.c +++ b/arch/powerpc/kernel/idle.c @@ -34,11 +34,7 @@ #include #ifdef CONFIG_HOTPLUG_CPU -/* this is used for software suspend, and that shuts down - * CPUs even while the system is still booting... */ -#define cpu_should_die() (cpu_is_offline(smp_processor_id()) && \ - (system_state == SYSTEM_RUNNING \ - || system_state == SYSTEM_BOOTING)) +#define cpu_should_die() cpu_is_offline(smp_processor_id()) #else #define cpu_should_die() 0 #endif -- cgit v1.2.3 From 3dcd7e269d2223126f6ee9bc893f5a6166e1770d Mon Sep 17 00:00:00 2001 From: "J.A. Magallón" Date: Tue, 30 Sep 2008 10:02:52 +0200 Subject: x86: fix typo in enable_mtrr_cleanup early parameter Correct typo for 'enable_mtrr_cleanup' early boot param name. Signed-off-by: J.A. Magallon Signed-off-by: Ingo Molnar --- arch/x86/kernel/cpu/mtrr/main.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'arch') diff --git a/arch/x86/kernel/cpu/mtrr/main.c b/arch/x86/kernel/cpu/mtrr/main.c index b117d7f8a564..885c8265e6b5 100644 --- a/arch/x86/kernel/cpu/mtrr/main.c +++ b/arch/x86/kernel/cpu/mtrr/main.c @@ -834,7 +834,7 @@ static int __init enable_mtrr_cleanup_setup(char *str) enable_mtrr_cleanup = 1; return 0; } -early_param("enble_mtrr_cleanup", enable_mtrr_cleanup_setup); +early_param("enable_mtrr_cleanup", enable_mtrr_cleanup_setup); struct var_mtrr_state { unsigned long range_startk; -- cgit v1.2.3 From de59985e3a623d4d5d6207f1777398ca0606ab1c Mon Sep 17 00:00:00 2001 From: Zachary Amsden Date: Tue, 30 Sep 2008 11:02:12 -0700 Subject: x86: Fix broken LDT access in VMI After investigating a JRE failure, I found this bug was introduced a long time ago, and had already managed to survive another bugfix which occurred on the same line. The result is a total failure of the JRE due to LDT selectors not working properly. This one took a long time to rear up because LDT usage is not very common, but the bug is quite serious. It got introduced along with another bug, already fixed, by 75b8bb3e56ca09a467fbbe5229bc68627f7445be Signed-off-by: Zachary Amsden Cc: Ingo Molnar Cc: Glauber de Oliveira Costa Cc: Signed-off-by: Linus Torvalds --- arch/x86/kernel/vmi_32.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'arch') diff --git a/arch/x86/kernel/vmi_32.c b/arch/x86/kernel/vmi_32.c index 6ca515d6db54..edfb09f30479 100644 --- a/arch/x86/kernel/vmi_32.c +++ b/arch/x86/kernel/vmi_32.c @@ -235,7 +235,7 @@ static void vmi_write_ldt_entry(struct desc_struct *dt, int entry, const void *desc) { u32 *ldt_entry = (u32 *)desc; - vmi_ops.write_idt_entry(dt, entry, ldt_entry[0], ldt_entry[1]); + vmi_ops.write_ldt_entry(dt, entry, ldt_entry[0], ldt_entry[1]); } static void vmi_load_sp0(struct tss_struct *tss, -- cgit v1.2.3 From dc63b52673d71f9d49b9d72d263a9f32df18c3ee Mon Sep 17 00:00:00 2001 From: Zachary Amsden Date: Tue, 30 Sep 2008 11:02:12 -0700 Subject: x86, vmi: fix broken LDT access This one took a long time to rear up because LDT usage is not very common, but the bug is quite serious. It got introduced along with another bug, already fixed, by 75b8bb3e56ca09a467fbbe5229bc68627f7445be After investigating a JRE failure, I found this bug was introduced a long time ago, and had already managed to survive another bugfix which occurred on the same line. The result is a total failure of the JRE due to LDT selectors not working properly. Signed-off-by: Zachary Amsden Cc: Glauber de Oliveira Costa Cc: stable@kernel.org Signed-off-by: Ingo Molnar --- arch/x86/kernel/vmi_32.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'arch') diff --git a/arch/x86/kernel/vmi_32.c b/arch/x86/kernel/vmi_32.c index 6ca515d6db54..edfb09f30479 100644 --- a/arch/x86/kernel/vmi_32.c +++ b/arch/x86/kernel/vmi_32.c @@ -235,7 +235,7 @@ static void vmi_write_ldt_entry(struct desc_struct *dt, int entry, const void *desc) { u32 *ldt_entry = (u32 *)desc; - vmi_ops.write_idt_entry(dt, entry, ldt_entry[0], ldt_entry[1]); + vmi_ops.write_ldt_entry(dt, entry, ldt_entry[0], ldt_entry[1]); } static void vmi_load_sp0(struct tss_struct *tss, -- cgit v1.2.3 From d6478fad430e37148b56f642c87301ba72476675 Mon Sep 17 00:00:00 2001 From: David Howells Date: Wed, 1 Oct 2008 13:47:06 +0100 Subject: MN10300: Fix IRQ handling Fix the IRQ handling on the MN10300 arch. This patch makes a number of significant changes: (1) It separates the irq_chip definition for edge-triggered interrupts from the one for level-triggered interrupts. This is necessary because the MN10300 PIC latches the IRQ channel's interrupt request bit (GxICR_REQUEST), even after the device has ceased to assert its interrupt line and the interrupt channel has been disabled in the PIC. So for level-triggered interrupts we need to clear this bit when we re-enable - which is achieved by setting GxICR_DETECT but not GxICR_REQUEST when writing to the register. Not doing this results in spurious interrupts occurring because calling mask_ack() at the start of handle_level_irq() is insufficient - it fails to clear the REQUEST latch because the device that caused the interrupt is still asserting its interrupt line at this point. (2) IRQ disablement [irq_chip::disable_irq()] shouldn't clear the interrupt request flag for edge-triggered interrupts lest it lose an interrupt. (3) IRQ unmasking [irq_chip::unmask_irq()] also shouldn't clear the interrupt request flag for edge-triggered interrupts lest it lose an interrupt. (4) The end() operation is now left to the default (no-operation) as __do_IRQ() is compiled out. This may affect misrouted_irq(), but according to Thomas Gleixner it's the correct thing to do. (5) handle_level_irq() is used for edge-triggered interrupts rather than handle_edge_irq() as the MN10300 PIC latches interrupt events even on masked IRQ channels, thus rendering IRQ_PENDING unnecessary. It is sufficient to call mask_ack() at the start and unmask() at the end. (6) For level-triggered interrupts, ack() is now NULL as it's not used, and there is no effective ACK function on the PIC. mask_ack() is now the same as mask() as the latch continues to latch, even when the channel is masked. Further, the patch discards the disable() op implementation as its now the same as the mask() op implementation, which is used instead. It also discards the enable() op implementations as they're now the same as the unmask() op implementations, which are used instead. Signed-off-by: David Howells Signed-off-by: Linus Torvalds --- arch/mn10300/kernel/irq.c | 71 ++++++++++++++++++++++------------- arch/mn10300/unit-asb2303/unit-init.c | 2 +- arch/mn10300/unit-asb2305/unit-init.c | 2 +- 3 files changed, 47 insertions(+), 28 deletions(-) (limited to 'arch') diff --git a/arch/mn10300/kernel/irq.c b/arch/mn10300/kernel/irq.c index 761c434a2488..56c64ccc9c21 100644 --- a/arch/mn10300/kernel/irq.c +++ b/arch/mn10300/kernel/irq.c @@ -20,22 +20,8 @@ EXPORT_SYMBOL(__mn10300_irq_enabled_epsw); atomic_t irq_err_count; /* - * MN10300 INTC controller operations + * MN10300 interrupt controller operations */ -static void mn10300_cpupic_disable(unsigned int irq) -{ - u16 tmp = GxICR(irq); - GxICR(irq) = (tmp & GxICR_LEVEL) | GxICR_DETECT; - tmp = GxICR(irq); -} - -static void mn10300_cpupic_enable(unsigned int irq) -{ - u16 tmp = GxICR(irq); - GxICR(irq) = (tmp & GxICR_LEVEL) | GxICR_ENABLE; - tmp = GxICR(irq); -} - static void mn10300_cpupic_ack(unsigned int irq) { u16 tmp; @@ -60,26 +46,54 @@ static void mn10300_cpupic_mask_ack(unsigned int irq) static void mn10300_cpupic_unmask(unsigned int irq) { u16 tmp = GxICR(irq); - GxICR(irq) = (tmp & GxICR_LEVEL) | GxICR_ENABLE | GxICR_DETECT; + GxICR(irq) = (tmp & GxICR_LEVEL) | GxICR_ENABLE; tmp = GxICR(irq); } -static void mn10300_cpupic_end(unsigned int irq) +static void mn10300_cpupic_unmask_clear(unsigned int irq) { + /* the MN10300 PIC latches its interrupt request bit, even after the + * device has ceased to assert its interrupt line and the interrupt + * channel has been disabled in the PIC, so for level-triggered + * interrupts we need to clear the request bit when we re-enable */ u16 tmp = GxICR(irq); - GxICR(irq) = (tmp & GxICR_LEVEL) | GxICR_ENABLE; + GxICR(irq) = (tmp & GxICR_LEVEL) | GxICR_ENABLE | GxICR_DETECT; tmp = GxICR(irq); } -static struct irq_chip mn10300_cpu_pic = { - .name = "cpu", - .disable = mn10300_cpupic_disable, - .enable = mn10300_cpupic_enable, +/* + * MN10300 PIC level-triggered IRQ handling. + * + * The PIC has no 'ACK' function per se. It is possible to clear individual + * channel latches, but each latch relatches whether or not the channel is + * masked, so we need to clear the latch when we unmask the channel. + * + * Also for this reason, we don't supply an ack() op (it's unused anyway if + * mask_ack() is provided), and mask_ack() just masks. + */ +static struct irq_chip mn10300_cpu_pic_level = { + .name = "cpu_l", + .disable = mn10300_cpupic_mask, + .enable = mn10300_cpupic_unmask_clear, + .ack = NULL, + .mask = mn10300_cpupic_mask, + .mask_ack = mn10300_cpupic_mask, + .unmask = mn10300_cpupic_unmask_clear, +}; + +/* + * MN10300 PIC edge-triggered IRQ handling. + * + * We use the latch clearing function of the PIC as the 'ACK' function. + */ +static struct irq_chip mn10300_cpu_pic_edge = { + .name = "cpu_e", + .disable = mn10300_cpupic_mask, + .enable = mn10300_cpupic_unmask, .ack = mn10300_cpupic_ack, .mask = mn10300_cpupic_mask, .mask_ack = mn10300_cpupic_mask_ack, .unmask = mn10300_cpupic_unmask, - .end = mn10300_cpupic_end, }; /* @@ -114,7 +128,8 @@ void set_intr_level(int irq, u16 level) */ void set_intr_postackable(int irq) { - set_irq_handler(irq, handle_level_irq); + set_irq_chip_and_handler(irq, &mn10300_cpu_pic_level, + handle_level_irq); } /* @@ -126,8 +141,12 @@ void __init init_IRQ(void) for (irq = 0; irq < NR_IRQS; irq++) if (irq_desc[irq].chip == &no_irq_type) - set_irq_chip_and_handler(irq, &mn10300_cpu_pic, - handle_edge_irq); + /* due to the PIC latching interrupt requests, even + * when the IRQ is disabled, IRQ_PENDING is superfluous + * and we can use handle_level_irq() for edge-triggered + * interrupts */ + set_irq_chip_and_handler(irq, &mn10300_cpu_pic_edge, + handle_level_irq); unit_init_IRQ(); } diff --git a/arch/mn10300/unit-asb2303/unit-init.c b/arch/mn10300/unit-asb2303/unit-init.c index 14b2c817cff8..70e8cb4ea266 100644 --- a/arch/mn10300/unit-asb2303/unit-init.c +++ b/arch/mn10300/unit-asb2303/unit-init.c @@ -51,7 +51,7 @@ void __init unit_init_IRQ(void) switch (GET_XIRQ_TRIGGER(extnum)) { case XIRQ_TRIGGER_HILEVEL: case XIRQ_TRIGGER_LOWLEVEL: - set_irq_handler(XIRQ2IRQ(extnum), handle_level_irq); + set_intr_postackable(XIRQ2IRQ(extnum)); break; default: break; diff --git a/arch/mn10300/unit-asb2305/unit-init.c b/arch/mn10300/unit-asb2305/unit-init.c index 6a352414a358..72812a9439ac 100644 --- a/arch/mn10300/unit-asb2305/unit-init.c +++ b/arch/mn10300/unit-asb2305/unit-init.c @@ -52,7 +52,7 @@ void __init unit_init_IRQ(void) switch (GET_XIRQ_TRIGGER(extnum)) { case XIRQ_TRIGGER_HILEVEL: case XIRQ_TRIGGER_LOWLEVEL: - set_irq_handler(XIRQ2IRQ(extnum), handle_level_irq); + set_intr_postackable(XIRQ2IRQ(extnum)); break; default: