From 5cd0c3442021fbf39c7152b341a952aa24054be9 Mon Sep 17 00:00:00 2001 From: Russell King Date: Tue, 3 May 2005 12:18:46 +0100 Subject: [PATCH] ARM: decompressor: use platform debug macros Rather than duplicate the assembly for debug macros in the decompressor head.S, use asm/arch/debug-macros.S instead. Signed-off-by: Russell King --- arch/arm/boot/compressed/head.S | 105 +++++++++------------------------------- 1 file changed, 22 insertions(+), 83 deletions(-) (limited to 'arch') diff --git a/arch/arm/boot/compressed/head.S b/arch/arm/boot/compressed/head.S index c0e7aff3dec..7c7f475e213 100644 --- a/arch/arm/boot/compressed/head.S +++ b/arch/arm/boot/compressed/head.S @@ -18,48 +18,30 @@ * Please select one of the following when turning on debugging. */ #ifdef DEBUG -#if defined(CONFIG_DEBUG_DC21285_PORT) - .macro loadsp, rb - mov \rb, #0x42000000 - .endm - .macro writeb, rb - str \rb, [r3, #0x160] - .endm -#elif defined(CONFIG_DEBUG_ICEDCC) + +#include + +#if defined(CONFIG_DEBUG_ICEDCC) .macro loadsp, rb .endm - .macro writeb, rb - mcr p14, 0, \rb, c0, c1, 0 - .endm -#elif defined(CONFIG_FOOTBRIDGE) - .macro loadsp, rb - mov \rb, #0x7c000000 + .macro writeb, ch, rb + mcr p14, 0, \ch, c0, c1, 0 .endm - .macro writeb, rb - strb \rb, [r3, #0x3f8] +#else + .macro writeb, ch, rb + senduart \ch, \rb .endm -#elif defined(CONFIG_ARCH_RPC) + +#if defined(CONFIG_FOOTBRIDGE) || \ + defined(CONFIG_ARCH_RPC) || \ + defined(CONFIG_ARCH_INTEGRATOR) || \ + defined(CONFIG_ARCH_PXA) || \ + defined(CONFIG_ARCH_IXP4XX) || \ + defined(CONFIG_ARCH_IXP2000) || \ + defined(CONFIG_ARCH_LH7A40X) || \ + defined(CONFIG_ARCH_OMAP) .macro loadsp, rb - mov \rb, #0x03000000 - orr \rb, \rb, #0x00010000 - .endm - .macro writeb, rb - strb \rb, [r3, #0x3f8 << 2] - .endm -#elif defined(CONFIG_ARCH_INTEGRATOR) - .macro loadsp, rb - mov \rb, #0x16000000 - .endm - .macro writeb, rb - strb \rb, [r3, #0] - .endm -#elif defined(CONFIG_ARCH_PXA) /* Xscale-type */ - .macro loadsp, rb - mov \rb, #0x40000000 - orr \rb, \rb, #0x00100000 - .endm - .macro writeb, rb - strb \rb, [r3, #0] + addruart \rb .endm #elif defined(CONFIG_ARCH_SA1100) .macro loadsp, rb @@ -70,64 +52,21 @@ add \rb, \rb, #0x00010000 @ Ser1 # endif .endm - .macro writeb, rb - str \rb, [r3, #0x14] @ UTDR - .endm -#elif defined(CONFIG_ARCH_IXP4XX) - .macro loadsp, rb - mov \rb, #0xc8000000 - .endm - .macro writeb, rb - str \rb, [r3, #0] -#elif defined(CONFIG_ARCH_IXP2000) - .macro loadsp, rb - mov \rb, #0xc0000000 - orr \rb, \rb, #0x00030000 - .endm - .macro writeb, rb - str \rb, [r3, #0] - .endm -#elif defined(CONFIG_ARCH_LH7A40X) - .macro loadsp, rb - ldr \rb, =0x80000700 @ UART2 UARTBASE - .endm - .macro writeb, rb - strb \rb, [r3, #0] - .endm -#elif defined(CONFIG_ARCH_OMAP) - .macro loadsp, rb - mov \rb, #0xff000000 @ physical base address - add \rb, \rb, #0x00fb0000 -#if defined(CONFIG_OMAP_LL_DEBUG_UART2) || defined(CONFIG_OMAP_LL_DEBUG_UART3) - add \rb, \rb, #0x00000800 -#endif -#ifdef CONFIG_OMAP_LL_DEBUG_UART3 - add \rb, \rb, #0x00009000 -#endif - .endm - .macro writeb, rb - strb \rb, [r3] - .endm #elif defined(CONFIG_ARCH_IOP331) .macro loadsp, rb mov \rb, #0xff000000 orr \rb, \rb, #0x00ff0000 orr \rb, \rb, #0x0000f700 @ location of the UART .endm - .macro writeb, rb - str \rb, [r3, #0] - .endm #elif defined(CONFIG_ARCH_S3C2410) - .macro loadsp, rb + .macro loadsp, rb mov \rb, #0x50000000 add \rb, \rb, #0x4000 * CONFIG_S3C2410_LOWLEVEL_UART_PORT .endm - .macro writeb, rb - strb \rb, [r3, #0x20] - .endm #else #error no serial architecture defined #endif +#endif #endif .macro kputc,val @@ -734,7 +673,7 @@ puts: loadsp r3 1: ldrb r2, [r0], #1 teq r2, #0 moveq pc, lr -2: writeb r2 +2: writeb r2, r3 mov r1, #0x00020000 3: subs r1, r1, #1 bne 3b -- cgit v1.2.3-18-g5258 From 1f9c381fa3e0b9b9042e310c69df87eaf9b46ea4 Mon Sep 17 00:00:00 2001 From: Russell King Date: Tue, 3 May 2005 12:22:19 +0100 Subject: [PATCH] ARM: Clean up commenting/spacing for Integrator Signed-off-by: Russell King --- arch/arm/mach-integrator/core.c | 6 ++++-- arch/arm/mach-integrator/leds.c | 2 +- 2 files changed, 5 insertions(+), 3 deletions(-) (limited to 'arch') diff --git a/arch/arm/mach-integrator/core.c b/arch/arm/mach-integrator/core.c index 86c50c3889b..bd17b515431 100644 --- a/arch/arm/mach-integrator/core.c +++ b/arch/arm/mach-integrator/core.c @@ -216,7 +216,9 @@ integrator_timer_interrupt(int irq, void *dev_id, struct pt_regs *regs) write_seqlock(&xtime_lock); - // ...clear the interrupt + /* + * clear the interrupt + */ timer1->TimerClear = 1; timer_tick(regs); @@ -264,7 +266,7 @@ void __init integrator_time_init(unsigned long reload, unsigned int ctrl) timer1->TimerValue = timer_reload; timer1->TimerControl = timer_ctrl; - /* + /* * Make irqs happen for the system timer */ setup_irq(IRQ_TIMERINT1, &integrator_timer_irq); diff --git a/arch/arm/mach-integrator/leds.c b/arch/arm/mach-integrator/leds.c index 9d182b77b31..d2c0ab21150 100644 --- a/arch/arm/mach-integrator/leds.c +++ b/arch/arm/mach-integrator/leds.c @@ -37,7 +37,7 @@ static void integrator_leds_event(led_event_t ledevt) unsigned long flags; const unsigned int dbg_base = IO_ADDRESS(INTEGRATOR_DBG_BASE); unsigned int update_alpha_leds; - + // yup, change the LEDs local_irq_save(flags); update_alpha_leds = 0; -- cgit v1.2.3-18-g5258 From 52292c9b8c16aa9024a65aaeeca434bb8fec7d24 Mon Sep 17 00:00:00 2001 From: Benjamin Herrenschmidt Date: Tue, 3 May 2005 15:34:58 +1000 Subject: [PATCH] ppc64: fix gcc 4.0 vs CONFIG_ALTIVEC gcc-4.0 generates altivec code implicitly when -mcpu indicates an altivec capable CPU which is not suitable for the kernel. However, we used to set -mcpu=970 when CONFIG_ALTIVEC was set because a gcc-3.x bug prevented from using -maltivec along with -mcpu=power4, thus prevented building the RAID6 altivec code. This patch fixes all of this by testing for the gcc version. If 4.0 or later, just normally use -mcpu=power4 and let the RAID6 code add -maltivec to the few files it needs to be compiled with altivec support. For 3.x, we still use -mcpu=970 to work around the above problem, which is fine as 3.x will never implicitly generate altivec code. The Makefile hackery may not be the most lovely, I welcome anybody more skilled than me to improve it. Signed-off-by: Benjamin Herrenschmidt Signed-off-by: Linus Torvalds --- arch/ppc64/Makefile | 7 +++++++ 1 file changed, 7 insertions(+) (limited to 'arch') diff --git a/arch/ppc64/Makefile b/arch/ppc64/Makefile index d33e20bcc52..691f3008e69 100644 --- a/arch/ppc64/Makefile +++ b/arch/ppc64/Makefile @@ -56,12 +56,19 @@ LDFLAGS_vmlinux := -Bstatic -e $(KERNELLOAD) -Ttext $(KERNELLOAD) CFLAGS += -msoft-float -pipe -mminimal-toc -mtraceback=none \ -mcall-aixdesc +GCC_VERSION := $(call cc-version) +GCC_BROKEN_VEC := $(shell if [ $(GCC_VERSION) -lt 0400 ] ; then echo "y"; fi ;) + ifeq ($(CONFIG_POWER4_ONLY),y) ifeq ($(CONFIG_ALTIVEC),y) +ifeq ($(GCC_BROKEN_VEC),y) CFLAGS += $(call cc-option,-mcpu=970) else CFLAGS += $(call cc-option,-mcpu=power4) endif +else + CFLAGS += $(call cc-option,-mcpu=power4) +endif else CFLAGS += $(call cc-option,-mtune=power4) endif -- cgit v1.2.3-18-g5258 From 7a9bdd8842d0847f3b15068550a3632a2d259057 Mon Sep 17 00:00:00 2001 From: Suresh Siddha Date: Tue, 5 Apr 2005 18:05:00 -0700 Subject: [IA64] Add config SCHED_SMT Now that we have MC/MT detection patches in, appended patch allows us to configure MT scheduler optimizations. For now, we will this option off by default. There is some discussion going on lkml about setting up sched-domains which are absolutely needed (like for example, we shouldn't setup SMT domain for non MT processors). Once that patch goes in, we