Kernel updates since December 2009

include/l4/generic/bootmem.h

@@ -9,6 +9,4 @@ unsigned long bootmem_free_pages(void);
 void *alloc_bootmem(int size, int alignment);
 pmd_table_t *alloc_boot_pmd(void);
 
-extern pgd_table_t init_pgd;
-
 #endif /* __BOOTMEM_H__ */

include/l4/generic/cap-types.h

@@ -46,7 +46,6 @@
  */
 #define CAP_DEVTYPE_TIMER		1
 #define CAP_DEVTYPE_UART		2
-#define CAP_DEVTYPE_CLCD		3
 #define CAP_DEVTYPE_OTHER		0xF
 #define CAP_DEVTYPE_MASK		0xFFFF
 #define CAP_DEVNUM_MASK			0xFFFF0000
@@ -102,6 +101,10 @@
 #define CAP_MAP_UNMAP		(1 << 5)
 #define CAP_MAP_UTCB		(1 << 6)
 
+/* Cache operations, applicable to (virtual) memory regions */
+#define CAP_CACHE_INVALIDATE		(1 << 7)
+#define CAP_CACHE_CLEAN			(1 << 8)
+
 /*
  * IRQ Control capability
  */
@@ -137,4 +140,6 @@
 #define CAP_CAP_DESTROY		(1 << 6)
 #define CAP_CAP_MODIFY		(CAP_CAP_DEDUCE | CAP_CAP_SPLIT \
 				 | CAP_CAP_DESTROY)
+
+
 #endif /* __CAP_TYPES_H__ */

include/l4/generic/capability.h

@@ -104,6 +104,8 @@ struct capability *cap_find_by_capid(l4id_t capid, struct cap_list **clist);
 /* Capability checking on system calls */
 int cap_map_check(struct ktcb *task, unsigned long phys, unsigned long virt,
 		  unsigned long npages, unsigned int flags);
+int cap_unmap_check(struct ktcb *task, unsigned long virt,
+		    unsigned long npages);
 int cap_thread_check(struct ktcb *task, unsigned int flags,
 		     struct task_ids *ids);
 int cap_exregs_check(struct ktcb *task, struct exregs_data *exregs);
@@ -114,5 +116,7 @@ int cap_mutex_check(unsigned long mutex_address, int mutex_op);
 
 int cap_irq_check(struct ktcb *registrant, unsigned int req,
 		  unsigned int flags, l4id_t irq);
+int cap_cache_check(unsigned long start, unsigned long end,
+		    unsigned int flags);
 
 #endif /* __GENERIC_CAPABILITY_H__ */

include/l4/generic/container.h

@@ -106,8 +106,7 @@ void kres_insert_container(struct container *c,
 
 struct container *container_create(void);
 
-int container_init_pagers(struct kernel_resources *kres,
-			  pgd_table_t *current_pgd);
+int container_init_pagers(struct kernel_resources *kres);
 
 int init_containers(struct kernel_resources *kres);
 struct container *container_find(struct kernel_resources *kres, l4id_t cid);