break; -- cgit v1.2.3 From 1fce2d01dff65a76cd08c1b145acd9d3c20021d2 Mon Sep 17 00:00:00 2001 From: Kumar Gala Date: Thu, 2 Oct 2008 00:58:49 -0500 Subject: powerpc: Fix boot hang regression on MPC8544DS Commit 00c5372d37a78990c1530184a9c792ee60a30067 caused the MPC8544DS board to hang at boot. The MPC8544DS is unique in that it doesn't use the PCI slots on the ULI (unlike the MPC8572DS or MPC8610HPCD). So the dummy read at the end of the address space causes us to hang. We can detect the situation by comparing the bridge's BARs versus the root complex. Signed-off-by: Kumar Gala --- arch/powerpc/platforms/fsl_uli1575.c | 12 +++++++++++- 1 file changed, 11 insertions(+), 1 deletion(-) (limited to 'arch') diff --git a/arch/powerpc/platforms/fsl_uli1575.c b/arch/powerpc/platforms/fsl_uli1575.c index ef74a0763ec1..8c619963becc 100644 --- a/arch/powerpc/platforms/fsl_uli1575.c +++ b/arch/powerpc/platforms/fsl_uli1575.c @@ -219,11 +219,21 @@ static void __devinit quirk_final_uli5249(struct pci_dev *dev) int i; u8 *dummy; struct pci_bus *bus = dev->bus; + resource_size_t end = 0; + + for (i = PCI_BRIDGE_RESOURCES; i < PCI_BRIDGE_RESOURCES+3; i++) { + unsigned long flags = pci_resource_flags(dev, i); + if ((flags & (IORESOURCE_MEM|IORESOURCE_PREFETCH)) == IORESOURCE_MEM) + end = pci_resource_end(dev, i); + } for (i = 0; i < PCI_BUS_NUM_RESOURCES; i++) { if ((bus->resource[i]) && (bus->resource[i]->flags & IORESOURCE_MEM)) { - dummy = ioremap(bus->resource[i]->end - 3, 0x4); + if (bus->resource[i]->end == end) + dummy = ioremap(bus->resource[i]->start, 0x4); + else + dummy = ioremap(bus->resource[i]->end - 3, 0x4); if (dummy) { in_8(dummy); iounmap(dummy); -- cgit v1.2.3 From b7e4226e4f427b59dc8e9c45a2a1a1ed1353a140 Mon Sep 17 00:00:00 2001 From: Ralf Baechle Date: Wed, 1 Oct 2008 21:52:41 +0100 Subject: [MIPS] Build fix: Fix irq flags type Though from a hardware perspective it would be sensible to use only a 32-bit unsigned int type Linux defines interrupt flags to be stored in an unsigned long and nothing else. Signed-off-by: Ralf Baechle --- arch/mips/kernel/ptrace.c | 2 +- arch/mips/kernel/smtc.c | 6 +++--- include/asm-mips/mipsregs.h | 6 +++--- 3 files changed, 7 insertions(+), 7 deletions(-) (limited to 'arch') diff --git a/arch/mips/kernel/ptrace.c b/arch/mips/kernel/ptrace.c index 35234b92b9a5..96ffc9c6d194 100644 --- a/arch/mips/kernel/ptrace.c +++ b/arch/mips/kernel/ptrace.c @@ -238,7 +238,7 @@ long arch_ptrace(struct task_struct *child, long request, long addr, long data) case FPC_EIR: { /* implementation / version register */ unsigned int flags; #ifdef CONFIG_MIPS_MT_SMTC - unsigned int irqflags; + unsigned long irqflags; unsigned int mtflags; #endif /* CONFIG_MIPS_MT_SMTC */ diff --git a/arch/mips/kernel/smtc.c b/arch/mips/kernel/smtc.c index a516286532ab..05f2708a9029 100644 --- a/arch/mips/kernel/smtc.c +++ b/arch/mips/kernel/smtc.c @@ -556,7 +556,7 @@ void mipsmt_prepare_cpus(void) void __cpuinit smtc_boot_secondary(int cpu, struct task_struct *idle) { extern u32 kernelsp[NR_CPUS]; - long flags; + unsigned long flags; int mtflags; LOCK_MT_PRA(); @@ -753,7 +753,7 @@ void smtc_send_ipi(int cpu, int type, unsigned int action) { int tcstatus; struct smtc_ipi *pipi; - long flags; + unsigned long flags; int mtflags; if (cpu == smp_processor_id()) { @@ -975,7 +975,7 @@ static irqreturn_t ipi_interrupt(int irq, void *dev_idm) struct smtc_ipi *pipi; unsigned long tcstatus; int sent; - long flags; + unsigned long flags; unsigned int mtflags; unsigned int vpflags; diff --git a/include/asm-mips/mipsregs.h b/include/asm-mips/mipsregs.h index a46f8e258e6b..979866000da4 100644 --- a/include/asm-mips/mipsregs.h +++ b/include/asm-mips/mipsregs.h @@ -1462,7 +1462,7 @@ set_c0_##name(unsigned int set) \ { \ unsigned int res; \ unsigned int omt; \ - unsigned int flags; \ + unsigned long flags; \ \ local_irq_save(flags); \ omt = __dmt(); \ @@ -1480,7 +1480,7 @@ clear_c0_##name(unsigned int clear) \ { \ unsigned int res; \ unsigned int omt; \ - unsigned int flags; \ + unsigned long flags; \ \ local_irq_save(flags); \ omt = __dmt(); \ @@ -1498,7 +1498,7 @@ change_c0_##name(unsigned int change, unsigned int new) \ { \ unsigned int res; \ unsigned int omt; \ - unsigned int flags; \ + unsigned long flags; \ \ local_irq_save(flags); \ \ -- cgit v1.2.3 From 498a863fdfb3e60d2e0e964cd202c2030bda1005 Mon Sep 17 00:00:00 2001 From: Ralf Baechle Date: Wed, 1 Oct 2008 22:23:52 +0100 Subject: [MIPS] SMTC: Build fix: Fix filename in Makefile Signed-off-by: Ralf Baechle --- arch/mips/mti-malta/Makefile | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'arch') diff --git a/arch/mips/mti-malta/Makefile b/arch/mips/mti-malta/Makefile index 3b7dd722c32a..cef2db8d2225 100644 --- a/arch/mips/mti-malta/Makefile +++ b/arch/mips/mti-malta/Makefile @@ -15,6 +15,6 @@ obj-$(CONFIG_EARLY_PRINTK) += malta-console.o obj-$(CONFIG_PCI) += malta-pci.o # FIXME FIXME FIXME -obj-$(CONFIG_MIPS_MT_SMTC) += malta_smtc.o +obj-$(CONFIG_MIPS_MT_SMTC) += malta-smtc.o EXTRA_CFLAGS += -Werror -- cgit v1.2.3 From 9cc123631b6630948eb5059d218f44424888daa7 Mon Sep 17 00:00:00 2001 From: "Kevin D. Kissell" Date: Tue, 9 Sep 2008 21:33:36 +0200 Subject: [MIPS] SMTC: Fix holes in SMTC and FPU affinity support. Signed-off-by: Kevin D. Kissell Signed-off-by: Ralf Baechle --- arch/mips/kernel/mips-mt-fpaff.c | 2 +- arch/mips/kernel/process.c | 19 +++++++++---------- arch/mips/kernel/traps.c | 6 ++++-- 3 files changed, 14 insertions(+), 13 deletions(-) (limited to 'arch') diff --git a/arch/mips/kernel/mips-mt-fpaff.c b/arch/mips/kernel/mips-mt-fpaff.c index df4d3f2f740c..dc9eb72ed9de 100644 --- a/arch/mips/kernel/mips-mt-fpaff.c +++ b/arch/mips/kernel/mips-mt-fpaff.c @@ -159,7 +159,7 @@ __setup("fpaff=", fpaff_thresh); /* * FPU Use Factor empirically derived from experiments on 34K */ -#define FPUSEFACTOR 333 +#define FPUSEFACTOR 2000 static __init int mt_fp_affinity_init(void) { diff --git a/arch/mips/kernel/process.c b/arch/mips/kernel/process.c index ce7684335a41..22fc19bbe87f 100644 --- a/arch/mips/kernel/process.c +++ b/arch/mips/kernel/process.c @@ -55,7 +55,7 @@ void __noreturn cpu_idle(void) while (1) { tick_nohz_stop_sched_tick(1); while (!need_resched()) { -#ifdef CONFIG_SMTC_IDLE_HOOK_DEBUG +#ifdef CONFIG_MIPS_MT_SMTC extern void smtc_idle_loop_hook(void); smtc_idle_loop_hook(); @@ -145,19 +145,18 @@ int copy_thread(int nr, unsigned long clone_flags, unsigned long usp, */ p->thread.cp0_status = read_c0_status() & ~(ST0_CU2|ST0_CU1); childregs->cp0_status &= ~(ST0_CU2|ST0_CU1); + +#ifdef CONFIG_MIPS_MT_SMTC + /* + * SMTC restores TCStatus after Status, and the CU bits + * are aliased there. + */ + childregs->cp0_tcstatus &= ~(ST0_CU2|ST0_CU1); +#endif clear_tsk_thread_flag(p, TIF_USEDFPU); #ifdef CONFIG_MIPS_MT_FPAFF clear_tsk_thread_flag(p, TIF_FPUBOUND); - - /* - * FPU affinity support is cleaner if we track the - * user-visible CPU affinity from the very beginning. - * The generic cpus_allowed mask will already have - * been copied from the parent before copy_thread - * is invoked. - */ - p->thread.user_cpus_allowed = p->cpus_allowed; #endif /* CONFIG_MIPS_MT_FPAFF */ if (clone_flags & CLONE_SETTLS) diff --git a/arch/mips/kernel/traps.c b/arch/mips/kernel/traps.c index 5fd0cd020af5..b602ac6eb47d 100644 --- a/arch/mips/kernel/traps.c +++ b/arch/mips/kernel/traps.c @@ -825,8 +825,10 @@ static void mt_ase_fp_affinity(void) if (cpus_intersects(current->cpus_allowed, mt_fpu_cpumask)) { cpumask_t tmask; - cpus_and(tmask, current->thread.user_cpus_allowed, - mt_fpu_cpumask); + current->thread.user_cpus_allowed + = current->cpus_allowed; + cpus_and(tmask, current->cpus_allowed, + mt_fpu_cpumask); set_cpus_allowed(current, tmask); set_thread_flag(TIF_FPUBOUND); } -- cgit v1.2.3 From d2bb01b042a38219fbddaafc214c5beb96248d2f Mon Sep 17 00:00:00 2001 From: "Kevin D. Kissell" Date: Tue, 9 Sep 2008 21:35:01 +0200 Subject: [MIPS] SMTC: Close tiny holes in the SMTC IPI replay system. Signed-off-by: Kevin D. Kissell Signed-off-by: Ralf Baechle --- arch/mips/kernel/entry.S | 10 +++--- arch/mips/kernel/smtc.c | 2 +- include/asm-mips/stackframe.h | 72 ++++++++++++++++++++++++++++++++++++------- 3 files changed, 67 insertions(+), 17 deletions(-) (limited to 'arch') diff --git a/arch/mips/kernel/entry.S b/arch/mips/kernel/entry.S index e29598ae939d..ffa331029e08 100644 --- a/arch/mips/kernel/entry.S +++ b/arch/mips/kernel/entry.S @@ -79,11 +79,6 @@ FEXPORT(syscall_exit) FEXPORT(restore_all) # restore full frame #ifdef CONFIG_MIPS_MT_SMTC -/* Detect and execute deferred IPI "interrupts" */ - LONG_L s0, TI_REGS($28) - LONG_S sp, TI_REGS($28) - jal deferred_smtc_ipi - LONG_S s0, TI_REGS($28) #ifdef CONFIG_MIPS_MT_SMTC_IM_BACKSTOP /* Re-arm any temporarily masked interrupts not explicitly "acked" */ mfc0 v0, CP0_TCSTATUS @@ -112,6 +107,11 @@ FEXPORT(restore_all) # restore full frame xor t0, t0, t3 mtc0 t0, CP0_TCCONTEXT #endif /* CONFIG_MIPS_MT_SMTC_IM_BACKSTOP */ +/* Detect and execute deferred IPI "interrupts" */ + LONG_L s0, TI_REGS($28) + LONG_S sp, TI_REGS($28) + jal deferred_smtc_ipi + LONG_S s0, TI_REGS($28) #endif /* CONFIG_MIPS_MT_SMTC */ .set noat RESTORE_TEMP diff --git a/arch/mips/kernel/smtc.c b/arch/mips/kernel/smtc.c index 05f2708a9029..39b491b9ad87 100644 --- a/arch/mips/kernel/smtc.c +++ b/arch/mips/kernel/smtc.c @@ -70,7 +70,7 @@ static atomic_t ipi_timer_latch[NR_CPUS]; #define IPIBUF_PER_CPU 4 -static struct smtc_ipi_q IPIQ[NR_CPUS]; +struct smtc_ipi_q IPIQ[NR_CPUS]; static struct smtc_ipi_q freeIPIq; diff --git a/include/asm-mips/stackframe.h b/include/asm-mips/stackframe.h index 051e1af0bb95..4c37c4e5f72e 100644 --- a/include/asm-mips/stackframe.h +++ b/include/asm-mips/stackframe.h @@ -297,14 +297,31 @@ #ifdef CONFIG_MIPS_MT_SMTC .set mips32r2 /* - * This may not really be necessary if ints are already - * inhibited here. + * We need to make sure the read-modify-write + * of Status below isn't perturbed by an interrupt + * or cross-TC access, so we need to do at least a DMT, + * protected by an interrupt-inhibit. But setting IXMT + * also creates a few-cycle window where an IPI could + * be queued and not be detected before potentially + * returning to a WAIT or user-mode loop. It must be + * replayed. + * + * We're in the middle of a context switch, and + * we can't dispatch it directly without trashing + * some registers, so we'll try to detect this unlikely + * case and program a software interrupt in the VPE, + * as would be done for a cross-VPE IPI. To accomodate + * the handling of that case, we're doing a DVPE instead + * of just a DMT here to protect against other threads. + * This is a lot of cruft to cover a tiny window. + * If you can find a better design, implement it! + * */ mfc0 v0, CP0_TCSTATUS ori v0, TCSTATUS_IXMT mtc0 v0, CP0_TCSTATUS _ehb - DMT 5 # dmt a1 + DVPE 5 # dvpe a1 jal mips_ihb #endif /* CONFIG_MIPS_MT_SMTC */ mfc0 a0, CP0_STATUS @@ -325,17 +342,50 @@ */ LONG_L v1, PT_TCSTATUS(sp) _ehb - mfc0 v0, CP0_TCSTATUS + mfc0 a0, CP0_TCSTATUS andi v1, TCSTATUS_IXMT - /* We know that