can enable this option by default. Signed-off-by: Suresh Siddha Signed-off-by: Tony Luck --- arch/ia64/Kconfig | 9 +++++++++ 1 file changed, 9 insertions(+) (limited to 'arch') diff --git a/arch/ia64/Kconfig b/arch/ia64/Kconfig index 468dbe8a6b9..cad8346def1 100644 --- a/arch/ia64/Kconfig +++ b/arch/ia64/Kconfig @@ -261,6 +261,15 @@ config HOTPLUG_CPU can be controlled through /sys/devices/system/cpu/cpu#. Say N if you want to disable CPU hotplug. +config SCHED_SMT + bool "SMT scheduler support" + depends on SMP + default off + help + Improves the CPU scheduler's decision making when dealing with + Intel IA64 chips with MultiThreading at a cost of slightly increased + overhead in some places. If unsure say N here. + config PREEMPT bool "Preemptible Kernel" help -- cgit v1.2.3-18-g5258 From 512f64295f2f0049515dcc4e97c1f1ae0df9629c Mon Sep 17 00:00:00 2001 From: Andreas Schwab Date: Tue, 26 Apr 2005 23:00:00 -0700 Subject: [IA64] Fix memcpy_mck.S for current binutils The current ia64 assembler complains about mismatching .proc/.endp pairs. (Same patch also sent by H.J. Lu) Signed-off-by: Andreas Schwab Signed-off-by: Tony Luck --- arch/ia64/lib/memcpy_mck.S | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'arch') diff --git a/arch/ia64/lib/memcpy_mck.S b/arch/ia64/lib/memcpy_mck.S index 3c2cd2f04db..6f308e62c13 100644 --- a/arch/ia64/lib/memcpy_mck.S +++ b/arch/ia64/lib/memcpy_mck.S @@ -75,6 +75,7 @@ GLOBAL_ENTRY(memcpy) mov f6=f0 br.cond.sptk .common_code ;; +END(memcpy) GLOBAL_ENTRY(__copy_user) .prologue // check dest alignment @@ -524,7 +525,6 @@ EK(.ex_handler, (p17) st8 [dst1]=r39,8); \ #undef B #undef C #undef D -END(memcpy) /* * Due to lack of local tag support in gcc 2.x assembler, it is not clear which -- cgit v1.2.3-18-g5258 From 9df6f705c0934a49d0f0a3468a5b5044c8aec4f1 Mon Sep 17 00:00:00 2001 From: David Mosberger-Tang Date: Fri, 25 Mar 2005 00:16:00 -0700 Subject: [IA64] fix typos caught by new assembler Patch below fixes 3 trivial typos which are caught by the new assembler (v2.169.90). Please apply. [Note: fix to memcpy that was also part of this patch was separately applied from patches by H.J. and Andreas ... so the delta here only has the other two fixes. -Tony] Signed-off-by: David Mosberger-Tang Signed-off-by: Tony Luck --- arch/ia64/kernel/entry.S | 2 +- arch/ia64/lib/memset.S | 2 +- 2 files changed, 2 insertions(+), 2 deletions(-) (limited to 'arch') diff --git a/arch/ia64/kernel/entry.S b/arch/ia64/kernel/entry.S index d3f093820bc..81c45d44739 100644 --- a/arch/ia64/kernel/entry.S +++ b/arch/ia64/kernel/entry.S @@ -782,7 +782,7 @@ GLOBAL_ENTRY(ia64_ret_from_ia32_execve) st8.spill [r2]=r8 // store return value in slot for r8 and set unat bit .mem.offset 8,0 st8.spill [r3]=r0 // clear error indication in slot for r10 and set unat bit -END(ia64_ret_from_ia32_execve_syscall) +END(ia64_ret_from_ia32_execve) // fall through #endif /* CONFIG_IA32_SUPPORT */ GLOBAL_ENTRY(ia64_leave_kernel) diff --git a/arch/ia64/lib/memset.S b/arch/ia64/lib/memset.S index bd8cf907fe2..f26c16aefb1 100644 --- a/arch/ia64/lib/memset.S +++ b/arch/ia64/lib/memset.S @@ -57,10 +57,10 @@ GLOBAL_ENTRY(memset) { .mmi .prologue alloc tmp = ar.pfs, 3, 0, 0, 0 - .body lfetch.nt1 [dest] // .save ar.lc, save_lc mov.i save_lc = ar.lc + .body } { .mmi mov ret0 = dest // return value cmp.ne p_nz, p_zr = value, r0 // use stf.spill if value is zero -- cgit v1.2.3-18-g5258 From 2074615a13a4f250e0a4e3f6ec8e3733b950a783 Mon Sep 17 00:00:00 2001 From: David Mosberger-Tang Date: Fri, 18 Feb 2005 19:09:00 -0700 Subject: [IA64] use fc.i for fluch_icache_range() This is a small patch to switch fluch_icache_range() to use fc.i instead of fc. This would save time on processors which can establish i-cache coherency without flushing the cache-line out to memory (not that any current processors do). On existing processors, fc.i behaves like fc. The only caveat is that very old assemblers may not know about fc.i yet. Signed-off-by: David Mosberger-Tang Signed-off-by: Tony Luck --- arch/ia64/lib/flush.S | 6 +++--- 1 file changed, 3 insertions(+), 3 deletions(-) (limited to 'arch') diff --git a/arch/ia64/lib/flush.S b/arch/ia64/lib/flush.S index 29c802b1966..a1af9146cfd 100644 --- a/arch/ia64/lib/flush.S +++ b/arch/ia64/lib/flush.S @@ -1,8 +1,8 @@ /* * Cache flushing routines. * - * Copyright (C) 1999-2001 Hewlett-Packard Co - * Copyright (C) 1999-2001 David Mosberger-Tang + * Copyright (C) 1999-2001, 2005 Hewlett-Packard Co + * David Mosberger-Tang */ #include #include @@ -26,7 +26,7 @@ GLOBAL_ENTRY(flush_icache_range) mov ar.lc=r8 ;; -.Loop: fc in0 // issuable on M0 only +.Loop: fc.i in0 // issuable on M2 only add in0=32,in0 br.cloop.sptk.few .Loop ;; -- cgit v1.2.3-18-g5258 From c2d1d65ad441c8abe624bdb1c2cff2e47c8c1ee1 Mon Sep 17 00:00:00 2001 From: Bruce Losure Date: Thu, 24 Mar 2005 12:28:00 -0700 Subject: [IA64-SGI] Altix only: Remove hubdev SAL call Hi Tony, This patch against ia64-test-2.6.12 fixes a bug where the tiocx code was inadvertently un-doing some address modifications done in earlier fixup code. This patch just removes the offending code. Signed-off-by: Bruce Losure Signed-off-by: Tony Luck --- arch/ia64/sn/kernel/tiocx.c | 19 ------------------- 1 file changed, 19 deletions(-) (limited to 'arch') diff --git a/arch/ia64/sn/kernel/tiocx.c b/arch/ia64/sn/kernel/tiocx.c index 66190d7e492..128ccf23973 100644 --- a/arch/ia64/sn/kernel/tiocx.c +++ b/arch/ia64/sn/kernel/tiocx.c @@ -330,19 +330,6 @@ EXPORT_SYMBOL(tiocx_bus_type); EXPORT_SYMBOL(tiocx_dma_addr); EXPORT_SYMBOL(tiocx_swin_base); -static uint64_t tiocx_get_hubdev_info(u64 handle, u64 address) -{ - - struct ia64_sal_retval ret_stuff; - ret_stuff.status = 0; - ret_stuff.v0 = 0; - - ia64_sal_oemcall_nolock(&ret_stuff, - SN_SAL_IOIF_GET_HUBDEV_INFO, - handle, address, 0, 0, 0, 0, 0); - return ret_stuff.v0; -} - static void tio_conveyor_set(nasid_t nasid, int enable_flag) { uint64_t ice_frz; @@ -477,18 +464,12 @@ static int __init tiocx_init(void) if (nasid & 0x1) { /* TIO's are always odd */ struct hubdev_info *hubdev; - uint64_t status; struct xwidget_info *widgetp; DBG("Found TIO at nasid 0x%x\n", nasid); hubdev = (struct hubdev_info *)(NODEPDA(cnodeid)->pdinfo); - status = - tiocx_get_hubdev_info(nasid, - (uint64_t) __pa(hubdev)); - if (status) - continue; widgetp = &hubdev->hdi_xwidget_info[TIOCX_CORELET]; -- cgit v1.2.3-18-g5258 From ce0a3956b32650e229b68964c4400bbdc5ad3ca1 Mon Sep 17 00:00:00 2001 From: Bruce Losure Date: Mon, 25 Apr 2005 19:41:00 -0700 Subject: [IA64-SGI] Altix patch to add bricktype knowledge to tiocx Here is a patch to enable the SGI tiocx bus driver to distingush between FPGA-attached h/w and non-FPGA-attached h/w. Signed-off-by: Bruce Losure Signed-off-by: Tony Luck --- arch/ia64/sn/kernel/tiocx.c | 41 ++++++++++++++++++++++++++++++++--------- 1 file changed, 32 insertions(+), 9 deletions(-) (limited to 'arch') diff --git a/arch/ia64/sn/kernel/tiocx.c b/arch/ia64/sn/kernel/tiocx.c index 128ccf23973..ab9b5f35c2a 100644 --- a/arch/ia64/sn/kernel/tiocx.c +++ b/arch/ia64/sn/kernel/tiocx.c @@ -21,6 +21,8 @@ #include #include #include +#include +#include #include "tio.h" #include "xtalk/xwidgetdev.h" #include "xtalk/hubdev.h" @@ -308,14 +310,12 @@ void tiocx_irq_free(struct sn_irq_info *sn_irq_info) } } -uint64_t -tiocx_dma_addr(uint64_t addr) +uint64_t tiocx_dma_addr(uint64_t addr) { return PHYS_TO_TIODMA(addr); } -uint64_t -tiocx_swin_base(int nasid) +uint64_t tiocx_swin_base(int nasid) { return TIO_SWIN_BASE(nasid, TIOCX_CORELET); } @@ -366,7 +366,29 @@ static void tio_corelet_reset(nasid_t nasid, int corelet) udelay(2000); } -static int fpga_attached(nasid_t