Hexagon: Add init_task and process functions
authorRichard Kuo <rkuo@codeaurora.org>
Mon, 31 Oct 2011 23:41:49 +0000 (18:41 -0500)
committerLinus Torvalds <torvalds@linux-foundation.org>
Tue, 1 Nov 2011 14:34:19 +0000 (07:34 -0700)
Signed-off-by: Richard Kuo <rkuo@codeaurora.org>
Acked-by: Arnd Bergmann <arnd@arndb.de>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
arch/hexagon/kernel/init_task.c [new file with mode: 0644]
arch/hexagon/kernel/process.c [new file with mode: 0644]

diff --git a/arch/hexagon/kernel/init_task.c b/arch/hexagon/kernel/init_task.c
new file mode 100644 (file)
index 0000000..73283d3
--- /dev/null
@@ -0,0 +1,54 @@
+/*
+ * Init task definition
+ *
+ * Copyright (c) 2010-2011, Code Aurora Forum. All rights reserved.
+ *
+ * This program is free software; you can redistribute it and/or modify
+ * it under the terms of the GNU General Public License version 2 and
+ * only version 2 as published by the Free Software Foundation.
+ *
+ * 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., 51 Franklin Street, Fifth Floor, Boston, MA
+ * 02110-1301, USA.
+ */
+
+#include <linux/mm.h>
+#include <linux/module.h>
+#include <linux/sched.h>
+#include <linux/init_task.h>
+#include <linux/fs.h>
+#include <linux/mqueue.h>
+#include <asm/thread_info.h>
+#include <asm/uaccess.h>
+#include <asm/pgtable.h>
+
+static struct signal_struct init_signals = INIT_SIGNALS(init_signals);
+static struct sighand_struct init_sighand = INIT_SIGHAND(init_sighand);
+
+/*
+ * Initial thread structure.
+ *
+ * We need to make sure that this is 8192-byte aligned due to the
+ * way process stacks are handled. This is done by making sure
+ * the linker maps this in the .text segment right after head.S,
+ * and making head.S ensure the proper alignment.
+ */
+union thread_union init_thread_union
+       __attribute__((__section__(".data.init_task"),
+               __aligned__(THREAD_SIZE))) = {
+                       INIT_THREAD_INFO(init_task)
+               };
+
+/*
+ * Initial task structure.
+ *
+ * All other task structs will be allocated on slabs in fork.c
+ */
+struct task_struct init_task = INIT_TASK(init_task);
+EXPORT_SYMBOL(init_task);
diff --git a/arch/hexagon/kernel/process.c b/arch/hexagon/kernel/process.c
new file mode 100644 (file)
index 0000000..18c4f0b
--- /dev/null
@@ -0,0 +1,279 @@
+/*
+ * Process creation support for Hexagon
+ *
+ * Copyright (c) 2010-2011, Code Aurora Forum. All rights reserved.
+ *
+ * This program is free software; you can redistribute it and/or modify
+ * it under the terms of the GNU General Public License version 2 and
+ * only version 2 as published by the Free Software Foundation.
+ *
+ * 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., 51 Franklin Street, Fifth Floor, Boston, MA
+ * 02110-1301, USA.
+ */
+
+#include <linux/sched.h>
+#include <linux/types.h>
+#include <linux/module.h>
+#include <linux/tick.h>
+#include <linux/uaccess.h>
+#include <linux/slab.h>
+
+/*
+ * Kernel thread creation.  The desired kernel function is "wrapped"
+ * in the kernel_thread_helper function, which does cleanup
+ * afterwards.
+ */
+static void __noreturn kernel_thread_helper(void *arg, int (*fn)(void *))
+{
+       do_exit(fn(arg));
+}
+
+int kernel_thread(int (*fn)(void *), void *arg, unsigned long flags)
+{
+       struct pt_regs regs;
+
+       memset(&regs, 0, sizeof(regs));
+       /*
+        * Yes, we're exploting illicit knowledge of the ABI here.
+        */
+       regs.r00 = (unsigned long) arg;
+       regs.r01 = (unsigned long) fn;
+       pt_set_elr(&regs, (unsigned long)kernel_thread_helper);
+       pt_set_kmode(&regs);
+
+       return do_fork(flags|CLONE_VM|CLONE_UNTRACED, 0, &regs, 0, NULL, NULL);
+}
+EXPORT_SYMBOL(kernel_thread);
+
+/*
+ * Program thread launch.  Often defined as a macro in processor.h,
+ * but we're shooting for a small footprint and it's not an inner-loop
+ * performance-critical operation.
+ *
+ * The Hexagon ABI specifies that R28 is zero'ed before program launch,
+ * so that gets automatically done here.  If we ever stop doing that here,
+ * we'll probably want to define the ELF_PLAT_INIT macro.
+ */
+void start_thread(struct pt_regs *regs, unsigned long pc, unsigned long sp)
+{
+       /* Set to run with user-mode data segmentation */
+       set_fs(USER_DS);
+       /* We want to zero all data-containing registers. Is this overkill? */
+       memset(regs, 0, sizeof(*regs));
+       /* We might want to also zero all Processor registers here */
+       pt_set_usermode(regs);
+       pt_set_elr(regs, pc);
+       pt_set_rte_sp(regs, sp);
+}
+
+/*
+ *  Spin, or better still, do a hardware or VM wait instruction
+ *  If hardware or VM offer wait termination even though interrupts
+ *  are disabled.