TCStatua.IXMT should be set from above */ - xori v0, v0, TCSTATUS_IXMT - or v0, v0, v1 - mtc0 v0, CP0_TCSTATUS - _ehb - andi a1, a1, VPECONTROL_TE + bnez v1, 0f + +/* + * We'd like to detect any IPIs queued in the tiny window + * above and request an software interrupt to service them + * when we ERET. + * + * Computing the offset into the IPIQ array of the executing + * TC's IPI queue in-line would be tedious. We use part of + * the TCContext register to hold 16 bits of offset that we + * can add in-line to find the queue head. + */ + mfc0 v0, CP0_TCCONTEXT + la a2, IPIQ + srl v0, v0, 16 + addu a2, a2, v0 + LONG_L v0, 0(a2) + beqz v0, 0f +/* + * If we have a queue, provoke dispatch within the VPE by setting C_SW1 + */ + mfc0 v0, CP0_CAUSE + ori v0, v0, C_SW1 + mtc0 v0, CP0_CAUSE +0: + /* + * This test should really never branch but + * let's be prudent here. Having atomized + * the shared register modifications, we can + * now EVPE, and must do so before interrupts + * are potentially re-enabled. + */ + andi a1, a1, MVPCONTROL_EVP beqz a1, 1f - emt + evpe 1: + /* We know that TCStatua.IXMT should be set from above */ + xori a0, a0, TCSTATUS_IXMT + or a0, a0, v1 + mtc0 a0, CP0_TCSTATUS + _ehb + .set mips0 #endif /* CONFIG_MIPS_MT_SMTC */ LONG_L v1, PT_EPC(sp) -- cgit v1.2.3 From 8531a35e5e275b17c57c39b7911bc2b37025f28c Mon Sep 17 00:00:00 2001 From: "Kevin D. Kissell" Date: Tue, 9 Sep 2008 21:48:52 +0200 Subject: [MIPS] SMTC: Fix SMTC dyntick support. Rework of SMTC support to make it work with the new clock event system, allowing "tickless" operation, and to make it compatible with the use of the "wait_irqoff" idle loop. The new clocking scheme means that the previously optional IPI instant replay mechanism is now required, and has been made more robust. Signed-off-by: Kevin D. Kissell Signed-off-by: Ralf Baechle --- arch/mips/Kconfig | 26 +--- arch/mips/kernel/Makefile | 1 + arch/mips/kernel/cevt-r4k.c | 173 ++++++--------------- arch/mips/kernel/cevt-smtc.c | 321 +++++++++++++++++++++++++++++++++++++++ arch/mips/kernel/cpu-probe.c | 10 +- arch/mips/kernel/genex.S | 4 +- arch/mips/kernel/smtc.c | 252 +++++++++++++++--------------- arch/mips/mti-malta/malta-smtc.c | 9 +- include/asm-mips/cevt-r4k.h | 46 ++++++ include/asm-mips/irqflags.h | 26 +++- include/asm-mips/smtc.h | 8 +- 11 files changed, 597 insertions(+), 279 deletions(-) create mode 100644 arch/mips/kernel/cevt-smtc.c create mode 100644 include/asm-mips/cevt-r4k.h (limited to 'arch') diff --git a/arch/mips/Kconfig b/arch/mips/Kconfig index 49896a2a1d72..c930b8ceb418 100644 --- a/arch/mips/Kconfig +++ b/arch/mips/Kconfig @@ -1403,7 +1403,6 @@ config MIPS_MT_SMTC depends on CPU_MIPS32_R2 #depends on CPU_MIPS64_R2 # once there is hardware ... depends on SYS_SUPPORTS_MULTITHREADING - select GENERIC_CLOCKEVENTS_BROADCAST select CPU_MIPSR2_IRQ_VI select CPU_MIPSR2_IRQ_EI select MIPS_MT @@ -1451,32 +1450,17 @@ config MIPS_VPE_LOADER Includes a loader for loading an elf relocatable object onto another VPE and running it. -config MIPS_MT_SMTC_INSTANT_REPLAY - bool "Low-latency Dispatch of Deferred SMTC IPIs" - depends on MIPS_MT_SMTC && !PREEMPT - default y - help - SMTC pseudo-interrupts between TCs are deferred and queued - if the target TC is interrupt-inhibited (IXMT). In the first - SMTC prototypes, these queued IPIs were serviced on return - to user mode, or on entry into the kernel idle loop. The - INSTANT_REPLAY option dispatches them as part of local_irq_restore() - processing, which adds runtime overhead (hence the option to turn - it off), but ensures that IPIs are handled promptly even under - heavy I/O interrupt load. - config MIPS_MT_SMTC_IM_BACKSTOP bool "Use per-TC register bits as backstop for inhibited IM bits" depends on MIPS_MT_SMTC - default y + default n help To support multiple TC microthreads acting as "CPUs" within a VPE, VPE-wide interrupt mask bits must be specially manipulated during interrupt handling. To support legacy drivers and interrupt controller management code, SMTC has a "backstop" to track and if necessary restore the interrupt mask. This has some performance - impact on interrupt service overhead. Disable it only if you know - what you are doing. + impact on interrupt service overhead. config MIPS_MT_SMTC_IRQAFF bool "Support IRQ affinity API" @@ -1486,10 +1470,8 @@ config MIPS_MT_SMTC_IRQAFF Enables SMP IRQ affinity API (/proc/irq/*/smp_affinity, etc.) for SMTC Linux kernel. Requires platform support, of which an example can be found in the MIPS kernel i8259 and Malta - platform code. It is recommended that MIPS_MT_SMTC_INSTANT_REPLAY - be enabled if MIPS_MT_SMTC_IRQAFF is used. Adds overhead to - interrupt dispatch, and should be used only if you know what - you are doing. + platform code. Adds some overhead to interrupt dispatch, and + should be used only if you know what you are doing. config MIPS_VPE_LOADER_TOM bool "Load VPE program into memory hidden from linux" diff --git a/arch/mips/kernel/Makefile b/arch/mips/kernel/Makefile index 706f93974797..25775cb54000 100644 --- a/arch/mips/kernel/Makefile +++ b/arch/mips/kernel/Makefile @@ -10,6 +10,7 @@ obj-y += cpu-probe.o branch.o entry.o genex.o irq.o process.o \ obj-$(CONFIG_CEVT_BCM1480) += cevt-bcm1480.o obj-$(CONFIG_CEVT_R4K) += cevt-r4k.o +obj-$(CONFIG_MIPS_MT_SMTC) += cevt-smtc.o obj-$(CONFIG_CEVT_DS1287) += cevt-ds1287.o obj-$(CONFIG_CEVT_GT641XX) += cevt-gt641xx.o obj-$(CONFIG_CEVT_SB1250) += cevt-sb1250.o diff --git a/arch/mips/kernel/cevt-r4k.c b/arch/mips/kernel/cevt-r4k.c index 24a2d907aa0d..4a4c59f2737a 100644 --- a/arch/mips/kernel/cevt-r4k.c +++ b/arch/mips/kernel/cevt-r4k.c @@ -12,6 +12,14 @@ #include #include +#include + +/* + * The SMTC Kernel for the 34K, 1004K, et. al. replaces several + * of these routines with SMTC-specific variants. + */ + +#ifndef CONFIG_MIPS_MT_SMTC static int mips_next_event(unsigned long delta, struct clock_event_device *evt) @@ -19,60 +27,27 @@ static int mips_next_event(unsigned long delta, unsigned int cnt; int res; -#ifdef CONFIG_MIPS_MT_SMTC - { - unsigned long flags, vpflags; - local_irq_save(flags); - vpflags = dvpe(); -#endif cnt = read_c0_count(); cnt += delta; write_c0_compare(cnt); res = ((int)(read_c0_count() - cnt) > 0) ? -ETIME : 0; -#ifdef CONFIG_MIPS_MT_SMTC - evpe(vpflags); - local_irq_restore(flags); - } -#endif return res; } -static void mips_set_mode(enum clock_event_mode mode, - struct clock_event_device *evt) +#endif /* CONFIG_MIPS_MT_SMTC */ + +void mips_set_clock_mode(enum clock_event_mode mode, + struct clock_event_device *evt) { /* Nothing to do ... */ } -static DEFINE_PER_CPU(struct clock_event_device, mips_clockevent_device); -static int cp0_timer_irq_installed; +DEFINE_PER_CPU(struct clock_event_device, mips_clockevent_device); +int cp0_timer_irq_installed; -/* - * Timer ack for an R4k-compatible timer of a known frequency. - */ -static void c0_timer_ack(void) -{ - write_c0_compare(read_c0_compare()); -} +#ifndef CONFIG_MIPS_MT_SMTC -/* - * Possibly handle a performance counter interrupt. - * Return true if the timer interrupt should not be checked - */ -static inline int handle_perf_irq(int r2) -{ - /* - * The performance counter overflow interrupt may be shared with the - * timer interrupt (cp0_perfcount_irq < 0). If it is and a - * performance counter has overflowed (perf_irq() == IRQ_HANDLED) - * and we can't reliably determine if a counter interrupt has also - * happened (!r2) then don't check for a timer interrupt. - */ - return (cp0_perfcount_irq < 0) && - perf_irq() == IRQ_HANDLED && - !r2; -} - -static irqreturn_t c0_compare_interrupt(int irq, void *dev_id) +irqreturn_t c0_compare_interrupt(int irq, void *dev_id) { const int r2 = cpu_has_mips_r2; struct clock_event_device *cd; @@ -93,12 +68,8 @@ static irqreturn_t c0_compare_interrupt(int irq, void *dev_id) * interrupt. Being the paranoiacs we are we check anyway. */ if (!r2 || (read_c0_cause() & (1 << 30))) { - c0_timer_ack(); -#ifdef CONFIG_MIPS_MT_SMTC - if (cpu_data[cpu].vpe_id) - goto out; - cpu = 0; -#endif + /* Clear Count/Compare Interrupt */ + write_c0_compare(read_c0_compare()); cd = &per_cpu(mips_clockevent_device, cpu); cd->event_handler(cd); } @@ -107,65 +78,16 @@ out: return IRQ_HANDLED; } -static struct irqaction c0_compare_irqaction = { +#endif /* Not CONFIG_MIPS_MT_SMTC */ + +struct irqaction c0_compare_irqaction = { .handler = c0_compare_interrupt, -#ifdef CONFIG_MIPS_MT_SMTC - .flags = IRQF_DISABLED, -#else .flags = IRQF_DISABLED | IRQF_PERCPU, -#endif .name = "timer", }; -#ifdef CONFIG_MIPS_MT_SMTC -DEFINE_PER_CPU(struct clock_event_device, smtc_dummy_clockevent_device); - -static void smtc_set_mode(enum clock_event_mode mode, - struct clock_event_device *evt) -{ -} - -static void mips_broadcast(cpumask_t mask) -{ - unsigned int cpu; - - for_each_cpu_mask(cpu, mask) - smtc_send_ipi(cpu, SMTC_CLOCK_TICK, 0); -} - -static void setup_smtc_dummy_clockevent_device(void) -{ - //uint64_t mips_freq = mips_hpt_^frequency; - unsigned int cpu = smp_processor_id(); - struct clock_event_device *cd; - cd = &per_cpu(smtc_dummy_clockevent_device, cpu); - - cd->name = "SMTC"; - cd->features = CLOCK_EVT_FEAT_DUMMY; - - /* Calculate the min / max delta */ - cd->mult = 0; //div_sc((unsigned long) mips_freq, NSEC_PER_SEC, 32); - cd->shift = 0; //32; - cd->max_delta_ns = 0; //clockevent_delta2ns(0x7fffffff, cd); - cd->min_delta_ns = 0; //clockevent_delta2ns(0x30, cd); - - cd->rating = 200; - cd->irq = 17; //-1; -// if (cpu) -// cd->cpumask = CPU_MASK_ALL; // cpumask_of_cpu(cpu); -// else - cd->cpumask = cpumask_of_cpu(cpu); - - cd->set_mode = smtc_set_mode; - - cd->broadcast = mips_broadcast; - - clockevents_register_device(cd); -} -#endif - -static void mips_event_handler(struct clock_event_device *dev) +void mips_event_handler(struct clock_event_device *dev) { } @@ -177,7 +99,23 @@ static int c0_compare_int_pending(void) return (read_c0_cause() >> cp0_compare_irq) & 0x100; } -static int c0_compare_int_usable(void) +/* + * Compare interrupt can be routed and latched outside the core, + * so a single execution hazard barrier may not be enough to give + * it time to clear as seen in the Cause register. 