nasid) +static int tiocx_btchar_get(int nasid) +{ + moduleid_t module_id; + geoid_t geoid; + int cnodeid; + + cnodeid = nasid_to_cnodeid(nasid); + geoid = cnodeid_get_geoid(cnodeid); + module_id = geo_module(geoid); + return MODULE_GET_BTCHAR(module_id); +} + +static int is_fpga_brick(int nasid) +{ + switch (tiocx_btchar_get(nasid)) { + case L1_BRICKTYPE_SA: + case L1_BRICKTYPE_ATHENA: + return 1; + } + return 0; +} + +static int bitstream_loaded(nasid_t nasid) { uint64_t cx_credits; @@ -383,7 +405,7 @@ static int tiocx_reload(struct cx_dev *cx_dev) int mfg_num = CX_DEV_NONE; nasid_t nasid = cx_dev->cx_id.nasid; - if (fpga_attached(nasid)) { + if (bitstream_loaded(nasid)) { uint64_t cx_id; cx_id = @@ -414,9 +436,10 @@ static ssize_t show_cxdev_control(struct device *dev, char *buf) { struct cx_dev *cx_dev = to_cx_dev(dev); - return sprintf(buf, "0x%x 0x%x 0x%x\n", + return sprintf(buf, "0x%x 0x%x 0x%x %d\n", cx_dev->cx_id.nasid, - cx_dev->cx_id.part_num, cx_dev->cx_id.mfg_num); + cx_dev->cx_id.part_num, cx_dev->cx_id.mfg_num, + tiocx_btchar_get(cx_dev->cx_id.nasid)); } static ssize_t store_cxdev_control(struct device *dev, const char *buf, @@ -462,7 +485,7 @@ static int __init tiocx_init(void) if ((nasid = cnodeid_to_nasid(cnodeid)) < 0) break; /* No more nasids .. bail out of loop */ - if (nasid & 0x1) { /* TIO's are always odd */ + if ((nasid & 0x1) && is_fpga_brick(nasid)) { struct hubdev_info *hubdev; struct xwidget_info *widgetp; -- cgit v1.2.3-18-g5258 From 2e34f07ff0c944399a6456e2d91cf0ca1d9a497c Mon Sep 17 00:00:00 2001 From: Dean Nelson Date: Mon, 21 Mar 2005 19:41:00 -0700 Subject: [PATCH] move cnodeid_to_nasid_table out of pda Another step in the effort to eliminate the SN pda structure. This patch moves the cnodeid_to_nasid_table field out of the pda, making it a standalone per-cpu data item, and exports it so it can be accessed by kernel modules. Signed-off-by: Dean Nelson Signed-off-by: Tony Luck --- arch/ia64/sn/kernel/setup.c | 31 +++++++++++++++++-------------- 1 file changed, 17 insertions(+), 14 deletions(-) (limited to 'arch') diff --git a/arch/ia64/sn/kernel/setup.c b/arch/ia64/sn/kernel/setup.c index d35f2a6f9c9..fea71ee891e 100644 --- a/arch/ia64/sn/kernel/setup.c +++ b/arch/ia64/sn/kernel/setup.c @@ -3,7 +3,7 @@ * License. See the file "COPYING" in the main directory of this archive * for more details. * - * Copyright (C) 1999,2001-2004 Silicon Graphics, Inc. All rights reserved. + * Copyright (C) 1999,2001-2005 Silicon Graphics, Inc. All rights reserved. */ #include @@ -73,6 +73,9 @@ EXPORT_SYMBOL(sn_rtc_cycles_per_second); DEFINE_PER_CPU(struct sn_hub_info_s, __sn_hub_info); EXPORT_PER_CPU_SYMBOL(__sn_hub_info); +DEFINE_PER_CPU(short, __sn_cnodeid_to_nasid[MAX_NUMNODES]); +EXPORT_PER_CPU_SYMBOL(__sn_cnodeid_to_nasid); + partid_t sn_partid = -1; EXPORT_SYMBOL(sn_partid); char sn_system_serial_number_string[128]; @@ -373,11 +376,11 @@ static void __init sn_init_pdas(char **cmdline_p) { cnodeid_t cnode; - memset(pda->cnodeid_to_nasid_table, -1, - sizeof(pda->cnodeid_to_nasid_table)); + memset(sn_cnodeid_to_nasid, -1, + sizeof(__ia64_per_cpu_var(__sn_cnodeid_to_nasid))); for_each_online_node(cnode) - pda->cnodeid_to_nasid_table[cnode] = - pxm_to_nasid(nid_to_pxm_map[cnode]); + sn_cnodeid_to_nasid[cnode] = + pxm_to_nasid(nid_to_pxm_map[cnode]); numionodes = num_online_nodes(); scan_for_ionodes(); @@ -486,15 +489,18 @@ void __init sn_cpu_init(void) pda->idle_flag = 0; if (cpuid != 0) { - memcpy(pda->cnodeid_to_nasid_table, - pdacpu(0)->cnodeid_to_nasid_table, - sizeof(pda->cnodeid_to_nasid_table)); + /* copy cpu 0's sn_cnodeid_to_nasid table to this cpu's */ + memcpy(sn_cnodeid_to_nasid, + (&per_cpu(__sn_cnodeid_to_nasid, 0)), + sizeof(__ia64_per_cpu_var(__sn_cnodeid_to_nasid))); } /* * Check for WARs. * Only needs to be done once, on BSP. - * Has to be done after loop above, because it uses pda.cnodeid_to_nasid_table[i]. + * Has to be done after loop above, because it uses this cpu's + * sn_cnodeid_to_nasid table which was just initialized if this + * isn't cpu 0. * Has to be done before assignment below. */ if (!wars_have_been_checked) { @@ -580,8 +586,7 @@ static void __init scan_for_ionodes(void) brd = find_lboard_any(brd, KLTYPE_SNIA); while (brd) { - pda->cnodeid_to_nasid_table[numionodes] = - brd->brd_nasid; + sn_cnodeid_to_nasid[numionodes] = brd->brd_nasid; physical_node_map[brd->brd_nasid] = numionodes; root_lboard[numionodes] = brd; numionodes++; @@ -602,8 +607,7 @@ static void __init scan_for_ionodes(void) root_lboard[nasid_to_cnodeid(nasid)], KLTYPE_TIO); while (brd) { - pda->cnodeid_to_nasid_table[numionodes] = - brd->brd_nasid; + sn_cnodeid_to_nasid[numionodes] = brd->brd_nasid; physical_node_map[brd->brd_nasid] = numionodes; root_lboard[numionodes] = brd; numionodes++; @@ -614,7 +618,6 @@ static void __init scan_for_ionodes(void) brd = find_lboard_any(brd, KLTYPE_TIO); } } - } int -- cgit v1.2.3-18-g5258 From b0d82bd5df874f7dadbeced1b0163473387da37c Mon Sep 17 00:00:00 2001 From: Dean Nelson Date: Wed, 23 Mar 2005 19:46:00 -0700 Subject: [IA64-SGI] SGI Altix cross partition functionality (2nd This patch contains the shim module (XP) which interfaces between the communication module (XPC) and the functional support modules (like XPNET). Signed-off-by: Dean Nelson Signed-off-by: Tony Luck --- arch/ia64/Kconfig | 10 ++ arch/ia64/sn/kernel/xp_main.c | 289 +++++++++++++++++++++++++++++++++++++++ arch/ia64/sn/kernel/xp_nofault.S | 31 +++++ 3 files changed, 330 insertions(+) create mode 100644 arch/ia64/sn/kernel/xp_main.c create mode 100644 arch/ia64/sn/kernel/xp_nofault.S (limited to 'arch') diff --git a/arch/ia64/Kconfig b/arch/ia64/Kconfig index cad8346def1..ce13ad689d1 100644 --- a/arch/ia64/Kconfig +++ b/arch/ia64/Kconfig @@ -217,6 +217,16 @@ config IA64_SGI_SN_SIM If you are compiling a kernel that will run under SGI's IA-64 simulator (Medusa) then say Y, otherwise say N. +config IA64_SGI_SN_XP + tristate "Support communication between SGI SSIs" + depends on MSPEC + help + An SGI machine can be divided into multiple Single System + Images which act independently of each other and have + hardware based memory protection from the others. Enabling + this feature will allow for direct communication between SSIs + based on a network adapter and DMA messaging. + config FORCE_MAX_ZONEORDER int default "18" diff --git a/arch/ia64/sn/kernel/xp_main.c b/arch/ia64/sn/kernel/xp_main.c new file mode 100644 index 00000000000..3be52a34c80 --- /dev/null +++ b/arch/ia64/sn/kernel/xp_main.c @@ -0,0 +1,289 @@ +/* + * 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) 2004-2005 Silicon Graphics, Inc. All Rights Reserved. + */ + + +/* + * Cross Partition (XP) base. + * + * XP provides a base from which its users can interact + * with XPC, yet not be dependent on XPC. + * + */ + + +#include +#include +#include +#include +#include +#include + + +/* + * Target of nofault PIO read. + */ +u64 xp_nofault_PIOR_target; + + +/* + * xpc_registrations[] keeps track of xpc_connect()'s done by the kernel-level + * users of XPC. + */ +struct xpc_registration xpc_registrations[XPC_NCHANNELS]; + + +/* + * Initialize the XPC interface to indicate that XPC isn't loaded. + */ +static enum xpc_retval xpc_notloaded(void) { return xpcNotLoaded; } + +struct xpc_interface xpc_interface = { + (void (*)(int)) xpc_notloaded, + (void (*)(int)) xpc_notloaded, + (enum xpc_retval (*)(partid_t, int, u32, void **)) xpc_notloaded, + (enum xpc_retval (*)(partid_t, int, void *)) xpc_notloaded, + (enum xpc_retval (*)(partid_t, int, void *, xpc_notify_func, void *)) + xpc_notloaded, + (void (*)(partid_t, int, void *)) xpc_notloaded, + (enum