include/l4/generic/debug.h

@@ -0,0 +1,162 @@
+/*
+ * Definitions for kernel entry accounting.
+ *
+ * Copyright (C) 2010 B Labs Ltd.
+ *
+ * Written by Bahadir Balban
+ */
+#ifndef __GENERIC_DEBUG_H__
+#define __GENERIC_DEBUG_H__
+
+#include INC_ARCH(types.h)
+#include INC_SUBARCH(cache.h)
+#include <l4/lib/printk.h>
+
+#if defined(CONFIG_DEBUG_ACCOUNTING)
+
+struct exception_count {
+	u64 syscall;
+	u64 data_abort;
+	u64 prefetch_abort;
+	u64 irq;
+	u64 undefined_abort;
+};
+
+/*
+ * Note these are packed to match systable offsets
+ * so that they're incremented with an auccess
+ */
+struct syscall_count {
+	u64 ipc;
+	u64 tswitch;
+	u64 tctrl;
+	u64 exregs;
+	u64 emtpy;
+	u64 unmap;
+	u64 irqctrl;
+	u64 empty1;
+	u64 map;
+	u64 getid;
+	u64 capctrl;
+	u64 empty2;
+	u64 time;
+	u64 mutexctrl;
+	u64 cachectrl;
+} __attribute__ ((__packed__));
+
+struct task_op_count {
+	u64 context_switch;
+	u64 space_switch;
+};
+
+struct cache_op_count {
+	u64 dcache_clean_mva;
+	u64 dcache_inval_mva;
+	u64 icache_clean_mva;
+	u64 icache_inval_mva;
+	u64 dcache_clean_setway;
+	u64 dcache_inval_setway;
+	u64 tlb_mva;
+};
+
+#if defined(CONFIG_DEBUG_PERFMON_KERNEL)
+
+/* Minimum, maximum and average timings for the call */
+struct syscall_timing {
+	u64 total;
+	u32 min;
+	u32 max;
+	u32 avg;
+};
+
+struct syscall_timings {
+	struct syscall_timing ipc;
+	struct syscall_timing tswitch;
+	struct syscall_timing tctrl;
+	struct syscall_timing exregs;
+	struct syscall_timing emtpy;
+	struct syscall_timing unmap;
+	struct syscall_timing irqctrl;
+	struct syscall_timing empty1;
+	struct syscall_timing map;
+	struct syscall_timing getid;
+	struct syscall_timing capctrl;
+	struct syscall_timing empty2;
+	struct syscall_timing time;
+	struct syscall_timing mutexctrl;
+	struct syscall_timing cachectrl;
+	u64 all_total;
+} __attribute__ ((__packed__));
+
+extern struct syscall_timings syscall_timings;
+
+
+#endif /* End of CONFIG_DEBUG_PERFMON_KERNEL */
+
+struct system_accounting {
+	struct syscall_count syscalls;
+
+#if defined(CONFIG_DEBUG_PERFMON_KERNEL)
+	struct syscall_timings syscall_timings;
+#endif
+
+	struct exception_count exceptions;
+	struct cache_op_count cache_ops;
+	struct task_op_count task_ops;
+} __attribute__ ((__packed__));
+
+
+extern struct system_accounting system_accounting;
+
+static inline void system_account_dabort(void)
+{
+	system_accounting.exceptions.data_abort++;
+}
+
+static inline void system_account_pabort(void)
+{
+	system_accounting.exceptions.prefetch_abort++;
+}
+
+static inline void system_account_undef_abort(void)
+{
+	system_accounting.exceptions.undefined_abort++;
+}
+
+static inline void system_account_irq(void)
+{
+	system_accounting.exceptions.irq++;
+}
+
+static inline void system_account_syscall(void)
+{
+	system_accounting.exceptions.syscall++;
+}
+
+static inline void system_account_context_switch(void)
+{
+	system_accounting.task_ops.context_switch++;
+}
+
+static inline void system_account_space_switch(void)
+{
+	system_accounting.task_ops.space_switch++;
+}
+
+#include INC_SUBARCH(debug.h)
+
+#else /* End of CONFIG_DEBUG_ACCOUNTING */
+
+static inline void system_account_cache_op(int op) { }
+static inline void system_account_irq(void) { }
+static inline void system_account_syscall(void) { }
+static inline void system_account_dabort(void) { }
+static inline void system_account_pabort(void) { }
+static inline void system_account_undef_abort(void) { }
+static inline void system_account_space_switch(void) { }
+static inline void system_account_context_switch(void) { }
+
+#endif /* End of !CONFIG_DEBUG_ACCOUNTING */
+
+
+#endif /* __GENERIC_DEBUG_H__ */

include/l4/generic/irq.h

@@ -19,8 +19,8 @@
 
 typedef void (*irq_op_t)(l4id_t irq);
 struct irq_chip_ops {
-	void (*init)(void);
-	l4id_t (*read_irq)(void);
+	void (*init)();
+	l4id_t (*read_irq)(void *data);
 	irq_op_t ack_and_mask;
 	irq_op_t unmask;
 };
@@ -31,6 +31,7 @@ struct irq_chip {
 	int cascade;		/* The irq that lower chip uses on this chip */
 	int start;		/* The global irq offset for this chip */
 	int end;		/* End of this chip's irqs */
+	void *data;		/* Anything that a of interest to a driver */
 	struct irq_chip_ops ops;
 };
 

include/l4/generic/platform.h

@@ -2,21 +2,32 @@
 #define __PLATFORM_H__
 /*
  * Generic functions to be provided by every platform.
+ *
+ * Include only those API's that are needed by sources
+ * outside the src/platform code.
  */
 