4 time the + * pipeline depth seems reasonably conservative, and empirically + * works better in configurations with high CPU/bus clock ratios. + */ + +#define compare_change_hazard() \ + do { \ + irq_disable_hazard(); \ + irq_disable_hazard(); \ + irq_disable_hazard(); \ + irq_disable_hazard(); \ + } while (0) + +int c0_compare_int_usable(void) { unsigned int delta; unsigned int cnt; @@ -187,7 +125,7 @@ static int c0_compare_int_usable(void) */ if (c0_compare_int_pending()) { write_c0_compare(read_c0_count()); - irq_disable_hazard(); + compare_change_hazard(); if (c0_compare_int_pending()) return 0; } @@ -196,7 +134,7 @@ static int c0_compare_int_usable(void) cnt = read_c0_count(); cnt += delta; write_c0_compare(cnt); - irq_disable_hazard(); + compare_change_hazard(); if ((int)(read_c0_count() - cnt) < 0) break; /* increase delta if the timer was already expired */ @@ -205,11 +143,12 @@ static int c0_compare_int_usable(void) while ((int)(read_c0_count() - cnt) <= 0) ; /* Wait for expiry */ + compare_change_hazard(); if (!c0_compare_int_pending()) return 0; write_c0_compare(read_c0_count()); - irq_disable_hazard(); + compare_change_hazard(); if (c0_compare_int_pending()) return 0; @@ -219,6 +158,8 @@ static int c0_compare_int_usable(void) return 1; } +#ifndef CONFIG_MIPS_MT_SMTC + int __cpuinit mips_clockevent_init(void) { uint64_t mips_freq = mips_hpt_frequency; @@ -229,17 +170,6 @@ int __cpuinit mips_clockevent_init(void) if (!cpu_has_counter || !mips_hpt_frequency) return -ENXIO; -#ifdef CONFIG_MIPS_MT_SMTC - setup_smtc_dummy_clockevent_device(); - - /* - * On SMTC we only register VPE0's compare interrupt as clockevent - * device. - */ - if (cpu) - return 0; -#endif - if (!c0_compare_int_usable()) return -ENXIO; @@ -265,13 +195,9 @@ int __cpuinit mips_clockevent_init(void) cd->rating = 300; cd->irq = irq; -#ifdef CONFIG_MIPS_MT_SMTC - cd->cpumask = CPU_MASK_ALL; -#else cd->cpumask = cpumask_of_cpu(cpu); -#endif cd->set_next_event = mips_next_event; - cd->set_mode = mips_set_mode; + cd->set_mode = mips_set_clock_mode; cd->event_handler = mips_event_handler; clockevents_register_device(cd); @@ -281,12 +207,9 @@ int __cpuinit mips_clockevent_init(void) cp0_timer_irq_installed = 1; -#ifdef CONFIG_MIPS_MT_SMTC -#define CPUCTR_IMASKBIT (0x100 << cp0_compare_irq) - setup_irq_smtc(irq, &c0_compare_irqaction, CPUCTR_IMASKBIT); -#else setup_irq(irq, &c0_compare_irqaction); -#endif return 0; } + +#endif /* Not CONFIG_MIPS_MT_SMTC */ diff --git a/arch/mips/kernel/cevt-smtc.c b/arch/mips/kernel/cevt-smtc.c new file mode 100644 index 000000000000..5162fe4b5952 --- /dev/null +++ b/arch/mips/kernel/cevt-smtc.c @@ -0,0 +1,321 @@ +/* + * This file is subject to the terms and conditions of the GNU General Public + * License. See the file "COPYING" in the main directory of this archive + * for more details. + * + * Copyright (C) 2007 MIPS Technologies, Inc. + * Copyright (C) 2007 Ralf Baechle + * Copyright (C) 2008 Kevin D. Kissell, Paralogos sarl + */ +#include +#include +#include + +#include +#include +#include + +/* + * Variant clock event timer support for SMTC on MIPS 34K, 1004K + * or other MIPS MT cores. + * + * Notes on SMTC Support: + * + * SMTC has multiple microthread TCs pretending to be Linux CPUs. + * But there's only one Count/Compare pair per VPE, and Compare + * interrupts are taken opportunisitically by available TCs + * bound to the VPE with the Count register. The new timer + * framework provides for global broadcasts, but we really + * want VPE-level multicasts for best behavior. So instead + * of invoking the high-level clock-event broadcast code, + * this version of SMTC support uses the historical SMTC + * multicast mechanisms "under the hood", appearing to the + * generic clock layer as if the interrupts are per-CPU. + * + * The approach taken here is to maintain a set of NR_CPUS + * virtual timers, and track which "CPU" needs to be alerted + * at each event. + * + * It's unlikely that we'll see a MIPS MT core with more than + * 2 VPEs, but we *know* that we won't need to handle more + * VPEs than we have "CPUs". So NCPUs arrays of NCPUs elements + * is always going to be overkill, but always going to be enough. + */ + +unsigned long smtc_nexttime[NR_CPUS][NR_CPUS]; +static int smtc_nextinvpe[NR_CPUS]; + +/* + * Timestamps stored are absolute values to be programmed + * into Count register. Valid timestamps will never be zero. + * If a Zero Count value is actually calculated, it is converted + * to be a 1, which will introduce 1 or two CPU cycles of error + * roughly once every four billion events, which at 1000 HZ means + * about once every 50 days. If that's actually a problem, one + * could alternate squashing 0 to 1 and to -1. + */ + +#define MAKEVALID(x) (((x) == 0L) ? 1L : (x)) +#define ISVALID(x) ((x) != 0L) + +/* + * Time comparison is subtle, as it's really truncated + * modular arithmetic. + */ + +#define IS_SOONER(a, b, reference) \ + (((a) - (unsigned long)(reference)) < ((b) - (unsigned long)(reference))) + +/* + * CATCHUP_INCREMENT, used when the function falls behind the counter. + * Could be an increasing function instead of a constant; + */ + +#define CATCHUP_INCREMENT 64 + +static int mips_next_event(unsigned long delta, + struct clock_event_device *evt) +{ + unsigned long flags; + unsigned int mtflags; + unsigned long timestamp, reference, previous; + unsigned long nextcomp = 0L; + int vpe = current_cpu_data.vpe_id; + int cpu = smp_processor_id(); + local_irq_save(flags); + mtflags = dmt(); + + /* + * Maintain the per-TC virtual timer + * and program the per-VPE shared Count register + * as appropriate here... + */ + reference = (unsigned long)read_c0_count(); + timestamp = MAKEVALID(reference + delta); + /* + * To really model the clock, we have to catch the case + * where the current next-in-VPE timestamp is the old + * timestamp for the calling CPE, but the new value is + * in fact later. In that case, we have to do a full + * scan and discover the new next-in-VPE CPU id and + * timestamp. + */ + previous = smtc_nexttime[vpe][cpu]; + if (cpu == smtc_nextinvpe[vpe] && ISVALID(previous) + && IS_SOONER(previous, timestamp, reference)) { + int i; + int soonest = cpu; + + /* + * Update timestamp array here, so that new + * value gets considered along with those of + * other virtual CPUs on the VPE. + */ + smtc_nexttime[vpe][cpu] = timestamp; + for_each_online_cpu(i) { + if (ISVALID(smtc_nexttime[vpe][i]) + && IS_SOONER(smtc_nexttime[vpe][i], + smtc_nexttime[vpe][soonest], reference)) { + soonest = i; + } + } + smtc_nextinvpe[vpe] = soonest; + nextcomp = smtc_nexttime[vpe][soonest]; + /* + * Otherwise, we don't have to process the whole array rank, + * we just have to see if the event horizon has gotten closer. + */ + } else { + if (!ISVALID(smtc_nexttime[vpe][smtc_nextinvpe[vpe]]) || + IS_SOONER(timestamp, + smtc_nexttime[vpe][smtc_nextinvpe[vpe]], reference)) { + smtc_nextinvpe[vpe] = cpu; + nextcomp = timestamp; + } + /* + * Since next-in-VPE may me the same as the executing + * virtual CPU, we update the array *after* checking + * its value. + */ + smtc_nexttime[vpe][cpu] = timestamp; + } + + /* + * It may be that, in fact, we don't need to update Compare, + * but if we do, we want to make sure we didn't fall into + * a crack just behind Count. + */ + if (ISVALID(nextcomp)) { + write_c0_compare(nextcomp); + ehb(); + /* + * We never return an error, we just make sure + * that we trigger the handlers as quickly as + * we can if we fell behind. + */ + while ((nextcomp - (unsigned long)read_c0_count()) + > (unsigned long)LONG_MAX) { + nextcomp += CATCHUP_INCREMENT; + write_c0_compare(nextcomp); + ehb(); + } + } + emt(mtflags); + local_irq_restore(flags); + return 0; +} + + +void smtc_distribute_timer(int vpe) +{ + unsigned long flags; + unsigned int mtflags; + int cpu; + struct clock_event_device *cd; + unsigned long nextstamp = 0L; + unsigned long reference; + + +repeat: + for_each_online_cpu(cpu) { + /* + * Find virtual CPUs within the current VPE who have + * unserviced timer requests whose time is now past. + */ + local_irq_save(flags); + mtflags = dmt(); + if (cpu_data[cpu].vpe_id == vpe && + ISVALID(smtc_nexttime[vpe][cpu])) { + reference = (unsigned long)read_c0_count(); + if ((smtc_nexttime[vpe][cpu] - reference) + > (unsigned long)LONG_MAX) { + smtc_nexttime[vpe][cpu] = 0L; + emt(mtflags); + local_irq_restore(flags); + /* + * We don't send IPIs to ourself. + */ + if (cpu != smp_processor_id()) { + smtc_send_ipi(cpu, SMTC_CLOCK_TICK, 0); + } else { + cd = &per_cpu(mips_clockevent_device, cpu); + cd->event_handler(cd); + } + } else { + /* Local to VPE but Valid Time not yet reached. */ + if (!ISVALID(nextstamp) || + IS_SOONER(smtc_nexttime[vpe][cpu], nextstamp, + reference)) { + smtc_nextinvpe[vpe] = cpu; + nextstamp = smtc_nexttime[vpe][cpu]; + } + emt(mtflags); + local_irq_restore(flags); + } + } else { + emt(mtflags); + local_irq_restore(flags); + + } + } + /* Reprogram for interrupt at next soonest timestamp for VPE */ + if (ISVALID(nextstamp)) { + write_c0_compare(nextstamp); + ehb(); + if ((nextstamp - (unsigned long)read_c0_count()) + > (unsigned long)LONG_MAX) + goto repeat; + } +} + + +irqreturn_t c0_compare_interrupt(int irq, void *dev_id) +{ + int cpu = smp_processor_id(); + + /* If we're running SMTC, we've got MIPS MT and therefore MIPS32R2 */ + handle_perf_irq(1); + + if (read_c0_cause() & (1 << 30)) { + /* Clear Count/Compare Interrupt */ + write_c0_compare(read_c0_compare()); + smtc_distribute_timer(cpu_data[cpu].vpe_id); + } + return IRQ_HANDLED; +} + + +int __cpuinit mips_clockevent_init(void) +{ + uint64_t mips_freq = mips_hpt_frequency; + unsigned int cpu = smp_processor_id(); + struct clock_event_device *cd; + unsigned int irq; + int i; + int j; + + if (!cpu_has_counter || !mips_hpt_frequency) + return -ENXIO; + if (cpu == 0) { + for (i = 0; i < num_possible_cpus(); i++) { + smtc_nextinvpe[i] = 0; + for (j = 0; j < num_possible_cpus(); j++) + smtc_nexttime[i][j] = 0L; + } + /* + * SMTC also can't have the usablility test + * run by secondary TCs once Compare is in use. + */ + if (!c0_compare_int_usable()) + return -ENXIO; + } + + /* + * With vectored interrupts things are getting platform specific. + * get_c0_compare_int is a hook to allow a platform to return the + * interrupt number of it's liking. + */ + irq = MIPS_CPU_IRQ_BASE + cp0_compare_irq; + if (get_c0_compare_int) + irq = get_c0_compare_int(); + + cd = &per_cpu(mips_clockevent_device, cpu); + + cd->name = "MIPS"; + cd->features = CLOCK_EVT_FEAT_ONESHOT; + + /* Calculate the min / max delta */ + cd->mult = div_sc((unsigned long) mips_freq, NSEC_PER_SEC, 32); + cd->shift = 32; + cd->max_delta_ns = clockevent_delta2ns(0x7fffffff, cd); + cd->min_delta_ns = clockevent_delta2ns(0x300, cd); + + cd->rating = 300; + cd->irq = irq; + cd->cpumask = cpumask_of_cpu(cpu); + cd->set_next_event = mips_next_event; + cd->set_mode = mips_set_clock_mode; + cd->event_handler = mips_event_handler; + + clockevents_register_device(cd); + + /* + * On SMTC we only want to do the data structure + * initialization and IRQ setup once. + */ + if (cpu) + return 0; + /* + * And we need the hwmask associated with the c0_compare + * vector to be initialized. + */ + irq_hwmask[irq] = (0x100 << cp0_compare_irq); + if (cp0_timer_irq_installed) + return 0; + + cp0_timer_irq_installed = 1; + + setup_irq(irq, &c0_compare_irqaction); + + return 0; +} diff --git a/arch/mips/kernel/cpu-probe.c b/arch/mips/kernel/cpu-probe.c index 11c92dc53791..e621fda8ab37 100644 --- a/arch/mips/kernel/cpu-probe.c +++ b/arch/mips/kernel/cpu-probe.c @@ -54,14 +54,18 @@ extern void r4k_wait(void); * interrupt is requested" restriction in the MIPS32/MIPS64 architecture makes * using this version a gamble. */ -static void r4k_wait_irqoff(void) +void r4k_wait_irqoff(void) { local_irq_disable(); if (!need_resched()) - __asm__(" .set mips3 \n" + __asm__(" .set push \n" + " .set