xpc_retval (*)(partid_t, void *)) xpc_notloaded +}; + + +/* + * XPC calls this when it (the XPC module) has been loaded. + */ +void +xpc_set_interface(void (*connect)(int), + void (*disconnect)(int), + enum xpc_retval (*allocate)(partid_t, int, u32, void **), + enum xpc_retval (*send)(partid_t, int, void *), + enum xpc_retval (*send_notify)(partid_t, int, void *, + xpc_notify_func, void *), + void (*received)(partid_t, int, void *), + enum xpc_retval (*partid_to_nasids)(partid_t, void *)) +{ + xpc_interface.connect = connect; + xpc_interface.disconnect = disconnect; + xpc_interface.allocate = allocate; + xpc_interface.send = send; + xpc_interface.send_notify = send_notify; + xpc_interface.received = received; + xpc_interface.partid_to_nasids = partid_to_nasids; +} + + +/* + * XPC calls this when it (the XPC module) is being unloaded. + */ +void +xpc_clear_interface(void) +{ + xpc_interface.connect = (void (*)(int)) xpc_notloaded; + xpc_interface.disconnect = (void (*)(int)) xpc_notloaded; + xpc_interface.allocate = (enum xpc_retval (*)(partid_t, int, u32, + void **)) xpc_notloaded; + xpc_interface.send = (enum xpc_retval (*)(partid_t, int, void *)) + xpc_notloaded; + xpc_interface.send_notify = (enum xpc_retval (*)(partid_t, int, void *, + xpc_notify_func, void *)) xpc_notloaded; + xpc_interface.received = (void (*)(partid_t, int, void *)) + xpc_notloaded; + xpc_interface.partid_to_nasids = (enum xpc_retval (*)(partid_t, void *)) + xpc_notloaded; +} + + +/* + * Register for automatic establishment of a channel connection whenever + * a partition comes up. + * + * Arguments: + * + * ch_number - channel # to register for connection. + * func - function to call for asynchronous notification of channel + * state changes (i.e., connection, disconnection, error) and + * the arrival of incoming messages. + * key - pointer to optional user-defined value that gets passed back + * to the user on any callouts made to func. + * payload_size - size in bytes of the XPC message's payload area which + * contains a user-defined message. The user should make + * this large enough to hold their largest message. + * nentries - max #of XPC message entries a message queue can contain. + * The actual number, which is determined when a connection + * is established and may be less then requested, will be + * passed to the user via the xpcConnected callout. + * assigned_limit - max number of kthreads allowed to be processing + * messages (per connection) at any given instant. + * idle_limit - max number of kthreads allowed to be idle at any given + * instant. + */ +enum xpc_retval +xpc_connect(int ch_number, xpc_channel_func func, void *key, u16 payload_size, + u16 nentries, u32 assigned_limit, u32 idle_limit) +{ + struct xpc_registration *registration; + + + DBUG_ON(ch_number < 0 || ch_number >= XPC_NCHANNELS); + DBUG_ON(payload_size == 0 || nentries == 0); + DBUG_ON(func == NULL); + DBUG_ON(assigned_limit == 0 || idle_limit > assigned_limit); + + registration = &xpc_registrations[ch_number]; + + if (down_interruptible(®istration->sema) != 0) { + return xpcInterrupted; + } + + /* if XPC_CHANNEL_REGISTERED(ch_number) */ + if (registration->func != NULL) { + up(®istration->sema); + return xpcAlreadyRegistered; + } + + /* register the channel for connection */ + registration->msg_size = XPC_MSG_SIZE(payload_size); + registration->nentries = nentries; + registration->assigned_limit = assigned_limit; + registration->idle_limit = idle_limit; + registration->key = key; + registration->func = func; + + up(®istration->sema); + + xpc_interface.connect(ch_number); + + return xpcSuccess; +} + + +/* + * Remove the registration for automatic connection of the specified channel + * when a partition comes up. + * + * Before returning this xpc_disconnect() will wait for all connections on the + * specified channel have been closed/torndown. So the caller can be assured + * that they will not be receiving any more callouts from XPC to their + * function registered via xpc_connect(). + * + * Arguments: + * + * ch_number - channel # to unregister. + */ +void +xpc_disconnect(int ch_number) +{ + struct xpc_registration *registration; + + + DBUG_ON(ch_number < 0 || ch_number >= XPC_NCHANNELS); + + registration = &xpc_registrations[ch_number]; + + /* + * We've decided not to make this a down_interruptible(), since we + * figured XPC's users will just turn around and call xpc_disconnect() + * again anyways, so we might as well wait, if need be. + */ + down(®istration->sema); + + /* if !XPC_CHANNEL_REGISTERED(ch_number) */ + if (registration->func == NULL) { + up(®istration->sema); + return; + } + + /* remove the connection registration for the specified channel */ + registration->func = NULL; + registration->key = NULL; + registration->nentries = 0; + registration->msg_size = 0; + registration->assigned_limit = 0; + registration->idle_limit = 0; + + xpc_interface.disconnect(ch_number); + + up(®istration->sema); + + return; +} + + +int __init +xp_init(void) +{ + int ret, ch_number; + u64 func_addr = *(u64 *) xp_nofault_PIOR; + u64 err_func_addr = *(u64 *) xp_error_PIOR; + + + if (!ia64_platform_is("sn2")) { + return -ENODEV; + } + + /* + * Register a nofault code region which performs a cross-partition + * PIO read. If the PIO read times out, the MCA handler will consume + * the error and return to a kernel-provided instruction to indicate + * an error. This PIO read exists because it is guaranteed to timeout + * if the destination is down (AMO operations do not timeout on at + * least some CPUs on Shubs <= v1.2, which unfortunately we have to + * work around). + */ + if ((ret = sn_register_nofault_code(func_addr, err_func_addr, + err_func_addr, 1, 1)) != 0) { + printk(KERN_ERR "XP: can't register nofault code, error=%d\n", + ret); + } + /* + * Setup the nofault PIO read target. (There is no special reason why + * SH_IPI_ACCESS was selected.) + */ + if (is_shub2()) { + xp_nofault_PIOR_target = SH2_IPI_ACCESS0; + } else { + xp_nofault_PIOR_target = SH1_IPI_ACCESS; + } + + /* initialize the connection registration semaphores */ + for (ch_number = 0; ch_number < XPC_NCHANNELS; ch_number++) { + sema_init(&xpc_registrations[ch_number].sema, 1); /* mutex */ + } + + return 0; +} +module_init(xp_init); + + +void __exit +xp_exit(void) +{ + u64 func_addr = *(u64 *) xp_nofault_PIOR; + u64 err_func_addr = *(u64 *) xp_error_PIOR; + + + /* unregister the PIO read nofault code region */ + (void) sn_register_nofault_code(func_addr, err_func_addr, + err_func_addr, 1, 0); +} +module_exit(xp_exit); + + +MODULE_AUTHOR("Silicon Graphics, Inc."); +MODULE_DESCRIPTION("Cross Partition (XP) base"); +MODULE_LICENSE("GPL"); + +EXPORT_SYMBOL(xp_nofault_PIOR); +EXPORT_SYMBOL(xp_nofault_PIOR_target); +EXPORT_SYMBOL(xpc_registrations); +EXPORT_SYMBOL(xpc_interface); +EXPORT_SYMBOL(xpc_clear_interface); +EXPORT_SYMBOL(xpc_set_interface); +EXPORT_SYMBOL(xpc_connect); +EXPORT_SYMBOL(xpc_disconnect); + diff --git a/arch/ia64/sn/kernel/xp_nofault.S b/arch/ia64/sn/kernel/xp_nofault.S new file mode 100644 index 00000000000..b772543053c --- /dev/null +++ b/arch/ia64/sn/kernel/xp_nofault.S @@ -0,0 +1,31 @@ +/* + * 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) 2004-2005 Silicon Graphics, Inc. All Rights Reserved. + */ + + +/* + * The xp_nofault_PIOR function takes a pointer to a remote PIO register + * and attempts to load and consume a value from it. This function + * will be registered as a nofault code block. In the event that the + * PIO read fails, the MCA handler will force the error to look + * corrected and vector to the xp_error_PIOR which will return an error. + * + * extern int