+ */
+static void default_idle(void)
+{
+       __vmwait();
+}
+
+void (*idle_sleep)(void) = default_idle;
+
+void cpu_idle(void)
+{
+       while (1) {
+               tick_nohz_stop_sched_tick(1);
+               local_irq_disable();
+               while (!need_resched()) {
+                       idle_sleep();
+                       /*  interrupts wake us up, but aren't serviced  */
+                       local_irq_enable();     /* service interrupt   */
+                       local_irq_disable();
+               }
+               local_irq_enable();
+               tick_nohz_restart_sched_tick();
+               schedule();
+       }
+}
+
+/*
+ *  Return saved PC of a blocked thread
+ */
+unsigned long thread_saved_pc(struct task_struct *tsk)
+{
+       return 0;
+}
+
+/*
+ * Copy architecture-specific thread state
+ */
+int copy_thread(unsigned long clone_flags, unsigned long usp,
+               unsigned long unused, struct task_struct *p,
+               struct pt_regs *regs)
+{
+       struct thread_info *ti = task_thread_info(p);
+       struct hexagon_switch_stack *ss;
+       struct pt_regs *childregs;
+       asmlinkage void ret_from_fork(void);
+
+       childregs = (struct pt_regs *) (((unsigned long) ti + THREAD_SIZE) -
+                                       sizeof(*childregs));
+
+       memcpy(childregs, regs, sizeof(*childregs));
+       ti->regs = childregs;
+
+       /*
+        * Establish kernel stack pointer and initial PC for new thread
+        */
+       ss = (struct hexagon_switch_stack *) ((unsigned long) childregs -
+                                                   sizeof(*ss));
+       ss->lr = (unsigned long)ret_from_fork;
+       p->thread.switch_sp = ss;
+
+       /* If User mode thread, set pt_reg stack pointer as per parameter */
+       if (user_mode(childregs)) {
+               pt_set_rte_sp(childregs, usp);
+
+               /* Child sees zero return value */
+               childregs->r00 = 0;
+
+               /*
+                * The clone syscall has the C signature:
+                * int [r0] clone(int flags [r0],
+                *           void *child_frame [r1],
+                *           void *parent_tid [r2],
+                *           void *child_tid [r3],
+                *           void *thread_control_block [r4]);
+                * ugp is used to provide TLS support.
+                */
+               if (clone_flags & CLONE_SETTLS)
+                       childregs->ugp = childregs->r04;
+
+               /*
+                * Parent sees new pid -- not necessary, not even possible at
+                * this point in the fork process
+                * Might also want to set things like ti->addr_limit
+                */
+       } else {
+               /*
+                * If kernel thread, resume stack is kernel stack base.
+                * Note that this is pointer arithmetic on pt_regs *
+                */
+               pt_set_rte_sp(childregs, (unsigned long)(childregs + 1));
+               /*
+                * We need the current thread_info fast path pointer
+                * set up in pt_regs.  The register to be used is
+                * parametric for assembler code, but the mechanism
+                * doesn't drop neatly into C.  Needs to be fixed.
+                */
+               childregs->THREADINFO_REG = (unsigned long) ti;
+       }
+
+       /*
+        * thread_info pointer is pulled out of task_struct "stack"
+        * field on switch_to.
+        */
+       p->stack = (void *)ti;
+
+       return 0;
+}
+
+/*
+ * Release any architecture-specific resources locked by thread
+ */
+void release_thread(struct task_struct *dead_task)
+{
+}
+
+/*
+ * Free any architecture-specific thread data structures, etc.
+ */
+void exit_thread(void)
+{
+}
+
+/*
+ * Some archs flush debug and FPU info here
+ */
+void flush_thread(void)
+{
+}
+
+/*
+ * The "wait channel" terminology is archaic, but what we want
+ * is an identification of the point at which the scheduler
+ * was invoked by a blocked thread.
+ */
+unsigned long get_wchan(struct task_struct *p)
+{
+       unsigned long fp, pc;
+       unsigned long stack_page;
+       int count = 0;
+       if (!p || p == current || p->state == TASK_RUNNING)
+               return 0;
+
+       stack_page = (unsigned long)task_stack_page(p);
+       fp = ((struct hexagon_switch_stack *)p->thread.switch_sp)->fp;
+       do {
+               if (fp < (stack_page + sizeof(struct thread_info)) ||
+                       fp >= (THREAD_SIZE - 8 + stack_page))
+                       return 0;
+               pc = ((unsigned long *)fp)[1];
+               if (!in_sched_functions(pc))
+                       return pc;
+               fp = *(unsigned long *) fp;
+       } while (count++ < 16);
+
+       return 0;
+}
+
+/*
+ * Borrowed from PowerPC -- basically allow smaller kernel stacks if we
+ * go crazy with the page sizes.
+ */
+#if THREAD_SHIFT < PAGE_SHIFT
+
+static struct kmem_cache *thread_info_cache;
+
+struct thread_info *alloc_thread_info_node(struct task_struct *tsk, int node)
+{
+       struct thread_info *ti;
+
+       ti = kmem_cache_alloc_node(thread_info_cache, GFP_KERNEL, node);
+       if (unlikely(ti == NULL))
+               return NULL;
+#ifdef CONFIG_DEBUG_STACK_USAGE
+       memset(ti, 0, THREAD_SIZE);
+#endif
+       return ti;
+}
+
+void free_thread_info(struct thread_info *ti)
+{
+       kmem_cache_free(thread_info_cache, ti);
+}
+
+/*  Weak symbol; called by init/main.c  */
+
+void thread_info_cache_init(void)
+{
+       thread_info_cache = kmem_cache_create("thread_info", THREAD_SIZE,
+                                             THREAD_SIZE, 0, NULL);
+       BUG_ON(thread_info_cache == NULL);
+}
+
+#endif /* THREAD_SHIFT < PAGE_SHIFT */
+
+/*
+ * Required placeholder.
+ */
+int dump_fpu(struct pt_regs *regs, elf_fpregset_t *fpu)
+{
+       return 0;
+}