+#include <l4/generic/resource.h>
+
 void platform_init(void);
 
-/* Uart APIs */
-void uart_init(void);
-void uart_putc(char c);
-
-/* Timer APIs */
-void timer_init(void);
-void timer_start(void);
-
 /* IRQ controller */
 void irq_controller_init(void);
 void platform_irq_enable(int irq);
 void platform_irq_disable(int irq);
 
+#define	dprintk(str, val)		\
+{					\
+	print_early(str);		\
+	printhex8((val));		\
+	print_early("\n");		\
+}
+
+void print_early(char *str);
+void printhex8(unsigned int);
+
+int platform_setup_device_caps(struct kernel_resources *kres);
+
+void platform_test_cpucycles(void);
+
 #endif /* __PLATFORM_H__ */

include/l4/generic/preempt.h

@@ -12,4 +12,5 @@ int preempt_count(void);
 int in_nested_irq_context(void);
 int in_irq_context(void);
 int in_task_context(void);
+
 #endif /* __PREEMPT_H__ */

include/l4/generic/resource.h

@@ -92,15 +92,20 @@ struct kernel_resources {
 	struct mem_cache *cont_cache;
 
 	/* Zombie thread list */
-	struct ktcb_list zombie_list;
+	DECLARE_PERCPU(struct ktcb_list, zombie_list);
+
+#if defined(CONFIG_SUBARCH_V7)
+	/* Global page tables on split page tables */
+	pgd_global_table_t *pgd_global;
+#endif
 };
 
 extern struct kernel_resources kernel_resources;
 
 void free_pgd(void *addr);
 void free_pmd(void *addr);
-void free_space(void *addr);
-void free_ktcb(void *addr);
+void free_space(void *addr, struct ktcb *task);
+void free_ktcb(void *addr, struct ktcb *task);
 void free_capability(void *addr);
 void free_container(void *addr);
 void free_user_mutex(void *addr);
@@ -118,4 +123,6 @@ int free_boot_memory(struct kernel_resources *kres);
 
 int init_system_resources(struct kernel_resources *kres);
 
+void setup_idle_caps(); /*TODO: Delete this when done with it */
+
 #endif /* __RESOURCES_H__ */

include/l4/generic/scheduler.h

@@ -7,8 +7,11 @@
 #define __SCHEDULER_H__
 
 #include <l4/generic/tcb.h>
+#include <l4/generic/smp.h>
+#include INC_SUBARCH(cpu.h)
 #include INC_SUBARCH(mm.h)
 #include INC_GLUE(memory.h)
+#include INC_GLUE(smp.h)
 
 /* Task priorities */
 #define TASK_PRIO_MAX		10
@@ -19,14 +22,16 @@
 #define TASK_PRIO_LOW		2
 #define TASK_PRIO_TOTAL		30
 
-/* Ticks per second, try ticks = 1000 + timeslice = 1 for regressed preemption test. */
-#define SCHED_TICKS				100
+/*
+ * CONFIG_SCHED_TICKS gives ticks per second.
+ * try ticks = 1000, and timeslice = 1 for regressed preemption test.
+ */
 
 /*
  * A task can run continuously at this granularity,
  * even if it has a greater total time slice.
  */
-#define SCHED_GRANULARITY			SCHED_TICKS/50
+#define SCHED_GRANULARITY			CONFIG_SCHED_TICKS/10
 
 static inline struct ktcb *current_task(void)
 {
@@ -37,13 +42,13 @@ static inline struct ktcb *current_task(void)
 #define current			current_task()
 #define need_resched		(current->ts_need_resched)
 
-#define SCHED_RQ_TOTAL					2
-
+#define SCHED_RQ_TOTAL			2
 
 /* A basic runqueue */
 struct runqueue {
+	struct scheduler *sched;
 	struct spinlock lock;		/* Lock */
-	struct link task_list;	/* List of tasks in rq */
+	struct link task_list;		/* List of tasks in rq */
 	unsigned int total;		/* Total tasks */
 };
 