mips3 \n" " wait \n" - " .set mips0 \n"); + " .set pop \n"); local_irq_enable(); + __asm__(" .globl __pastwait \n" + "__pastwait: \n"); + return; } /* diff --git a/arch/mips/kernel/genex.S b/arch/mips/kernel/genex.S index f886dd7f708e..01dcbe38fa01 100644 --- a/arch/mips/kernel/genex.S +++ b/arch/mips/kernel/genex.S @@ -282,8 +282,8 @@ NESTED(except_vec_vi_handler, 0, sp) and t0, a0, t1 #ifdef CONFIG_MIPS_MT_SMTC_IM_BACKSTOP mfc0 t2, CP0_TCCONTEXT - or t0, t0, t2 - mtc0 t0, CP0_TCCONTEXT + or t2, t0, t2 + mtc0 t2, CP0_TCCONTEXT #endif /* CONFIG_MIPS_MT_SMTC_IM_BACKSTOP */ xor t1, t1, t0 mtc0 t1, CP0_STATUS diff --git a/arch/mips/kernel/smtc.c b/arch/mips/kernel/smtc.c index 39b491b9ad87..897fb2b4751c 100644 --- a/arch/mips/kernel/smtc.c +++ b/arch/mips/kernel/smtc.c @@ -1,4 +1,21 @@ -/* Copyright (C) 2004 Mips Technologies, Inc */ +/* + * This program is free software; you can redistribute it and/or + * modify it under the terms of the GNU General Public License + * as published by the Free Software Foundation; either version 2 + * of the License, or (at your option) any later version. + * + * This program is distributed in the hope that it will be useful, + * but WITHOUT ANY WARRANTY; without even the implied warranty of + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the + * GNU General Public License for more details. + * + * You should have received a copy of the GNU General Public License + * along with this program; if not, write to the Free Software + * Foundation, Inc., 59 Temple Place - Suite 330, Boston, MA 02111-1307, USA. + * + * Copyright (C) 2004 Mips Technologies, Inc + * Copyright (C) 2008 Kevin D. Kissell + */ #include #include @@ -21,7 +38,6 @@ #include #include #include -#include #include /* @@ -58,11 +74,6 @@ unsigned long irq_hwmask[NR_IRQS]; asiduse smtc_live_asid[MAX_SMTC_TLBS][MAX_SMTC_ASIDS]; -/* - * Clock interrupt "latch" buffers, per "CPU" - */ - -static atomic_t ipi_timer_latch[NR_CPUS]; /* * Number of InterProcessor Interrupt (IPI) message buffers to allocate @@ -282,7 +293,7 @@ static void smtc_configure_tlb(void) * phys_cpu_present_map and the logical/physical mappings. */ -int __init mipsmt_build_cpu_map(int start_cpu_slot) +int __init smtc_build_cpu_map(int start_cpu_slot) { int i, ntcs; @@ -325,7 +336,12 @@ static void smtc_tc_setup(int vpe, int tc, int cpu) write_tc_c0_tcstatus((read_tc_c0_tcstatus() & ~(TCSTATUS_TKSU | TCSTATUS_DA | TCSTATUS_IXMT)) | TCSTATUS_A); - write_tc_c0_tccontext(0); + /* + * TCContext gets an offset from the base of the IPIQ array + * to be used in low-level code to detect the presence of + * an active IPI queue + */ + write_tc_c0_tccontext((sizeof(struct smtc_ipi_q) * cpu) << 16); /* Bind tc to vpe */ write_tc_c0_tcbind(vpe); /* In general, all TCs should have the same cpu_data indications */ @@ -336,10 +352,18 @@ static void smtc_tc_setup(int vpe, int tc, int cpu) cpu_data[cpu].options &= ~MIPS_CPU_FPU; cpu_data[cpu].vpe_id = vpe; cpu_data[cpu].tc_id = tc; + /* Multi-core SMTC hasn't been tested, but be prepared */ + cpu_data[cpu].core = (read_vpe_c0_ebase() >> 1) & 0xff; } +/* + * Tweak to get Count registes in as close a sync as possible. + * Value seems good for 34K-class cores. + */ + +#define CP0_SKEW 8 -void mipsmt_prepare_cpus(void) +void smtc_prepare_cpus(int cpus) { int i, vpe, tc, ntc, nvpe, tcpervpe[NR_CPUS], slop, cpu; unsigned long flags; @@ -363,13 +387,13 @@ void mipsmt_prepare_cpus(void) IPIQ[i].head = IPIQ[i].tail = NULL; spin_lock_init(&IPIQ[i].lock); IPIQ[i].depth = 0; - atomic_set(&ipi_timer_latch[i], 0); } /* cpu_data index starts at zero */ cpu = 0; cpu_data[cpu].vpe_id = 0; cpu_data[cpu].tc_id = 0; + cpu_data[cpu].core = (read_c0_ebase() >> 1) & 0xff; cpu++; /* Report on boot-time options */ @@ -484,7 +508,8 @@ void mipsmt_prepare_cpus(void) write_vpe_c0_compare(0); /* Propagate Config7 */ write_vpe_c0_config7(read_c0_config7()); - write_vpe_c0_count(read_c0_count()); + write_vpe_c0_count(read_c0_count() + CP0_SKEW); + ehb(); } /* enable multi-threading within VPE */ write_vpe_c0_vpecontrol(read_vpe_c0_vpecontrol() | VPECONTROL_TE); @@ -585,24 +610,22 @@ void __cpuinit smtc_boot_secondary(int cpu, struct task_struct *idle) void smtc_init_secondary(void) { - /* - * Start timer on secondary VPEs if necessary. - * plat_timer_setup has already have been invoked by init/main - * on "boot" TC. Like per_cpu_trap_init() hack, this assumes that - * SMTC init code assigns TCs consdecutively and in ascending order - * to across available VPEs. - */ - if (((read_c0_tcbind() & TCBIND_CURTC) != 0) && - ((read_c0_tcbind() & TCBIND_CURVPE) - != cpu_data[smp_processor_id() - 1].vpe_id)){ - write_c0_compare(read_c0_count() + mips_hpt_frequency/HZ); - } - local_irq_enable(); } void smtc_smp_finish(void) { + int cpu = smp_processor_id(); + + /* + * Lowest-numbered CPU per VPE starts a clock tick. + * Like per_cpu_trap_init() hack, this assumes that + * SMTC init code assigns TCs consdecutively and + * in ascending order across available VPEs. + */ + if (cpu > 0 && (cpu_data[cpu].vpe_id != cpu_data[cpu - 1].vpe_id)) + write_c0_compare(read_c0_count() + mips_hpt_frequency/HZ); + printk("TC %d going on-line as CPU %d\n", cpu_data[smp_processor_id()].tc_id, smp_processor_id()); } @@ -755,6 +778,8 @@ void smtc_send_ipi(int cpu, int type, unsigned int action) struct smtc_ipi *pipi; unsigned long flags; int mtflags; + unsigned long tcrestart; + extern void r4k_wait_irqoff(void), __pastwait(void); if (cpu == smp_processor_id()) { printk("Cannot Send IPI to self!\n"); @@ -771,8 +796,6 @@ void smtc_send_ipi(int cpu, int type, unsigned int action) pipi->arg = (void *)action; pipi->dest = cpu; if (cpu_data[cpu].vpe_id != cpu_data[smp_processor_id()].vpe_id) { - if (type == SMTC_CLOCK_TICK) - atomic_inc(&ipi_timer_latch[cpu]); /* If not on same VPE, enqueue and send cross-VPE interrupt */ smtc_ipi_nq(&IPIQ[cpu], pipi); LOCK_CORE_PRA(); @@ -800,22 +823,29 @@ void smtc_send_ipi(int cpu, int type, unsigned int action) if ((tcstatus & TCSTATUS_IXMT) != 0) { /* - * Spin-waiting here can deadlock, - * so we queue the message for the target TC. + * If we're in the the irq-off version of the wait + * loop, we need to force exit from the wait and + * do a direct post of the IPI. + */ + if (cpu_wait == r4k_wait_irqoff) { + tcrestart = read_tc_c0_tcrestart(); + if (tcrestart >= (unsigned long)r4k_wait_irqoff + && tcrestart < (unsigned long)__pastwait) { + write_tc_c0_tcrestart(__pastwait); + tcstatus &= ~TCSTATUS_IXMT; + write_tc_c0_tcstatus(tcstatus); + goto postdirect; + } + } + /* + * Otherwise we queue the message for the target TC + * to pick up when he does a local_irq_restore() */ write_tc_c0_tchalt(0); UNLOCK_CORE_PRA(); - /* Try to reduce redundant timer interrupt messages */ - if (type == SMTC_CLOCK_TICK) { - if (atomic_postincrement(&ipi_timer_latch[cpu])!=0){ - smtc_ipi_nq(&freeIPIq, pipi); - return; - } - } smtc_ipi_nq(&IPIQ[cpu], pipi); } else { - if (type == SMTC_CLOCK_TICK) - atomic_inc(&ipi_timer_latch[cpu]); +postdirect: post_direct_ipi(cpu, pipi); write_tc_c0_tchalt(0); UNLOCK_CORE_PRA(); @@ -883,7 +913,7 @@ static void ipi_call_interrupt(void) smp_call_function_interrupt(); } -DECLARE_PER_CPU(struct clock_event_device, smtc_dummy_clockevent_device); +DECLARE_PER_CPU(struct clock_event_device, mips_clockevent_device); void ipi_decode(struct smtc_ipi *pipi) { @@ -891,20 +921,13 @@ void ipi_decode(struct smtc_ipi *pipi) struct clock_event_device *cd; void *arg_copy = pipi->arg; int type_copy = pipi->type; - int ticks; - smtc_ipi_nq(&freeIPIq, pipi); switch (type_copy) { case SMTC_CLOCK_TICK: irq_enter(); kstat_this_cpu.irqs[MIPS_CPU_IRQ_BASE + 1]++; - cd = &per_cpu(smtc_dummy_clockevent_device, cpu); - ticks = atomic_read(&ipi_timer_latch[cpu]); - atomic_sub(ticks, &ipi_timer_latch[cpu]); - while (ticks) { - cd->event_handler(cd); - ticks--; - } + cd = &per_cpu(mips_clockevent_device, cpu); + cd->event_handler(cd); irq_exit(); break; @@ -937,24 +960,48 @@ void ipi_decode(struct smtc_ipi *pipi) } } +/* + * Similar to smtc_ipi_replay(), but invoked from context restore, + * so it reuses the current exception frame rather than set up a + * new one with self_ipi. + */ + void deferred_smtc_ipi(void) { - struct smtc_ipi *pipi; - unsigned long flags; -/* DEBUG */ - int q = smp_processor_id(); + int cpu = smp_processor_id(); /* * Test is not atomic, but much faster than a dequeue, * and the vast majority of invocations will have a null queue. + * If irq_disabled when this was called, then any IPIs queued + * after we test last will be taken on the next irq_enable/restore. + * If interrupts were enabled, then any IPIs added after the + * last test will be taken directly. */ - if (IPIQ[q].head != NULL) { - while((pipi = smtc_ipi_dq(&IPIQ[q])) != NULL) { - /* ipi_decode() should be called with interrupts off */ - local_irq_save(flags); + + while (IPIQ[cpu].head != NULL) { + struct smtc_ipi_q *q = &IPIQ[cpu]; + struct smtc_ipi *pipi; + unsigned long flags; + + /* + * It may be possible we'll come in with interrupts + * already enabled. + */ + local_irq_save(flags); + + spin_lock(&q->lock); + pipi = __smtc_ipi_dq(q); + spin_unlock(&q->lock); + if (pipi != NULL) ipi_decode(pipi); - local_irq_restore(flags); - } + /* + * The use of the __raw_local restore isn't + * as obviously necessary here as in smtc_ipi_replay(), + * but it's more efficient, given that we're already + * running down the IPI queue. + */ + __raw_local_irq_restore(flags); } } @@ -1066,55 +1113,53 @@ static void setup_cross_vpe_interrupts(unsigned int nvpe) /* * SMTC-specific hacks invoked from elsewhere in the kernel. - * - * smtc_ipi_replay is called from raw_local_irq_restore which is only ever - * called with interrupts disabled. We do rely on interrupts being disabled - * here because using spin_lock_irqsave()/spin_unlock_irqrestore() would - * result in a recursive call to raw_local_irq_restore(). */ -static void __smtc_ipi_replay(void) + /* + * smtc_ipi_replay is called from raw_local_irq_restore + */ + +void smtc_ipi_replay(void) { unsigned int cpu = smp_processor_id(); /* * To the extent that we've ever turned interrupts off, * we may have accumulated deferred IPIs. This is subtle. - * If we use the smtc_ipi_qdepth() macro, we'll get an - * exact number - but we'll also disable interrupts - * and create a window of failure where a new IPI gets - * queued after we test the depth but before we re-enable - * interrupts. So long as IXMT never gets set, however, * we should be OK: If we pick up something and dispatch * it here, that's great. If we see nothing, but concurrent * with this operation, another TC sends us an IPI, IXMT * is clear, and we'll handle it as a real pseudo-interrupt - * and not a pseudo-pseudo interrupt. + * and not a pseudo-pseudo interrupt. The important thing + * is to do the last check for queued message *after* the + * re-enabling of interrupts. */ - if (IPIQ[cpu].depth > 0) { - while (1) { - struct smtc_ipi_q *q = &IPIQ[cpu]; - struct smtc_ipi *pipi; - extern void self_ipi(struct smtc_ipi *); - - spin_lock(&q->lock); - pipi = __smtc_ipi_dq(q); - spin_unlock(&q->lock); - if (!pipi) - break; + while (IPIQ[cpu].head != NULL) { + struct smtc_ipi_q *q = &IPIQ[cpu]; + struct smtc_ipi *pipi; + unsigned long flags; + /* + * It's just possible we'll come in with interrupts + * already enabled. + */ + local_irq_save(flags); + + spin_lock(&q->lock); + pipi = __smtc_ipi_dq(q); + spin_unlock(&q->lock); + /* + ** But use a raw restore here to avoid recursion. + */ + __raw_local_irq_restore(flags); + + if (pipi) { self_ipi(pipi); smtc_cpu_stats[cpu].selfipis++; } } } -void smtc_ipi_replay(void) -{ - raw_local_irq_disable(); - __smtc_ipi_replay(); -} - EXPORT_SYMBOL(smtc_ipi_replay); void smtc_idle_loop_hook(void) @@ -1193,40 +1238,13 @@ void smtc_idle_loop_hook(void) } } - /* - * Now that we limit outstanding timer IPIs, check for hung TC - */ - for (tc = 0; tc < NR_CPUS; tc++) { - /* Don't check ourself - we'll dequeue IPIs just below */ - if ((tc != smp_processor_id()) && - atomic_read(&ipi_timer_latch[tc]) > timerq_limit) { - if (clock_hang_reported[tc] == 0) { - pdb_msg += sprintf(pdb_msg, - "TC %d looks hung with timer latch at %d\n", - tc, atomic_read(&ipi_timer_latch[tc])); - clock_hang_reported[tc]++; - } - } - } emt(mtflags); local_irq_restore(flags); if (pdb_msg != &id_ho_db_msg[0]) printk("CPU%d: %s", smp_processor_id(), id_ho_db_msg); #endif /* CONFIG_SMTC_IDLE_HOOK_DEBUG */ - /* - * Replay any accumulated deferred IPIs. If "Instant Replay" - * is in use, there should never be any. - */ -#ifndef CONFIG_MIPS_MT_SMTC_INSTANT_REPLAY - { - unsigned long flags; - - local_irq_save(flags); - __smtc_ipi_replay(); - local_irq_restore(flags); - } -#endif /* CONFIG_MIPS_MT_SMTC_INSTANT_REPLAY */ + smtc_ipi_replay(); } void smtc_soft_dump(void) @@ -1242,10 +1260,6 @@ void smtc_soft_dump(void) printk("%d: %ld\n", i, smtc_cpu_stats[i].selfipis); } smtc_ipi_qdump(); - printk("Timer IPI Backlogs:\n"); - for (i=0; i < NR_CPUS; i++) { - printk("%d: %d\n", i, atomic_read(&ipi_timer_latch[i])); - } printk("%d Recoveries of \"stolen\" FPU\n", atomic_read(&smtc_fpu_recoveries)); } diff --git a/arch/mips/mti-malta/malta-smtc.c b/arch/mips/mti-malta/malta-smtc.c index 5ea705e49454..f84a46a8ae6e 100644 --- a/arch/mips/mti-malta/malta-smtc.c +++ b/arch/mips/mti-malta/malta-smtc.c @@ -84,12 +84,17 @@ static void msmtc_cpus_done(void) static void __init msmtc_smp_setup(void) { - mipsmt_build_cpu_map(0); + /* + * we won't get the definitive value until + * we've run smtc_prepare_cpus later, but + * we would appear to need an upper bound now. + */ + smp_num_siblings = smtc_build_cpu_map(0); } static void __init msmtc_prepare_cpus(unsigned int max_cpus) { - mipsmt_prepare_cpus(); + smtc_prepare_cpus(max_cpus); } struct plat_smp_ops msmtc_smp_ops = { diff --git a/include/asm-mips/cevt-r4k.h b/include/asm-mips/cevt-r4k.h new file mode 100644 index 000000000000..fa4328f9124f --- /dev/null +++ b/include/asm-mips/cevt-r4k.h @@ -0,0 +1,46 @@ +/* + * This file is subject to the terms and conditions of the GNU General Public + * License. See the file "COPYING" in the main directory of this archive + * for more details. + * + * Copyright (C) 2008 Kevin D. Kissell + */ + +/* + * Definitions used for common event timer implementation + * for MIPS 4K-type processors and their MIPS MT variants. + * Avoids unsightly extern declarations in C files. + */ +#ifndef __ASM_CEVT_R4K_H +#define __ASM_CEVT_R4K_H + +DECLARE_PER_CPU(struct clock_event_device, mips_clockevent_device); + +void mips_event_handler(struct clock_event_device *dev); +int c0_compare_int_usable(void); +void mips_set_clock_mode(enum clock_event_mode, struct clock_event_device *); +irqreturn_t c0_compare_interrupt(int, void *); + +extern struct irqaction c0_compare_irqaction; +extern int cp0_timer_irq_installed; + +/* + * Possibly handle a performance counter interrupt. + * Return true if the timer interrupt should not be checked + */ + +static inline int handle_perf_irq(int r2) +{ + /* + * The performance counter overflow interrupt may be shared with the + * timer interrupt (cp0_perfcount_irq < 0). If it is and a + * performance counter has overflowed (perf_irq() == IRQ_HANDLED) + * and we can't reliably determine if a counter interrupt has also + * happened (!r2) then don't check for a timer interrupt. + */ + return (cp0_perfcount_irq < 0) && + perf_irq() == IRQ_HANDLED && + !r2; +} + +#endif /* __ASM_CEVT_R4K_H */ diff --git a/include/asm-mips/irqflags.h b/include/asm-mips/irqflags.h index 881e8866501d..701ec0ba8fa9 100644 --- a/include/asm-mips/irqflags.h +++ b/include/asm-mips/irqflags.h @@ -38,8 +38,17 @@ __asm__( " .set pop \n" " .endm"); +extern void smtc_ipi_replay(void); + static inline void raw_local_irq_enable(void) { +#ifdef CONFIG_MIPS_MT_SMTC + /* + * SMTC kernel needs to do a software replay of queued + * IPIs, at the cost of call overhead on each local_irq_enable() + */ + smtc_ipi_replay(); +#endif __asm__ __volatile__( "raw_local_irq_enable" : /* no outputs */ @@ -47,6 +56,7 @@ static inline void raw_local_irq_enable(void) : "memory"); } + /* * For cli() we have to insert nops to make sure that the new value * has actually arrived in the status register before the end of this @@ -185,15 +195,14 @@ __asm__( " .set pop \n" " .endm \n"); -extern void smtc_ipi_replay(void); static inline void raw_local_irq_restore(unsigned long flags) { unsigned long __tmp1; -#ifdef CONFIG_MIPS_MT_SMTC_INSTANT_REPLAY +#ifdef CONFIG_MIPS_MT_SMTC /* - * CONFIG_MIPS_MT_SMTC_INSTANT_REPLAY does prompt replay of deferred + * SMTC kernel needs to do a software replay of queued * IPIs, at the cost of branch and call overhead on each * local_irq_restore() */ @@ -208,6 +217,17 @@ static inline void raw_local_irq_restore(unsigned long flags) : "memory"); } +static inline void __raw_local_irq_restore(unsigned long flags) +{ + unsigned long __tmp1; + + __asm__ __volatile__( + "raw_local_irq_restore\t%0" + : "=r" (__tmp1) + : "0" (flags) + : "memory"); +} + static inline int raw_irqs_disabled_flags(unsigned long flags) { #ifdef CONFIG_MIPS_MT_SMTC diff --git a/include/asm-mips/smtc.h b/include/asm-mips/smtc.h index 3639b28f80db..ea60bf08dcb0 100644 --- a/include/asm-mips/smtc.h +++ b/include/asm-mips/smtc.h @@ -6,6 +6,7 @@ */ #include +#include /* * System-wide SMTC status information @@ -38,14 +39,15 @@ struct mm_struct; struct task_struct; void smtc_get_new_mmu_context(struct mm_struct *mm, unsigned long cpu); - +void self_ipi(struct smtc_ipi *); void smtc_flush_tlb_asid(unsigned long asid); -extern int mipsmt_build_cpu_map(int startslot); -extern void mipsmt_prepare_cpus(void); +extern int smtc_build_cpu_map(int startslot); +extern void smtc_prepare_cpus(int cpus); extern void smtc_smp_finish(void); extern void smtc_boot_secondary(int cpu, struct task_struct *t); extern void smtc_cpus_done(void); + /* * Sharing the TLB between multiple VPEs means that the * "random" index selection function is not allowed to -- cgit v1.2.3 From d3d238c7744d08c36a114a59cb537d4c0c6c9a86 Mon Sep 17 00:00:00 2001 From: Heiko Carstens Date: Fri, 3 Oct 2008 21:54:59 +0200 Subject: [S390] nohz: Fix __udelay. This fixes a regression that came with 934b2857cc576ae53c92a66e63fce7ddcfa74691 ("[S390] nohz/sclp: disable timer on synchronous waits."). If udelay() gets called from a disabled context it sets the clock comparator to a value where it expects the next interrupt. When the interrupt happens the clock comparator gets not reset and therefore the interrupt condition doesn't get cleared. The result is an endless timer interrupt loop. In addition this patch fixes also the following: rcutorture reveals that our __udelay implementation is still buggy, since it might schedule tasklets, but prevents their execution: NOHZ: local_softirq_pending 42 NOHZ: local_softirq_pending 02 NOHZ: local_softirq_pending 142 NOHZ: local_softirq_pending 02 To fix this we make sure that only the clock comparator interrupt is enabled when the enabled wait psw is loaded. Also no code gets called anymore which might schedule tasklets. Signed-off-by: Heiko Carstens Signed-off-by: Martin Schwidefsky --- arch/s390/kernel/time.c | 2 ++ arch/s390/lib/delay.c | 88 ++++++++++++++++++++++++++++++------------------- 2 files changed, 56 insertions(+), 34 deletions(-) (limited to 'arch') diff --git a/arch/s390/kernel/time.c b/arch/s390/kernel/time.c index ca114fe46ffb..06acb1a18bbc 100644 --- a/arch/s390/kernel/time.c +++ b/arch/s390/kernel/time.c @@ -169,6 +169,8 @@ void init_cpu_timer(void) static void clock_comparator_interrupt(__u16 code) { + if (S390_lowcore.clock_comparator == -1ULL) + set_clock_comparator(S390_lowcore.clock_comparator); } static void etr_timing_alert(struct etr_irq_parm *); diff --git a/arch/s390/lib/delay.c b/arch/s390/lib/delay.c index fc6ab6094df8..0953cee05efc 100644 --- a/arch/s390/lib/delay.c +++ b/arch/s390/lib/delay.c @@ -1,14 +1,9 @@ /* - * arch/s390/lib/delay.c * Precise Delay Loops for S390 * - * S390 version - * Copyright (C) 1999 IBM Deutschland Entwicklung GmbH, IBM Corporation - * Author(s): Martin Schwidefsky (schwidefsky@de.ibm.com), - * - * Derived from "arch/i386/lib/delay.c" - * Copyright (C) 1993 Linus Torvalds - * Copyright (C) 1997 Martin Mares + * Copyright IBM Corp. 1999,2008 + * Author(s): Martin Schwidefsky , + * Heiko Carstens , */ #include @@ -29,30 +24,31 @@ void __delay(unsigned long loops) asm volatile("0: brct %0,0b" : : "d" ((loops/2) + 1)); } -/* - * Waits for 'usecs' microseconds using the TOD clock comparator. - */ -void __udelay(unsigned long usecs) +static void __udelay_disabled(unsigned long usecs) { - u64 end, time, old_cc = 0; - unsigned long flags, cr0, mask, dummy; - int irq_context; + unsigned long mask, cr0, cr0_saved; + u64 clock_saved; - irq_context = in_interrupt(); - if (!irq_context) - local_bh_disable(); - local_irq_save(flags); - if (raw_irqs_disabled_flags(flags)) { - old_cc = local_tick_disable(); - S390_lowcore.clock_comparator = -1ULL; - __ctl_store(cr0, 0, 0); - dummy = (cr0 & 0xffff00e0) | 0x00000800; - __ctl_load(dummy , 0, 0); - mask = psw_kernel_bits | PSW_MASK_WAIT | PSW_MASK_EXT; - } else - mask = psw_kernel_bits | PSW_MASK_WAIT | - PSW_MASK_EXT | PSW_MASK_IO; + clock_saved = local_tick_disable(); + set_clock_comparator(get_clock() + ((u64) usecs << 12)); + __ctl_store(cr0_saved, 0, 0); + cr0 = (cr0_saved & 0xffff00e0) | 0x00000800; + __ctl_load(cr0 , 0, 0); + mask = psw_kernel_bits | PSW_MASK_WAIT | PSW_MASK_EXT; + trace_hardirqs_on(); + __load_psw_mask(mask); + local_irq_disable(); + __ctl_load(cr0_saved, 0, 0); + local_tick_enable(clock_saved); + set_clock_comparator(S390_lowcore.clock_comparator); +} +static void __udelay_enabled(unsigned long usecs) +{ + unsigned long mask; + u64 end, time; + + mask = psw_kernel_bits | PSW_MASK_WAIT | PSW_MASK_EXT | PSW_MASK_IO; end = get_clock() + ((u64) usecs << 12); do { time = end < S390_lowcore.clock_comparator ? @@ -62,13 +58,37 @@ void __udelay(unsigned long usecs) __load_psw_mask(mask); local_irq_disable(); } while (get_clock() < end); + set_clock_comparator(S390_lowcore.clock_comparator); +} - if (raw_irqs_disabled_flags(flags)) { - __ctl_load(cr0, 0, 0); - local_tick_enable(old_cc); +/* + * Waits for 'usecs' microseconds using the TOD clock comparator. + */ +void __udelay(unsigned long usecs) +{ + unsigned long flags; + + preempt_disable(); + local_irq_save(flags); + if (in_irq()) { + __udelay_disabled(usecs); + goto out; + } + if (in_softirq()) { + if (raw_irqs_disabled_flags(flags)) + __udelay_disabled(usecs); + else + __udelay_enabled(usecs); + goto out; } - if (!irq_context) + if (raw_irqs_disabled_flags(flags)) { + local_bh_disable(); + __udelay_disabled(usecs); _local_bh_enable(); - set_clock_comparator(S390_lowcore.clock_comparator); + goto out; + } + __udelay_enabled(usecs); +out: local_irq_restore(flags); + preempt_enable(); } -- cgit v1.2.3 From cc65f1ec192dc54de57483194502e9fa00934c39 Mon Sep 17 00:00:00 2001 From: "H. Peter Anvin" Date: Fri, 3 Oct 2008 13:00:56 -0700 Subject: x86 setup: correct segfault in generation of 32-bit reloc kernel Impact: segfault on build of a 32-bit relocatable kernel When converting arch/x86/boot/compressed/relocs.c to support unlimited sections, the computation of sym_strtab in walk_relocs() was done incorrectly. This causes a segfault for some people when building the relocatable 32-bit kernel. Pointed out by Anonymous . Signed-off-by: H. Peter Anvin --- arch/x86/boot/compressed/relocs.