xp_nofault_PIOR(void *remote_register); + */ + + .global xp_nofault_PIOR +xp_nofault_PIOR: + mov r8=r0 // Stage a success return value + ld8.acq r9=[r32];; // PIO Read the specified register + adds r9=1,r9 // Add to force a consume + br.ret.sptk.many b0;; // Return success + + .global xp_error_PIOR +xp_error_PIOR: + mov r8=1 // Return value of 1 + br.ret.sptk.many b0;; // Return failure + -- cgit v1.2.3-18-g5258 From 21223a9e78050919499d3d9039170e608eb939cc Mon Sep 17 00:00:00 2001 From: Tony Luck Date: Tue, 3 May 2005 12:25:50 -0700 Subject: [IA64] manually apply changes to arch/ia64/sn/kernel/Makefile cg-patch couldn't apply the patch to Makefile, and my dumb script rushed on and ran cg-commit without this change. Signed-off-by: Tony Luck --- arch/ia64/sn/kernel/Makefile | 4 +++- 1 file changed, 3 insertions(+), 1 deletion(-) (limited to 'arch') diff --git a/arch/ia64/sn/kernel/Makefile b/arch/ia64/sn/kernel/Makefile index 4f381fb2504..b1a4a23086b 100644 --- a/arch/ia64/sn/kernel/Makefile +++ b/arch/ia64/sn/kernel/Makefile @@ -4,10 +4,12 @@ # License. See the file "COPYING" in the main directory of this archive # for more details. # -# Copyright (C) 1999,2001-2003 Silicon Graphics, Inc. All Rights Reserved. +# Copyright (C) 1999,2001-2005 Silicon Graphics, Inc. All Rights Reserved. # obj-y += setup.o bte.o bte_error.o irq.o mca.o idle.o \ huberror.o io_init.o iomv.o klconflib.o sn2/ obj-$(CONFIG_IA64_GENERIC) += machvec.o obj-$(CONFIG_SGI_TIOCX) += tiocx.o +obj-$(CONFIG_IA64_SGI_SN_XP) += xp.o +xp-y := xp_main.o xp_nofault.o -- cgit v1.2.3-18-g5258 From 89eb8eb927e324366c3ac0458998aaf9953fc5cd Mon Sep 17 00:00:00 2001 From: Dean Nelson Date: Wed, 23 Mar 2005 19:50:00 -0700 Subject: [IA64-SGI] SGI Altix cross partition functionality [2/3] This patch contains the communication module (XPC) for cross partition communication on a partitioned SGI Altix. Signed-off-by: Dean Nelson Signed-off-by: Tony Luck --- arch/ia64/sn/kernel/Makefile | 2 + arch/ia64/sn/kernel/xpc.h | 991 +++++++++++++++ arch/ia64/sn/kernel/xpc_channel.c | 2297 +++++++++++++++++++++++++++++++++++ arch/ia64/sn/kernel/xpc_main.c | 1064 ++++++++++++++++ arch/ia64/sn/kernel/xpc_partition.c | 971 +++++++++++++++ 5 files changed, 5325 insertions(+) create mode 100644 arch/ia64/sn/kernel/xpc.h create mode 100644 arch/ia64/sn/kernel/xpc_channel.c create mode 100644 arch/ia64/sn/kernel/xpc_main.c create mode 100644 arch/ia64/sn/kernel/xpc_partition.c (limited to 'arch') diff --git a/arch/ia64/sn/kernel/Makefile b/arch/ia64/sn/kernel/Makefile index b1a4a23086b..6959736eade 100644 --- a/arch/ia64/sn/kernel/Makefile +++ b/arch/ia64/sn/kernel/Makefile @@ -13,3 +13,5 @@ obj-$(CONFIG_IA64_GENERIC) += machvec.o obj-$(CONFIG_SGI_TIOCX) += tiocx.o obj-$(CONFIG_IA64_SGI_SN_XP) += xp.o xp-y := xp_main.o xp_nofault.o +obj-$(CONFIG_IA64_SGI_SN_XP) += xpc.o +xpc-y := xpc_main.o xpc_channel.o xpc_partition.o diff --git a/arch/ia64/sn/kernel/xpc.h b/arch/ia64/sn/kernel/xpc.h new file mode 100644 index 00000000000..1a0aed8490d --- /dev/null +++ b/arch/ia64/sn/kernel/xpc.h @@ -0,0 +1,991 @@ +/* + * 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) 2004-2005 Silicon Graphics, Inc. All Rights Reserved. + */ + + +/* + * Cross Partition Communication (XPC) structures and macros. + */ + +#ifndef _IA64_SN_KERNEL_XPC_H +#define _IA64_SN_KERNEL_XPC_H + + +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include + + +/* + * XPC Version numbers consist of a major and minor number. XPC can always + * talk to versions with same major #, and never talk to versions with a + * different major #. + */ +#define _XPC_VERSION(_maj, _min) (((_maj) << 4) | ((_min) & 0xf)) +#define XPC_VERSION_MAJOR(_v) ((_v) >> 4) +#define XPC_VERSION_MINOR(_v) ((_v) & 0xf) + + +/* + * The next macros define word or bit representations for given + * C-brick nasid in either the SAL provided bit array representing + * nasids in the partition/machine or the AMO_t array used for + * inter-partition initiation communications. + * + * For SN2 machines, C-Bricks are alway even numbered NASIDs. As + * such, some space will be saved by insisting that nasid information + * passed from SAL always be packed for C-Bricks and the + * cross-partition interrupts use the same packing scheme. + */ +#define XPC_NASID_W_INDEX(_n) (((_n) / 64) / 2) +#define XPC_NASID_B_INDEX(_n) (((_n) / 2) & (64 - 1)) +#define XPC_NASID_IN_ARRAY(_n, _p) ((_p)[XPC_NASID_W_INDEX(_n)] & \ + (1UL << XPC_NASID_B_INDEX(_n))) +#define XPC_NASID_FROM_W_B(_w, _b) (((_w) * 64 + (_b)) * 2) + +#define XPC_HB_DEFAULT_INTERVAL 5 /* incr HB every x secs */ +#define XPC_HB_CHECK_DEFAULT_TIMEOUT 20 /* check HB every x secs */ + +/* define the process name of HB checker and the CPU it is pinned to */ +#define XPC_HB_CHECK_THREAD_NAME "xpc_hb" +#define XPC_HB_CHECK_CPU 0 + +/* define the process name of the discovery thread */ +#define XPC_DISCOVERY_THREAD_NAME "xpc_discovery" + + +#define XPC_HB_ALLOWED(_p, _v) ((_v)->heartbeating_to_mask & (1UL << (_p))) +#define XPC_ALLOW_HB(_p, _v) (_v)->heartbeating_to_mask |= (1UL << (_p)) +#define XPC_DISALLOW_HB(_p, _v) (_v)->heartbeating_to_mask &= (~(1UL << (_p))) + + +/* + * Reserved Page provided by SAL. + * + * SAL provides one page per partition of reserved memory. When SAL + * initialization is complete, SAL_signature, SAL_version, partid, + * part_nasids, and mach_nasids are set. + * + * Note: Until vars_pa is set, the partition XPC code has not been initialized. + */ +struct xpc_rsvd_page { + u64 SAL_signature; /* SAL unique signature */ + u64 SAL_version; /* SAL specified version */ + u8 partid; /* partition ID from SAL */ + u8 version; + u8 pad[6]; /* pad to u64 align */ + u64 vars_pa; + u64 part_nasids[XP_NASID_MASK_WORDS] ____cacheline_aligned; + u64 mach_nasids[XP_NASID_MASK_WORDS] ____cacheline_aligned; +}; +#define XPC_RP_VERSION _XPC_VERSION(1,0) /* version 1.0 of the reserved page */ + +#define XPC_RSVD_PAGE_ALIGNED_SIZE \ + (L1_CACHE_ALIGN(sizeof(struct xpc_rsvd_page))) + + +/* + * Define the structures by which XPC variables can be exported to other + * partitions. (There are two: struct xpc_vars and struct xpc_vars_part) + */ + +/* + * The following structure describes the partition generic variables + * needed by other partitions in order to properly initialize. + * + * struct xpc_vars version number also applies to struct xpc_vars_part. + * Changes to either structure and/or related functionality should be + * reflected by incrementing either the major or minor version numbers + * of struct xpc_vars. + */ +struct xpc_vars { + u8 version; + u64 heartbeat; + u64 heartbeating_to_mask; + u64 kdb_status; /* 0 = machine running */ + int act_nasid; + int act_phys_cpuid; + u64 vars_part_pa; + u64 amos_page_pa; /* paddr of page of AMOs from MSPEC driver */ + AMO_t *amos_page; /* vaddr of page of AMOs from MSPEC driver */ + AMO_t *act_amos; /* pointer to the first activation AMO */ +}; +#define XPC_V_VERSION _XPC_VERSION(3,0) /* version 3.0 of the cross vars */ + +#define XPC_VARS_ALIGNED_SIZE (L1_CACHE_ALIGN(sizeof(struct xpc_vars))) + +/* + * The following structure describes the per partition specific variables. + * + * An array of these structures, one per partition, will be defined. As a + * partition becomes active XPC will copy the array entry corresponding to + * itself from that partition. It is desirable that the size of this + * structure evenly divide into a cacheline, such that none of the entries + * in this array crosses a cacheline boundary. As it is now, each entry + * occupies half a cacheline. + */ +struct xpc_vars_part { + u64 magic; + + u64 openclose_args_pa; /* physical address of open and close