@@ -52,22 +57,25 @@ struct scheduler {
 	struct runqueue sched_rq[SCHED_RQ_TOTAL];
 	struct runqueue *rq_runnable;
 	struct runqueue *rq_expired;
+	struct ktcb *idle_task;
 
 	/* Total priority of all tasks in container */
 	int prio_total;
 };
-extern struct scheduler scheduler;
 
-void sched_init_runqueue(struct runqueue *rq);
+DECLARE_PERCPU(extern struct scheduler, scheduler);
+
+void sched_init_runqueue(struct scheduler *sched, struct runqueue *rq);
 void sched_init_task(struct ktcb *task, int priority);
 void sched_prepare_sleep(void);
-void sched_exit_sync(void);
 void sched_suspend_sync(void);
 void sched_suspend_async(void);
 void sched_resume_sync(struct ktcb *task);
 void sched_resume_async(struct ktcb *task);
+void sched_enqueue_task(struct ktcb *first_time_runner, int sync);
 void scheduler_start(void);
 void schedule(void);
-void sched_init(struct scheduler *scheduler);
+void sched_init(void);
+void idle_task(void);
 
 #endif /* __SCHEDULER_H__ */

include/l4/generic/smp.h

@@ -0,0 +1,23 @@
+/*
+ * Copyright 2010 B Labs Ltd.
+ *
+ * Author: Prem Mallappa <prem.mallappa@b-labs.co.uk>
+ */
+
+#ifndef __GENERIC_SMP_H__
+#define __GENERIC_SMP_H__
+
+#include INC_SUBARCH(cpu.h)
+
+/* IPIs, we define more as we go */
+/* we have limited IPI's on ARM, exactly 15 */
+#define IPI_TLB_FLUSH	0x00000001
+#define IPI_SCHEDULE	0x00000002
+#define IPI_CACH_FLUSH	0x00000003
+
+#if !defined (CONFIG_NCPU)
+#define CONFIG_NCPU	1
+#define smp_get_cpuid()		0
+#endif
+
+#endif	/* __GENERIC_SMP_H__ */

include/l4/generic/space.h

@@ -1,22 +1,31 @@
 /*
  * Generic address space related information.
  *
- * Copyright (C) 2007 Bahadir Balban
+ * Copyright (C) 2007-2010 Bahadir Balban
  */
 #ifndef __SPACE_H__
 #define __SPACE_H__
 
-/* The flags not embedded in the name behave as expected. E.g USR_RW is also */
-#define MAP_USR_RW_FLAGS	0	/* CB as one would expect */
-#define MAP_USR_RO_FLAGS	1	/* CB as one would expect */
-#define MAP_SVC_RW_FLAGS	2	/* CB as one would expect */
-#define MAP_USR_IO_FLAGS	3	/* Non-CB, RW TODO: How about RO one? */
-#define MAP_SVC_IO_FLAGS	4	/* Non-CB, RW */
+/*
+ * Generic mapping flags.
+ */
+#define MAP_FAULT			0
+#define MAP_USR_RW			1
+#define MAP_USR_RO			2
+#define MAP_KERN_RW			3
+#define MAP_USR_IO			4
+#define MAP_KERN_IO			5
+#define MAP_USR_RWX			6
+#define MAP_KERN_RWX			7
+#define MAP_USR_RX			8
+#define MAP_KERN_RX			9
+#define MAP_UNMAP			10	/* For unmap syscall */
+#define MAP_INVALID_FLAGS 		(1 << 31)
 
 /* Some default aliases */
-#define	MAP_USR_DEFAULT_FLAGS	MAP_USR_RW_FLAGS
-#define MAP_SVC_DEFAULT_FLAGS	MAP_SVC_RW_FLAGS
-#define MAP_IO_DEFAULT_FLAGS	MAP_SVC_IO_FLAGS
+#define	MAP_USR_DEFAULT		MAP_USR_RW
+#define MAP_KERN_DEFAULT	MAP_KERN_RW
+#define MAP_IO_DEFAULT		MAP_KERN_IO
 