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'arch') diff --git a/arch/x86/boot/compressed/relocs.c b/arch/x86/boot/compressed/relocs.c index a1310c52fc0c..857e492c571e 100644 --- a/arch/x86/boot/compressed/relocs.c +++ b/arch/x86/boot/compressed/relocs.c @@ -492,7 +492,7 @@ static void walk_relocs(void (*visit)(Elf32_Rel *rel, Elf32_Sym *sym)) continue; } sh_symtab = sec_symtab->symtab; - sym_strtab = sec->link->strtab; + sym_strtab = sec_symtab->link->strtab; for (j = 0; j < sec->shdr.sh_size/sizeof(Elf32_Rel); j++) { Elf32_Rel *rel; Elf32_Sym *sym; -- cgit v1.2.3 From d99e90164e6cf2eb85fa94d547d6336f8127a107 Mon Sep 17 00:00:00 2001 From: Yinghai Lu Date: Sat, 4 Oct 2008 15:55:12 -0700 Subject: x86: gart iommu have direct mapping when agp is present too move init_memory_mapping() out of init_k8_gatt. for: http://bugzilla.kernel.org/show_bug.cgi?id=11676 2.6.27-rc2 to rc8, apgart fails, iommu=soft works, regression This is needed because we need to map the GART aperture even if the GATT is not initialized. Signed-off-by: Yinghai Lu Signed-off-by: Ingo Molnar --- arch/x86/kernel/pci-gart_64.c | 20 +++++++++++--------- 1 file changed, 11 insertions(+), 9 deletions(-) (limited to 'arch') diff --git a/arch/x86/kernel/pci-gart_64.c b/arch/x86/kernel/pci-gart_64.c index 49285f8fd4d5..be33a5442d82 100644 --- a/arch/x86/kernel/pci-gart_64.c +++ b/arch/x86/kernel/pci-gart_64.c @@ -626,7 +626,6 @@ static __init int init_k8_gatt(struct agp_kern_info *info) struct pci_dev *dev; void *gatt; int i, error; - unsigned long start_pfn, end_pfn; printk(KERN_INFO "PCI-DMA: Disabling AGP.\n"); aper_size = aper_base = info->aper_size = 0; @@ -672,12 +671,6 @@ static __init int init_k8_gatt(struct agp_kern_info *info) printk(KERN_INFO "PCI-DMA: aperture base @ %x size %u KB\n", aper_base, aper_size>>10); - /* need to map that range */ - end_pfn = (aper_base>>PAGE_SHIFT) + (aper_size>>PAGE_SHIFT); - if (end_pfn > max_low_pfn_mapped) { - start_pfn = (aper_base>>PAGE_SHIFT); - init_memory_mapping(start_pfn<>PAGE_SHIFT) + (aper_size>>PAGE_SHIFT); + if (end_pfn > max_low_pfn_mapped) { + start_pfn = (aper_base>>PAGE_SHIFT); + init_memory_mapping(start_pfn<> PAGE_SHIFT; -- cgit v1.2.3 From 2fef357cf391b56b8b6952a412f7324e50427a77 Mon Sep 17 00:00:00 2001 From: Ralf Baechle Date: Sun, 5 Oct 2008 18:23:28 +0200 Subject: IDE: Fix platform device registration in Swarm IDE driver (v2) The Swarm IDE driver uses a release method which is defined in the driver itself thus potentially oopsable. The simple fix would be to just leak the device but this patch goes the full length and moves the entire handling of the platform device in the platform code and retains only the platform driver code in drivers/ide/mips/swarm.c. Signed-off-by: Ralf Baechle Cc: "Maciej W. Rozycki" Acked-by: Sergei Shtylyov [bart: remove no longer needed BLK_DEV_IDE_SWARM from ide/Kconfig] Signed-off-by: Bartlomiej Zolnierkiewicz --- arch/mips/sibyte/swarm/Makefile | 3 +- arch/mips/sibyte/swarm/platform.c | 81 ++++++++++++++++ drivers/ide/Kconfig | 4 - drivers/ide/mips/Makefile | 1 - drivers/ide/mips/swarm.c | 197 -------------------------------------- 5 files changed, 83 insertions(+), 203 deletions(-) create mode 100644 arch/mips/sibyte/swarm/platform.c delete mode 100644 drivers/ide/mips/swarm.c (limited to 'arch') diff --git a/arch/mips/sibyte/swarm/Makefile b/arch/mips/sibyte/swarm/Makefile index f18ba9201bbc..7b45f199d92a 100644 --- a/arch/mips/sibyte/swarm/Makefile +++ b/arch/mips/sibyte/swarm/Makefile @@ -1,3 +1,4 @@ -obj-y := setup.o rtc_xicor1241.o rtc_m41t81.o +obj-y := platform.o setup.o rtc_xicor1241.o \ + rtc_m41t81.o obj-$(CONFIG_I2C_BOARDINFO) += swarm-i2c.o diff --git a/arch/mips/sibyte/swarm/platform.c b/arch/mips/sibyte/swarm/platform.c new file mode 100644 index 000000000000..dd0e5b9b64e8 --- /dev/null +++ b/arch/mips/sibyte/swarm/platform.c @@ -0,0 +1,81 @@ +#include +#include +#include +#include +#include +#include + +#include +#include +#include + +#define DRV_NAME "pata-swarm" + +#define SWARM_IDE_SHIFT 5 +#define SWARM_IDE_BASE 0x1f0 +#define SWARM_IDE_CTRL 0x3f6 + +static struct resource swarm_pata_resource[] = { + { + .name = "Swarm GenBus IDE", + .flags = IORESOURCE_MEM, + }, { + .name = "Swarm GenBus IDE", + .flags = IORESOURCE_MEM, + }, { + .name = "Swarm GenBus IDE", + .flags = IORESOURCE_IRQ, + .start = K_INT_GB_IDE, + .end = K_INT_GB_IDE, + }, +}; + +static struct pata_platform_info pata_platform_data = { + .ioport_shift = SWARM_IDE_SHIFT, +}; + +static struct platform_device swarm_pata_device = { + .name = "pata_platform", + .id = -1, + .resource = swarm_pata_resource, + .num_resources = ARRAY_SIZE(swarm_pata_resource), + .dev = { + .platform_data = &pata_platform_data, + .coherent_dma_mask = ~0, /* grumble */ + }, +}; + +static int __init swarm_pata_init(void) +{ + u8 __iomem *base; + phys_t offset, size; + struct resource *r; + + if (!SIBYTE_HAVE_IDE) + return -ENODEV; + + base = ioremap(A_IO_EXT_BASE, 0x800); + offset = __raw_readq(base + R_IO_EXT_REG(R_IO_EXT_START_ADDR, IDE_CS)); + size = __raw_readq(base + R_IO_EXT_REG(R_IO_EXT_MULT_SIZE, IDE_CS)); + iounmap(base); + + offset = G_IO_START_ADDR(offset) << S_IO_ADDRBASE; + size = (G_IO_MULT_SIZE(size) + 1) << S_IO_REGSIZE; + if (offset < A_PHYS_GENBUS || offset >= A_PHYS_GENBUS_END) { + pr_info(DRV_NAME ": PATA interface at GenBus disabled\n"); + + return -EBUSY; + } + + pr_info(DRV_NAME ": PATA interface at GenBus slot %i\n", IDE_CS); + + r = swarm_pata_resource; + r[0].start = offset + (SWARM_IDE_BASE << SWARM_IDE_SHIFT); + r[0].end = offset + ((SWARM_IDE_BASE + 8) << SWARM_IDE_SHIFT) - 1; + r[1].start = offset + (SWARM_IDE_CTRL << SWARM_IDE_SHIFT); + r[1].end = offset + ((SWARM_IDE_CTRL + 1) << SWARM_IDE_SHIFT) - 1; + + return platform_device_register(&swarm_pata_device); +} + +device_initcall(swarm_pata_init); diff --git a/drivers/ide/Kconfig b/drivers/ide/Kconfig index 8e93a797c93d..052879a6f853 100644 --- a/drivers/ide/Kconfig +++ b/drivers/ide/Kconfig @@ -780,10 +780,6 @@ config BLK_DEV_IDEDMA_PMAC to transfer data to and from memory. Saying Y is safe and improves performance. -config BLK_DEV_IDE_SWARM - tristate "IDE for Sibyte evaluation boards" - depends on SIBYTE_SB1xxx_SOC - config BLK_DEV_IDE_AU1XXX bool "IDE for AMD Alchemy Au1200" depends on SOC_AU1200 diff --git a/drivers/ide/mips/Makefile b/drivers/ide/mips/Makefile index 677c7b2bac92..5873fa0b8769 100644 --- a/drivers/ide/mips/Makefile +++ b/drivers/ide/mips/Makefile @@ -1,4 +1,3 @@ -obj-$(CONFIG_BLK_DEV_IDE_SWARM) += swarm.o obj-$(CONFIG_BLK_DEV_IDE_AU1XXX) += au1xxx-ide.o EXTRA_CFLAGS := -Idrivers/ide diff --git a/drivers/ide/mips/swarm.c b/drivers/ide/mips/swarm.c deleted file mode 100644 index 39c9ee995857..000000000000 --- a/drivers/ide/mips/swarm.c +++ /dev/null @@ -1,197 +0,0 @@ -/* - * Copyright (C) 2001, 2002, 2003 Broadcom Corporation - * Copyright (C) 2004 MontaVista Software Inc. - * Author: Manish Lachwani, mlachwani@mvista.com - * Copyright (C) 2004 MIPS Technologies, Inc. All rights reserved. - * Author: Maciej W. Rozycki - * Copyright (c) 2006, 2008 Maciej W. Rozycki - * - * This program is free software; you can redistribute it and/or - * modify it under the terms of the GNU General Public License - * as published by the Free Software Foundation; either version 2 - * of the License, or (at your option) any later version. - * - * This program is distributed in the hope that it will be useful, - * but WITHOUT ANY WARRANTY; without even the implied warranty of - * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the - * GNU General Public License for more details. - * - * You should have received a copy of the GNU General Public License - * along with this program; if not, write to the Free Software - * Foundation, Inc., 59 Temple Place - Suite 330, Boston, MA 02111-1307, USA. - */ - -/* - * Derived loosely from ide-pmac.c, so: - * Copyright (C) 1998 Paul Mackerras. - * Copyright (C) 1995-1998 Mark Lord - */ - -/* - * Boards with SiByte processors so far have supported IDE devices via - * the Generic Bus, PCI bus, and built-in PCMCIA interface. In all - * cases, byte-swapping must be avoided for these devices (whereas - * other PCI devices, for example, will require swapping). Any - * SiByte-targetted kernel including IDE support will include this - * file. Probing of a Generic Bus for an IDE device is controlled by - * the definition of "SIBYTE_HAVE_IDE", which is provided by - * for Broadcom boards. - */ - -#include -#include -#include -#include -#include - -#include - -#include -#include -#include - -#define DRV_NAME "ide-swarm" - -static char swarm_ide_string[] = DRV_NAME; - -static struct resource swarm_ide_resource = { - .name = "SWARM GenBus IDE", - .flags = IORESOURCE_MEM, -}; - -static struct platform_device *swarm_ide_dev; - -static const struct ide_port_info swarm_port_info = { - .name = DRV_NAME, - .host_flags = IDE_HFLAG_MMIO | IDE_HFLAG_NO_DMA, -}; - -/* - * swarm_ide_probe - if the board header indicates the existence of - * Generic Bus IDE, allocate a HWIF for it. - */ -static int __devinit swarm_ide_probe(struct device *dev) -{ - u8 __iomem *base; - struct ide_host *host; - phys_t offset, size; - int i, rc; - hw_regs_t hw, *hws[] = { &hw, NULL, NULL, NULL }; - - if (!SIBYTE_HAVE_IDE) - return -ENODEV; - - base = ioremap(A_IO_EXT_BASE, 0x800); - offset = __raw_readq(base + R_IO_EXT_REG(R_IO_EXT_START_ADDR, IDE_CS)); - size = __raw_readq(base + R_IO_EXT_REG(R_IO_EXT_MULT_SIZE, IDE_CS)); - iounmap(base); - - offset = G_IO_START_ADDR(offset) << S_IO_ADDRBASE; - size = (G_IO_MULT_SIZE(size) + 1) << S_IO_REGSIZE; - if (offset < A_PHYS_GENBUS || offset >= A_PHYS_GENBUS_END) { - printk(KERN_INFO DRV_NAME - ": IDE interface at GenBus disabled\n"); - return -EBUSY; - } - - printk(KERN_INFO DRV_NAME ": IDE interface at GenBus slot %i\n", - IDE_CS); - - swarm_ide_resource.start = offset; - swarm_ide_resource.end = offset + size - 1; - if (request_resource(&iomem_resource, &swarm_ide_resource)) { - printk(KERN_ERR DRV_NAME - ": can't request I/O memory resource\n"); - return -EBUSY; - } - - base = ioremap(offset, size); - - memset(&hw, 0, sizeof(hw)); - for (i = 0; i <= 7; i++) - hw.io_ports_array[i] = - (unsigned long)(base + ((0x1f0 + i) << 5)); - hw.io_ports.ctl_addr = - (unsigned long)(base + (0x3f6 << 5)); - hw.irq = K_INT_GB_IDE; - hw.chipset = ide_generic; - - rc = ide_host_add(&swarm_port_info, hws, &host); - if (rc) - goto err; - - dev_set_drvdata(dev, host); - - return 0; -err: - release_resource(&swarm_ide_resource); - iounmap(base); - return rc; -} - -static struct device_driver swarm_ide_driver = { - .name = swarm_ide_string, - .bus = &platform_bus_type, - .probe = swarm_ide_probe, -}; - -static void swarm_ide_platform_release(struct device *device) -{ - struct platform_device *pldev; - - /* free device */ - pldev = to_platform_device(device); - kfree(pldev); -} - -static int __devinit swarm_ide_init_module(void) -{ - struct platform_device *pldev; - int err; - - printk(KERN_INFO "SWARM IDE driver\n"); - - if (driver_register(&swarm_ide_driver)) { - printk(KERN_ERR "Driver registration failed\n"); - err = -ENODEV; - goto out; - } - - if (!