args */ + u64 GPs_pa; /* physical address of Get/Put values */ + + u64 IPI_amo_pa; /* physical address of IPI AMO_t structure */ + int IPI_nasid; /* nasid of where to send IPIs */ + int IPI_phys_cpuid; /* physical CPU ID of where to send IPIs */ + + u8 nchannels; /* #of defined channels supported */ + + u8 reserved[23]; /* pad to a full 64 bytes */ +}; + +/* + * The vars_part MAGIC numbers play a part in the first contact protocol. + * + * MAGIC1 indicates that the per partition specific variables for a remote + * partition have been initialized by this partition. + * + * MAGIC2 indicates that this partition has pulled the remote partititions + * per partition variables that pertain to this partition. + */ +#define XPC_VP_MAGIC1 0x0053524156435058L /* 'XPCVARS\0'L (little endian) */ +#define XPC_VP_MAGIC2 0x0073726176435058L /* 'XPCvars\0'L (little endian) */ + + + +/* + * Functions registered by add_timer() or called by kernel_thread() only + * allow for a single 64-bit argument. The following macros can be used to + * pack and unpack two (32-bit, 16-bit or 8-bit) arguments into or out from + * the passed argument. + */ +#define XPC_PACK_ARGS(_arg1, _arg2) \ + ((((u64) _arg1) & 0xffffffff) | \ + ((((u64) _arg2) & 0xffffffff) << 32)) + +#define XPC_UNPACK_ARG1(_args) (((u64) _args) & 0xffffffff) +#define XPC_UNPACK_ARG2(_args) ((((u64) _args) >> 32) & 0xffffffff) + + + +/* + * Define a Get/Put value pair (pointers) used with a message queue. + */ +struct xpc_gp { + s64 get; /* Get value */ + s64 put; /* Put value */ +}; + +#define XPC_GP_SIZE \ + L1_CACHE_ALIGN(sizeof(struct xpc_gp) * XPC_NCHANNELS) + + + +/* + * Define a structure that contains arguments associated with opening and + * closing a channel. + */ +struct xpc_openclose_args { + u16 reason; /* reason why channel is closing */ + u16 msg_size; /* sizeof each message entry */ + u16 remote_nentries; /* #of message entries in remote msg queue */ + u16 local_nentries; /* #of message entries in local msg queue */ + u64 local_msgqueue_pa; /* physical address of local message queue */ +}; + +#define XPC_OPENCLOSE_ARGS_SIZE \ + L1_CACHE_ALIGN(sizeof(struct xpc_openclose_args) * XPC_NCHANNELS) + + + +/* struct xpc_msg flags */ + +#define XPC_M_DONE 0x01 /* msg has been received/consumed */ +#define XPC_M_READY 0x02 /* msg is ready to be sent */ +#define XPC_M_INTERRUPT 0x04 /* send interrupt when msg consumed */ + + +#define XPC_MSG_ADDRESS(_payload) \ + ((struct xpc_msg *)((u8 *)(_payload) - XPC_MSG_PAYLOAD_OFFSET)) + + + +/* + * Defines notify entry. + * + * This is used to notify a message's sender that their message was received + * and consumed by the intended recipient. + */ +struct xpc_notify { + struct semaphore sema; /* notify semaphore */ + u8 type; /* type of notification */ + + /* the following two fields are only used if type == XPC_N_CALL */ + xpc_notify_func func; /* user's notify function */ + void *key; /* pointer to user's key */ +}; + +/* struct xpc_notify type of notification */ + +#define XPC_N_CALL 0x01 /* notify function provided by user */ + + + +/* + * Define the structure that manages all the stuff required by a channel. In + * particular, they are used to manage the messages sent across the channel. + * + * This structure is private to a partition, and is NOT shared across the + * partition boundary. + * + * There is an array of these structures for each remote partition. It is + * allocated at the time a partition becomes active. The array contains one + * of these structures for each potential channel connection to that partition. + * + * Each of these structures manages two message queues (circular buffers). + * They are allocated at the time a channel connection is made. One of + * these message queues (local_msgqueue) holds the locally created messages + * that are destined for the remote partition. The other of these message + * queues (remote_msgqueue) is a locally cached copy of the remote partition's + * own local_msgqueue. + * + * The following is a description of the Get/Put pointers used to manage these + * two message queues. Consider the local_msgqueue to be on one partition + * and the remote_msgqueue to be its cached copy on another partition. A + * description of what each of the lettered areas contains is included. + * + * + * local_msgqueue remote_msgqueue + * + * |/////////| |/////////| + * w_remote_GP.get --> +---------+ |/////////| + * | F | |/////////| + * remote_GP.get --> +---------+ +---------+ <-- local_GP->get + * | | | | + * | | | E | + * | | | | + * | | +---------+ <-- w_local_GP.get + * | B | |/////////| + * | | |////D////| + * | | |/////////| + * | | +---------+ <-- w_remote_GP.put + * | | |////C////| + * local_GP->put --> +---------+ +---------+ <-- remote_GP.put + * | | |/////////| + * | A | |/////////| + * | | |/////////| + * w_local_GP.put --> +---------+ |/////////| + * |/////////| |/////////| + * + * + * ( remote_GP.[get|put] are cached copies of the remote + * partition's local_GP->[get|put], and thus their values can + * lag behind their counterparts on the remote partition. ) + * + * + * A - Messages that have been allocated, but have not yet been sent to the + * remote partition. + * + * B - Messages that have been sent, but have not yet been acknowledged by the + * remote partition as having been received. + * + * C - Area that needs to be prepared for the copying of sent messages, by + * the clearing of the message flags of any previously received messages. + * + * D - Area into which sent messages are to be copied from the remote + * partition's local_msgqueue and then delivered to their intended + * recipients. [ To allow for a multi-message copy, another pointer + * (next_msg_to_pull) has been added to keep track of the next message + * number needing to be copied (pulled). It chases after w_remote_GP.put. + * Any messages lying between w_local_GP.get and next_msg_to_pull have + * been copied and are ready to be delivered. ] + * + * E - Messages that have been copied and delivered, but have not yet been + * acknowledged by the recipient as having been received. + * + * F - Messages that have been acknowledged, but XPC has not yet notified the + * sender that the message was received by its intended recipient. + * This is also an area that needs to be prepared for the allocating of + * new messages, by the clearing of the message flags of the acknowledged + * messages. + */ +struct xpc_channel { + partid_t partid; /* ID of remote partition connected */ + spinlock_t lock; /* lock for updating this structure */ + u32 flags; /* general flags */ + + enum xpc_retval reason; /* reason why channel is disconnect'g */ + int reason_line; /* line# disconnect initiated from */ + + u16 number; /* channel # */ + + u16 msg_size; /* sizeof each msg entry */ + u16 local_nentries; /* #of msg entries in local msg queue */ + u16 remote_nentries; /* #of msg entries in remote msg queue*/ + + void *local_msgqueue_base; /* base address of kmalloc'd space */ + struct xpc_msg *local_msgqueue; /* local message queue */ + void *remote_msgqueue_base; /* base address of kmalloc'd space */ + struct xpc_msg *remote_msgqueue;/* cached copy of remote partition's */ + /* local message queue */ + u64 remote_msgqueue_pa; /* phys addr of remote partition's */ + /* local message queue */ + + atomic_t references; /* #of external references to queues */ + + atomic_t n_on_msg_allocate_wq; /* #on msg allocation wait queue */ + wait_queue_head_t msg_allocate_wq; /* msg allocation wait queue */ + + /* queue of msg senders who want to be notified when msg received */ + + atomic_t n_to_notify; /* #of msg senders to notify */ + struct