 #if defined (__KERNEL__)
 
@@ -46,11 +55,14 @@ struct address_space_list {
 };
 
 struct address_space *address_space_create(struct address_space *orig);
-void address_space_delete(struct address_space *space);
+void address_space_delete(struct address_space *space,
+			  struct ktcb *task_accounted);
 void address_space_attach(struct ktcb *tcb, struct address_space *space);
 struct address_space *address_space_find(l4id_t spid);
 void address_space_add(struct address_space *space);
-void address_space_remove(struct address_space *space);
+
+struct container;
+void address_space_remove(struct address_space *space, struct container *cont);
 void init_address_space_list(struct address_space_list *space_list);
 int check_access(unsigned long vaddr, unsigned long size,
 		 unsigned int flags, int page_in);

include/l4/generic/tcb.h

@@ -28,15 +28,20 @@
 #define TASK_INTERRUPTED		(1 << 0)
 #define TASK_SUSPENDING			(1 << 1)
 #define TASK_RESUMING			(1 << 2)
-#define TASK_EXITING			(1 << 3)
-#define TASK_PENDING_SIGNAL		(TASK_SUSPENDING | TASK_EXITING)
+#define TASK_PENDING_SIGNAL		(TASK_SUSPENDING)
+
+/*
+ * This is to indicate a task (either current or one of
+ * its children) exit has occured and cleanup needs to be
+ * called
+ */
+#define TASK_EXITED			(1 << 3)
 
 /* Task states */
 enum task_state {
 	TASK_INACTIVE	= 0,
 	TASK_SLEEPING	= 1,
 	TASK_RUNNABLE	= 2,
-	TASK_DEAD	= 3,
 };
 
 #define TASK_CID_MASK			0xFF000000
@@ -82,6 +87,9 @@ struct ktcb {
 	l4id_t tid;		/* Global thread id */
 	l4id_t tgid;		/* Global thread group id */
 
+	/* CPU affinity */
+	int affinity;
+
 	/* Other related threads */
 	l4id_t pagerid;
 
@@ -94,6 +102,9 @@ struct ktcb {
 	/* Lock for blocking thread state modifications via a syscall */
 	struct mutex thread_control_lock;
 
+	/* To protect against thread deletion/modification */
+	struct spinlock thread_lock;
+
 	u32 ts_need_resched;	/* Scheduling flag */
 	enum task_state state;
 
@@ -171,13 +182,14 @@ static inline void set_task_ids(struct ktcb *task, struct task_ids *ids)
 }
 
 struct ktcb *tcb_find(l4id_t tid);
+struct ktcb *tcb_find_lock(l4id_t tid);
 void tcb_add(struct ktcb *tcb);
 void tcb_remove(struct ktcb *tcb);
 
 void tcb_init(struct ktcb *tcb);
 struct ktcb *tcb_alloc_init(l4id_t cid);
 void tcb_delete(struct ktcb *tcb);
-
+void tcb_delete_zombies(void);
 
 void ktcb_list_remove(struct ktcb *task, struct ktcb_list *ktcb_list);
 void ktcb_list_add(struct ktcb *new, struct ktcb_list *ktcb_list);

include/l4/generic/thread.h

@@ -11,6 +11,7 @@ void thread_id_pool_init(void);
 int thread_id_new(void);
 int thread_id_del(int tid);
 
+void thread_setup_affinity(struct ktcb *task);
 void thread_destroy(struct ktcb *);
 
 #endif /* __GENERIC_THREAD_H__ */

include/l4/generic/time.h

@@ -4,8 +4,8 @@
  * Copyright (C) 2007 Bahadir Balban
  */
 
-#ifndef __GENERIC_TIMER_H__
-#define __GENERIC_TIMER_H__
+#ifndef __GENERIC_TIME_H__
+#define __GENERIC_TIME_H__
 
 /* Used by posix systems */
 struct timeval {
@@ -17,4 +17,4 @@ extern volatile u32 jiffies;
 
 int do_timer_irq(void);
 
-#endif /* __GENERIC_TIMER_H__ */
+#endif /* __GENERIC_TIME_H__ */