(pldev = kzalloc(sizeof (*pldev), GFP_KERNEL))) { - err = -ENOMEM; - goto out_unregister_driver; - } - - pldev->name = swarm_ide_string; - pldev->id = 0; - pldev->dev.release = swarm_ide_platform_release; - - if (platform_device_register(pldev)) { - err = -ENODEV; - goto out_free_pldev; - } - - if (!pldev->dev.driver) { - /* - * The driver was not bound to this device, there was - * no hardware at this address. Unregister it, as the - * release fuction will take care of freeing the - * allocated structure - */ - platform_device_unregister (pldev); - } - - swarm_ide_dev = pldev; - - return 0; - -out_free_pldev: - kfree(pldev); - -out_unregister_driver: - driver_unregister(&swarm_ide_driver); -out: - return err; -} - -module_init(swarm_ide_init_module); -- cgit v1.2.3 From 4a16ff4c692d49f0caeabe08e6d3ae28a0ad4d04 Mon Sep 17 00:00:00 2001 From: Ralf Baechle Date: Sat, 4 Oct 2008 00:06:29 +0100 Subject: [MIPS] Fix CMP Kconfig configuration and mark as broken. Because sync-r4k.c doesn't build. Signed-off-by: Ralf Baechle --- arch/mips/Kconfig | 27 ++++++++++++++++----------- 1 file changed, 16 insertions(+), 11 deletions(-) (limited to 'arch') diff --git a/arch/mips/Kconfig b/arch/mips/Kconfig index c930b8ceb418..1e06d233fa83 100644 --- a/arch/mips/Kconfig +++ b/arch/mips/Kconfig @@ -211,6 +211,7 @@ config MIPS_MALTA select SYS_SUPPORTS_64BIT_KERNEL select SYS_SUPPORTS_BIG_ENDIAN select SYS_SUPPORTS_LITTLE_ENDIAN + select SYS_SUPPORTS_MIPS_CMP if BROKEN # because SYNC_R4K is broken select SYS_SUPPORTS_MULTITHREADING select SYS_SUPPORTS_SMARTMIPS help @@ -1499,6 +1500,18 @@ config MIPS_APSP_KSPD "exit" syscall notifying other kernel modules the SP program is exiting. You probably want to say yes here. +config MIPS_CMP + bool "MIPS CMP framework support" + depends on SYS_SUPPORTS_MIPS_CMP + select SYNC_R4K if BROKEN + select SYS_SUPPORTS_SMP + select SYS_SUPPORTS_SCHED_SMT if SMP + select WEAK_ORDERING + default n + help + This is a placeholder option for the GCMP work. It will need to + be handled differently... + config SB1_PASS_1_WORKAROUNDS bool depends on CPU_SB1_PASS_1 @@ -1675,6 +1688,9 @@ config SMP config SMP_UP bool +config SYS_SUPPORTS_MIPS_CMP + bool + config SYS_SUPPORTS_SMP bool @@ -1722,17 +1738,6 @@ config NR_CPUS performance should round up your number of processors to the next power of two. -config MIPS_CMP - bool "MIPS CMP framework support" - depends on SMP - select SYNC_R4K - select SYS_SUPPORTS_SCHED_SMT - select WEAK_ORDERING - default n - help - This is a placeholder option for the GCMP work. It will need to - be handled differently... - source "kernel/time/Kconfig" # -- cgit v1.2.3 From fd3d2764ee5aad862e51c21b8239561acdea8c2f Mon Sep 17 00:00:00 2001 From: Ralf Baechle Date: Fri, 3 Oct 2008 22:43:38 +0100 Subject: [MIPS] IP27: Fix build errors if CONFIG_MAPPED_KERNEL=y Signed-off-by: Ralf Baechle --- arch/mips/kernel/head.S | 1 + include/asm-mips/sn/mapped_kernel.h | 8 ++++---- 2 files changed, 5 insertions(+), 4 deletions(-) (limited to 'arch') diff --git a/arch/mips/kernel/head.S b/arch/mips/kernel/head.S index 361364501d34..492a0a8d70fb 100644 --- a/arch/mips/kernel/head.S +++ b/arch/mips/kernel/head.S @@ -22,6 +22,7 @@ #include #include #include +#include #include #include diff --git a/include/asm-mips/sn/mapped_kernel.h b/include/asm-mips/sn/mapped_kernel.h index c3dd5d0d525f..721496a0bb92 100644 --- a/include/asm-mips/sn/mapped_kernel.h +++ b/include/asm-mips/sn/mapped_kernel.h @@ -5,6 +5,8 @@ #ifndef __ASM_SN_MAPPED_KERNEL_H #define __ASM_SN_MAPPED_KERNEL_H +#include + /* * Note on how mapped kernels work: the text and data section is * compiled at cksseg segment (LOADADDR = 0xc001c000), and the @@ -29,10 +31,8 @@ #define MAPPED_ADDR_RO_TO_PHYS(x) (x - REP_BASE) #define MAPPED_ADDR_RW_TO_PHYS(x) (x - REP_BASE - 16777216) -#define MAPPED_KERN_RO_PHYSBASE(n) \ - (PLAT_NODE_DATA(n)->kern_vars.kv_ro_baseaddr) -#define MAPPED_KERN_RW_PHYSBASE(n) \ - (PLAT_NODE_DATA(n)->kern_vars.kv_rw_baseaddr) +#define MAPPED_KERN_RO_PHYSBASE(n) (hub_data(n)->kern_vars.kv_ro_baseaddr) +#define MAPPED_KERN_RW_PHYSBASE(n) (hub_data(n)->kern_vars.kv_rw_baseaddr) #define MAPPED_KERN_RO_TO_PHYS(x) \ ((unsigned long)MAPPED_ADDR_RO_TO_PHYS(x) | \ -- cgit v1.2.3 From e84956f92a846246b09b34f2a728329c386d250f Mon Sep 17 00:00:00 2001 From: "Rafael J. Wysocki" Date: Mon, 6 Oct 2008 11:59:29 +0200 Subject: x86 ACPI: Blacklist two HP machines with buggy BIOSes There is a bug in the BIOSes of some HP boxes with AMD Turions which connects IO-APIC pins with ACPI thermal trip points in such a way that if the state of the IO-APIC is not as expected by the (buggy) BIOS, the thermal trip points are set to insanely low values (usually all of them become 16 degrees Celsius). As a result, thermal throttling kicks in and knock the system down to its shoes. Unfortunately some of the recent IO-APIC changes made the bug show up. To prevent this from happening, blacklist machines that are known to be affected (nx6115 and 6715b in this particular case). This fixes http://bugzilla.kernel.org/show_bug.cgi?id=11516 listed as a regression from 2.6.26. On my box it was caused by: commit 691874fa96d6349a8b60f8ea9c2bae52ece79941 Author: Maciej W. Rozycki Date: Tue May 27 21:19:51 2008 +0100 x86: I/O APIC: timer through 8259A second-chance Signed-off-by: Maciej W. Rozycki Signed-off-by: Ingo Molnar and the whole story is described in this (huge) thread: http://marc.info/?l=linux-kernel&m=121358440508410&w=4 Matthew Garrett told us about that happening on the nx6125: http://marc.info/?l=linux-kernel&m=121396307411930&w=4 and then Maciej analysed the breakage on the basis of a DSDT from the nx6325: http://marc.info/?l=linux-kernel&m=121401068718826&w=4 As far as the Dmitry's and Jason's boxes are concerned, I recognized the symptoms and asked them to verify that the blacklisting helped. It appears that the buggy BIOS code has been copy-pasted to the entire range of machines, for no good reason. Signed-off-by: Rafael J. Wysocki Tested-by: Dmitry Torokhov Tested-by: Jason Vas Dias Signed-off-by: Linus Torvalds --- arch/x86/kernel/acpi/boot.c | 16 ++++++++++++++++ 1 file changed, 16 insertions(+) (limited to 'arch') diff --git a/arch/x86/kernel/acpi/boot.c b/arch/x86/kernel/acpi/boot.c index bfd10fd211cd..c102af85df9c 100644 --- a/arch/x86/kernel/acpi/boot.c +++ b/arch/x86/kernel/acpi/boot.c @@ -1603,6 +1603,14 @@ static struct dmi_system_id __initdata acpi_dmi_table[] = { * is not connected at all. Force ignoring BIOS IRQ0 pin2 * override in that cases. */ + { + .callback = dmi_ignore_irq0_timer_override, + .ident = "HP nx6115 laptop", + .matches = { + DMI_MATCH(DMI_SYS_VENDOR, "Hewlett-Packard"), + DMI_MATCH(DMI_PRODUCT_NAME, "HP Compaq nx6115"), + }, + }, { .callback = dmi_ignore_irq0_timer_override, .ident = "HP NX6125 laptop", @@ -1619,6 +1627,14 @@ static struct dmi_system_id __initdata acpi_dmi_table[] = { DMI_MATCH(DMI_PRODUCT_NAME, "HP Compaq nx6325"), }, }, + { + .callback = dmi_ignore_irq0_timer_override, + .ident = "HP 6715b laptop", + .matches = { + DMI_MATCH(DMI_SYS_VENDOR, "Hewlett-Packard"), + DMI_MATCH(DMI_PRODUCT_NAME, "HP Compaq 6715b"), + }, + }, {} }; -- cgit v1.2.3 From e85ceae9102f6e3c1d707e7ac88fa48d252e9cfa Mon Sep 17 00:00:00 2001 From: Jan Kiszka Date: Mon, 6 Oct 2008 13:50:59 -0500 Subject: kgdb, x86: Avoid invoking kgdb_nmicallback twice per NMI Stress-testing KVM's latest NMI support with kgdbts inside an SMP guest, I came across spurious unhandled NMIs while running the singlestep test. Looking closer at the code path each NMI takes when KGDB is enabled, I noticed that kgdb_nmicallback is called twice per event: One time via DIE_NMI_IPI notification, the second time on DIE_NMI. Removing the first invocation cures the unhandled NMIs here. Signed-off-by: Jan Kiszka Signed-off-by: Jason Wessel --- arch/x86/kernel/kgdb.c | 7 +------ 1 file changed, 1 insertion(+), 6 deletions(-) (limited to 'arch') diff --git a/arch/x86/kernel/kgdb.c b/arch/x86/kernel/kgdb.c index 8282a2139681..10435a120d22 100644 --- a/arch/x86/kernel/kgdb.c +++ b/arch/x86/kernel/kgdb.c @@ -455,12 +455,7 @@ static int __kgdb_notify(struct die_args *args, unsigned long cmd) return NOTIFY_DONE; case DIE_NMI_IPI: - if (atomic_read(&kgdb_active) != -1) { - /* KGDB CPU roundup */ - kgdb_nmicallback(raw_smp_processor_id(), regs); - was_in_debug_nmi[raw_smp_processor_id()] = 1; - touch_nmi_watchdog(); - } + /* Just ignore, we will handle the roundup on DIE_NMI. */ return NOTIFY_DONE; case DIE_NMIUNKNOWN: -- cgit v1.2.3 From 8806048878fce541afe1d72cd2320f092519f90a Mon Sep 17 00:00:00 2001 From: Ralf Baechle Date: Wed, 8 Oct 2008 19:14:33 +0100 Subject: [MIPS] Sibyte: Register PIO PATA device only for Swarm and Litte Sur Symbol name spaghetti which is too complicated to cleanup on this stage of the release cycle breaks the build on BCM1480 platforms. Signed-off-by: Ralf Baechle --- arch/mips/sibyte/swarm/platform.c | 4 ++++ 1 file changed, 4 insertions(+) (limited to 'arch') diff --git a/arch/mips/sibyte/swarm/platform.c b/arch/mips/sibyte/swarm/platform.c index dd0e5b9b64e8..54847fe1e564 100644 --- a/arch/mips/sibyte/swarm/platform.c +++ b/arch/mips/sibyte/swarm/platform.c @@ -9,6 +9,8 @@ #include #include +#if defined(CONFIG_SIBYTE_SWARM) || defined(CONFIG_SIBYTE_LITTLESUR) + #define DRV_NAME "pata-swarm" #define SWARM_IDE_SHIFT 5 @@ -79,3 +81,5 @@ static int __init swarm_pata_init(void) } device_initcall(swarm_pata_init); + +#endif /* defined(CONFIG_SIBYTE_SWARM) || defined(CONFIG_SIBYTE_LITTLESUR) */ -- cgit v1.2.3