xpc_notify *notify_queue;/* notify queue for messages sent */ + + xpc_channel_func func; /* user's channel function */ + void *key; /* pointer to user's key */ + + struct semaphore msg_to_pull_sema; /* next msg to pull serialization */ + struct semaphore teardown_sema; /* wait for teardown completion */ + + struct xpc_openclose_args *local_openclose_args; /* args passed on */ + /* opening or closing of channel */ + + /* various flavors of local and remote Get/Put values */ + + struct xpc_gp *local_GP; /* local Get/Put values */ + struct xpc_gp remote_GP; /* remote Get/Put values */ + struct xpc_gp w_local_GP; /* working local Get/Put values */ + struct xpc_gp w_remote_GP; /* working remote Get/Put values */ + s64 next_msg_to_pull; /* Put value of next msg to pull */ + + /* kthread management related fields */ + +// >>> rethink having kthreads_assigned_limit and kthreads_idle_limit; perhaps +// >>> allow the assigned limit be unbounded and let the idle limit be dynamic +// >>> dependent on activity over the last interval of time + atomic_t kthreads_assigned; /* #of kthreads assigned to channel */ + u32 kthreads_assigned_limit; /* limit on #of kthreads assigned */ + atomic_t kthreads_idle; /* #of kthreads idle waiting for work */ + u32 kthreads_idle_limit; /* limit on #of kthreads idle */ + atomic_t kthreads_active; /* #of kthreads actively working */ + // >>> following field is temporary + u32 kthreads_created; /* total #of kthreads created */ + + wait_queue_head_t idle_wq; /* idle kthread wait queue */ + +} ____cacheline_aligned; + + +/* struct xpc_channel flags */ + +#define XPC_C_WASCONNECTED 0x00000001 /* channel was connected */ + +#define XPC_C_ROPENREPLY 0x00000002 /* remote open channel reply */ +#define XPC_C_OPENREPLY 0x00000004 /* local open channel reply */ +#define XPC_C_ROPENREQUEST 0x00000008 /* remote open channel request */ +#define XPC_C_OPENREQUEST 0x00000010 /* local open channel request */ + +#define XPC_C_SETUP 0x00000020 /* channel's msgqueues are alloc'd */ +#define XPC_C_CONNECTCALLOUT 0x00000040 /* channel connected callout made */ +#define XPC_C_CONNECTED 0x00000080 /* local channel is connected */ +#define XPC_C_CONNECTING 0x00000100 /* channel is being connected */ + +#define XPC_C_RCLOSEREPLY 0x00000200 /* remote close channel reply */ +#define XPC_C_CLOSEREPLY 0x00000400 /* local close channel reply */ +#define XPC_C_RCLOSEREQUEST 0x00000800 /* remote close channel request */ +#define XPC_C_CLOSEREQUEST 0x00001000 /* local close channel request */ + +#define XPC_C_DISCONNECTED 0x00002000 /* channel is disconnected */ +#define XPC_C_DISCONNECTING 0x00004000 /* channel is being disconnected */ + + + +/* + * Manages channels on a partition basis. There is one of these structures + * for each partition (a partition will never utilize the structure that + * represents itself). + */ +struct xpc_partition { + + /* XPC HB infrastructure */ + + u64 remote_rp_pa; /* phys addr of partition's rsvd pg */ + u64 remote_vars_pa; /* phys addr of partition's vars */ + u64 remote_vars_part_pa; /* phys addr of partition's vars part */ + u64 last_heartbeat; /* HB at last read */ + u64 remote_amos_page_pa; /* phys addr of partition's amos page */ + int remote_act_nasid; /* active part's act/deact nasid */ + int remote_act_phys_cpuid; /* active part's act/deact phys cpuid */ + u32 act_IRQ_rcvd; /* IRQs since activation */ + spinlock_t act_lock; /* protect updating of act_state */ + u8 act_state; /* from XPC HB viewpoint */ + enum xpc_retval reason; /* reason partition is deactivating */ + int reason_line; /* line# deactivation initiated from */ + int reactivate_nasid; /* nasid in partition to reactivate */ + + + /* XPC infrastructure referencing and teardown control */ + + u8 setup_state; /* infrastructure setup state */ + wait_queue_head_t teardown_wq; /* kthread waiting to teardown infra */ + atomic_t references; /* #of references to infrastructure */ + + + /* + * NONE OF THE PRECEDING FIELDS OF THIS STRUCTURE WILL BE CLEARED WHEN + * XPC SETS UP THE NECESSARY INFRASTRUCTURE TO SUPPORT CROSS PARTITION + * COMMUNICATION. ALL OF THE FOLLOWING FIELDS WILL BE CLEARED. (THE + * 'nchannels' FIELD MUST BE THE FIRST OF THE FIELDS TO BE CLEARED.) + */ + + + u8 nchannels; /* #of defined channels supported */ + atomic_t nchannels_active; /* #of channels that are not DISCONNECTED */ + struct xpc_channel *channels;/* array of channel structures */ + + void *local_GPs_base; /* base address of kmalloc'd space */ + struct xpc_gp *local_GPs; /* local Get/Put values */ + void *remote_GPs_base; /* base address of kmalloc'd space */ + struct xpc_gp *remote_GPs;/* copy of remote partition's local Get/Put */ + /* values */ + u64 remote_GPs_pa; /* phys address of remote partition's local */ + /* Get/Put values */ + + + /* fields used to pass args when opening or closing a channel */ + + void *local_openclose_args_base; /* base address of kmalloc'd space */ + struct xpc_openclose_args *local_openclose_args; /* local's args */ + void *remote_openclose_args_base; /* base address of kmalloc'd space */ + struct xpc_openclose_args *remote_openclose_args; /* copy of remote's */ + /* args */ + u64 remote_openclose_args_pa; /* phys addr of remote's args */ + + + /* IPI sending, receiving and handling related fields */ + + int remote_IPI_nasid; /* nasid of where to send IPIs */ + int remote_IPI_phys_cpuid; /* phys CPU ID of where to send IPIs */ + AMO_t *remote_IPI_amo_va; /* address of remote IPI AMO_t structure */ + + AMO_t *local_IPI_amo_va; /* address of IPI AMO_t structure */ + u64 local_IPI_amo; /* IPI amo flags yet to be handled */ + char IPI_owner[8]; /* IPI owner's name */ + struct timer_list dropped_IPI_timer; /* dropped IPI timer */ + + spinlock_t IPI_lock; /* IPI handler lock */ + + + /* channel manager related fields */ + + atomic_t channel_mgr_requests; /* #of requests to activate chan mgr */ + wait_queue_head_t channel_mgr_wq; /* channel mgr's wait queue */ + +} ____cacheline_aligned; + + +/* struct xpc_partition act_state values (for XPC HB) */ + +#define XPC_P_INACTIVE 0x00 /* partition is not active */ +#define XPC_P_ACTIVATION_REQ 0x01 /* created thread to activate */ +#define XPC_P_ACTIVATING 0x02 /* activation thread started */ +#define XPC_P_ACTIVE 0x03 /* xpc_partition_up() was called */ +#define XPC_P_DEACTIVATING 0x04 /* partition deactivation initiated */ + + +#define XPC_DEACTIVATE_PARTITION(_p, _reason) \ + xpc_deactivate_partition(__LINE__, (_p), (_reason)) + + +/* struct xpc_partition setup_state values */ + +#define XPC_P_UNSET 0x00 /* infrastructure was never setup */ +#define XPC_P_SETUP 0x01 /* infrastructure is setup */ +#define XPC_P_WTEARDOWN 0x02 /* waiting to teardown infrastructure */ +#define XPC_P_TORNDOWN 0x03 /* infrastructure is torndown */ + + +/* + * struct xpc_partition IPI_timer #of seconds to wait before checking for + * dropped IPIs. These occur whenever an IPI amo write doesn't complete until + * after the IPI was received. + */ +#define XPC_P_DROPPED_IPI_WAIT (0.25 * HZ) + + +#define XPC_PARTID(_p) ((partid_t) ((_p) - &xpc_partitions[0])) + + + +/* found in xp_main.c */ +extern struct xpc_registration xpc_registrations[]; + + +/* >>> found in xpc_main.c only */ +extern struct device *xpc_part; +extern struct device *xpc_chan; +extern irqreturn_t xpc_notify_IRQ_handler(int, void *, struct pt_regs *); +extern void xpc_dropped_IPI_check(struct xpc_partition *); +extern void xpc_activate_kthreads(struct xpc_channel *, int); +extern void xpc_create_kthreads(struct xpc_channel *, int); +extern void xpc_disconnect_wait(int); + + +/* found in xpc_main.c and efi-xpc.c */ +extern void xpc_activate_partition(struct xpc_partition *); + + +/* found in xpc_partition.c */ +extern int xpc_exiting; +extern int xpc_hb_interval; +extern int xpc_hb_check_interval; +extern struct xpc_vars *xpc_vars; +extern struct xpc_rsvd_page *xpc_rsvd_page; +extern struct xpc_vars_part *xpc_vars_part; +extern struct xpc_partition xpc_partitions[XP_MAX_PARTITIONS + 1]; +extern char xpc_remote_copy_buffer[]; +extern struct xpc_rsvd_page *xpc_rsvd_page_init(void); +extern void xpc_allow_IPI_ops(void); +extern void xpc_restrict_IPI_ops(void); +extern int xpc_identify_act_IRQ_sender(void); +extern enum xpc_retval xpc_mark_partition_active(struct xpc_partition *); +extern void xpc_mark_partition_inactive(struct xpc_partition *); +extern void xpc_discovery(void); +extern void xpc_check_remote_hb(void); +extern void xpc_deactivate_partition(const int, struct xpc_partition *, + enum xpc_retval); +extern enum xpc_retval xpc_initiate_partid_to_nasids(partid_t, void *); + + +/* found in xpc_channel.c */ +extern void xpc_initiate_connect(int); +extern void xpc_initiate_disconnect(int); +extern enum xpc_retval xpc_initiate_allocate(partid_t, int, u32, void **); +extern enum xpc_retval xpc_initiate_send(partid_t, int, void *); +extern enum xpc_retval xpc_initiate_send_notify(partid_t, int, void *, + xpc_notify_func, void *); +extern void xpc_initiate_received(partid_t, int, void *); +extern enum xpc_retval xpc_setup_infrastructure(struct xpc_partition *); +extern enum xpc_retval xpc_pull_remote_vars_part(struct xpc_partition *); +extern void xpc_process_channel_activity(struct xpc_partition *); +extern void xpc_connected_callout(struct xpc_channel *); +extern void xpc_deliver_msg(struct xpc_channel *); +extern void xpc_disconnect_channel(const int, struct xpc_channel *, + enum xpc_retval, unsigned long *); +extern void xpc_disconnected_callout(struct xpc_channel *); +extern void xpc_partition_down(struct xpc_partition *, enum xpc_retval); +extern void xpc_teardown_infrastructure(struct xpc_partition *); + + + +static inline void +xpc_wakeup_channel_mgr(struct xpc_partition *part) +{ + if (atomic_inc_return(&part->channel_mgr_requests) == 1) { + wake_up(&part->channel_mgr_wq); + } +} + + + +/* + * These next two inlines are used to keep us from tearing down a channel's + * msg queues while a thread may be referencing them. + */ +static inline void +xpc_msgqueue_ref(struct xpc_channel *ch) +{ + atomic_inc(&ch->references); +} + +static inline void +xpc_msgqueue_deref(struct xpc_channel *ch) +{ + s32 refs = atomic_dec_return(&ch->references); + + DBUG_ON(refs < 0); + if (refs == 0) { + xpc_wakeup_channel_mgr(&xpc_partitions[ch->partid]); + } +} + + + +#define XPC_DISCONNECT_CHANNEL(_ch, _reason, _irqflgs) \ + xpc_disconnect_channel(__LINE__, _ch, _reason, _irqflgs) + + +/* + * These two inlines are used to keep us from tearing down a partition's + * setup infrastructure while a thread may be referencing it. + */ +static inline void +xpc_part_deref(struct xpc_partition *part) +{ + s32 refs = atomic_dec_return(&part->references); + + + DBUG_ON(refs < 0); + if (refs == 0 && part->setup_state == XPC_P_WTEARDOWN) { + wake_up(&part->teardown_wq); + } +} + +static inline int +xpc_part_ref(struct xpc_partition *part) +{ + int setup; + + + atomic_inc(&part->references); + setup = (part->setup_state == XPC_P_SETUP); + if (!setup) { + xpc_part_deref(part); + } + return setup; +} + + + +/* + * The following macro is to be used for the setting of the reason and + * reason_line fields in both the struct xpc_channel and struct xpc_partition + * structures. + */ +#define XPC_SET_REASON(_p, _reason, _line) \ + { \ + (_p)->reason = _reason; \ + (_p)->reason_line = _line; \ + } + + + +/* + * The following set of macros and inlines are used for the sending and + * receiving of IPIs (also known as IRQs). There are two flavors of IPIs, + * one that is associated with partition activity (SGI_XPC_ACTIVATE) and + * the other that is associated with channel activity (SGI_XPC_NOTIFY). + */ + +static inline u64 +xpc_IPI_receive(AMO_t *amo) +{ + return FETCHOP_LOAD_OP(TO_AMO((u64) &amo->variable), FETCHOP_CLEAR); +} + + +static inline enum xpc_retval +xpc_IPI_send(AMO_t *amo, u64 flag, int nasid, int phys_cpuid, int vector) +{ + int ret = 0; + unsigned long irq_flags; + + + local_irq_save(irq_flags); + + FETCHOP_STORE_OP(TO_AMO((u64) &amo->variable), FETCHOP_OR, flag); + sn_send_IPI_phys(nasid, phys_cpuid, vector, 0); + + /* + * We must always use the nofault function regardless of whether we + * are on a Shub 1.1 system or a Shub 1.2 slice 0xc processor. If we + * didn't, we'd never know that the other partition is down and would + * keep sending IPIs and AMOs to it until the heartbeat times out. + */ + ret = xp_nofault_PIOR((u64 *) GLOBAL_MMR_ADDR(NASID_GET(&amo->variable), + xp_nofault_PIOR_target)); + + local_irq_restore(irq_flags); + + return ((ret == 0) ? xpcSuccess : xpcPioReadError); +} + + +/* + * IPIs associated with SGI_XPC_ACTIVATE IRQ. + */ + +/* + * Flag the appropriate AMO variable and send an IPI to the specified node. + */ +static inline void +xpc_activate_IRQ_send(u64 amos_page, int from_nasid, int to_nasid, + int to_phys_cpuid) +{ + int w_index = XPC_NASID_W_INDEX(from_nasid); + int b_index = XPC_NASID_B_INDEX(from_nasid); + AMO_t *amos = (AMO_t *) __va(amos_page + + (XP_MAX_PARTITIONS * sizeof(AMO_t))); + + + (void) xpc_IPI_send(&amos[w_index], (1UL << b_index), to_nasid, + to_phys_cpuid, SGI_XPC_ACTIVATE); +} + +static inline void +xpc_IPI_send_activate(struct xpc_vars *vars) +{ + xpc_activate_IRQ_send(vars->amos_page_pa, cnodeid_to_nasid(0), + vars->act_nasid, vars->act_phys_cpuid); +} + +static inline void +xpc_IPI_send_activated(struct xpc_partition *part) +{ + xpc_activate_IRQ_send(part->remote_amos_page_pa, cnodeid_to_nasid(0), + part->remote_act_nasid, part->remote_act_phys_cpuid); +} + +static inline void +xpc_IPI_send_reactivate(struct xpc_partition *part) +{ + xpc_activate_IRQ_send(xpc_vars->amos_page_pa, part->reactivate_nasid, + xpc_vars->act_nasid, xpc_vars->act_phys_cpuid); +} + + +/* + * IPIs associated with SGI_XPC_NOTIFY IRQ. + */ + +/* + * Send an IPI to the remote partition that is associated with the + * specified channel. + */ +#define XPC_NOTIFY_IRQ_SEND(_ch, _ipi_f, _irq_f) \ + xpc_notify_IRQ_send(_ch, _ipi_f, #_ipi_f, _irq_f) + +static inline void +xpc_notify_IRQ_send(struct xpc_channel *ch, u8 ipi_flag, char *ipi_flag_string, + unsigned long *irq_flags) +{ + struct xpc_partition *part = &xpc_partitions[ch->partid]; + enum xpc_retval ret; + + + if (likely(part->act_state != XPC_P_DEACTIVATING)) { + ret = xpc_IPI_send(part->remote_IPI_amo_va, + (u64) ipi_flag << (ch->number * 8), + part->remote_IPI_nasid, + part->remote_IPI_phys_cpuid, + SGI_XPC_NOTIFY); + dev_dbg(xpc_chan, "%s sent to partid=%d, channel=%d, ret=%d\n", + ipi_flag_string, ch->partid, ch->number, ret); + if (unlikely(ret != xpcSuccess)) { + if (irq_flags != NULL) { + spin_unlock_irqrestore(&ch->lock, *irq_flags); + } + XPC_DEACTIVATE_PARTITION(part, ret); + if (irq_flags != NULL) { + spin_lock_irqsave(&ch->lock, *irq_flags); + } + } + } +} + + +/* + * Make it look like the remote partition, which is associated with the + * specified channel, sent us an IPI. This faked IPI will be handled + * by xpc_dropped_IPI_check(). + */ +#define XPC_NOTIFY_IRQ_SEND_LOCAL(_ch, _ipi_f) \ + xpc_notify_IRQ_send_local(_ch, _ipi_f, #_ipi_f) + +static inline void +xpc_notify_IRQ_send_local(struct xpc_channel *ch, u8 ipi_flag, + char *ipi_flag_string) +{ + struct xpc_partition *part = &xpc_partitions[ch->partid]; + + + FETCHOP_STORE_OP(TO_AMO((u64) &part->local_IPI_amo_va->variable), + FETCHOP_OR, ((u64