Message-Id: <1243445589-32388-22-git-send-email-orenl@cs.columbia.edu>
Date:	Wed, 27 May 2009 13:32:47 -0400
From:	Oren Laadan <orenl@...columbia.edu>
To:	Andrew Morton <akpm@...ux-foundation.org>
Cc:	Linus Torvalds <torvalds@...l.org>,
	containers@...ts.linux-foundation.org,
	linux-kernel@...r.kernel.org, linux-mm@...ck.org,
	linux-api@...r.kernel.org, Serge Hallyn <serue@...ibm.com>,
	Dave Hansen <dave@...ux.vnet.ibm.com>,
	Ingo Molnar <mingo@...e.hu>, "H. Peter Anvin" <hpa@...or.com>,
	Alexander Viro <viro@...iv.linux.org.uk>,
	Pavel Emelyanov <xemul@...nvz.org>,
	Alexey Dobriyan <adobriyan@...il.com>,
	Oren Laadan <orenl@...columbia.edu>
Subject: [RFC v16][PATCH 21/43] c/r: restart-blocks

(Paraphrasing what's said this message:
http://lists.openwall.net/linux-kernel/2007/12/05/64)

Restart blocks are callbacks used cause a system call to be restarted
with the arguments specified in the system call restart block. It is
useful for system call that are not idempotent, i.e. the argument(s)
might be a relative timeout, where some adjustments are required when
restarting the system call. It relies on the system call itself to set
up its restart point and the argument save area.  They are rare: an
actual signal would turn that it an EINTR. The only case that should
ever trigger this is some kernel action that interrupts the system
call, but does not actually result in any user-visible state changes -
like freeze and thaw.

So restart blocks are about time remaining for the system call to
sleep/wait. Generally in c/r, there are two possible time models that
we can follow: absolute, relative. Here, I chose to save the relative
timeout, measured from the beginning of the checkpoint. The time when
the checkpoint (and restart) begin is also saved. This information is
sufficient to restart in either model (absolute or negative).

Which model to use should eventually be a per application choice (and
possible configurable via cradvise() or some sort). For now, we adopt
the relative model, namely, at restart the timeout is set relative to
the beginning of the restart.

To checkpoint, we check if a task has a valid restart block, and if so
we save the *remaining* time that is has to wait/sleep, and the type
of the restart block.

To restart, we fill in the data required at the proper place in the
thread information. If the system call return an error (which is
possibly an -ERESTARTSYS eg), we not only use that error as our own
return value, but also arrange for the task to execute the signal
handler (by faking a signal). The handler, in turn, already has the
code to handle these restart request gracefully.

Signed-off-by: Oren Laadan <orenl@...columbia.edu>
---
 arch/x86/include/asm/checkpoint_hdr.h |    1 -
 arch/x86/mm/checkpoint.c              |   10 +-
 checkpoint/checkpoint.c               |    1 +
 checkpoint/process.c                  |  226 +++++++++++++++++++++++++++++++++
 checkpoint/restart.c                  |   35 +++++-
 checkpoint/sys.c                      |    1 +
 include/linux/checkpoint.h            |    4 +
 include/linux/checkpoint_hdr.h        |   22 +++
 include/linux/checkpoint_types.h      |    3 +
 9 files changed, 293 insertions(+), 10 deletions(-)

diff --git a/arch/x86/include/asm/checkpoint_hdr.h b/arch/x86/include/asm/checkpoint_hdr.h
index cf90170..ee23df9 100644
--- a/arch/x86/include/asm/checkpoint_hdr.h
+++ b/arch/x86/include/asm/checkpoint_hdr.h
@@ -57,7 +57,6 @@ struct ckpt_hdr_header_arch {
 
 struct ckpt_hdr_thread {
 	struct ckpt_hdr h;
-	/* FIXME: restart blocks */
 	__u16 gdt_entry_tls_entries;
 	__u16 sizeof_tls_array;
 	__u16 ntls;	/* number of TLS entries to follow */
diff --git a/arch/x86/mm/checkpoint.c b/arch/x86/mm/checkpoint.c
index c781416..7cd7494 100644
--- a/arch/x86/mm/checkpoint.c
+++ b/arch/x86/mm/checkpoint.c
@@ -63,13 +63,9 @@ int checkpoint_thread(struct ckpt_ctx *ctx, struct task_struct *t)
 	 * FIXME: the TLS descriptors in the GDT should be called out and
 	 * not tied to the in-kernel representation.
 	 */
-	ret = ckpt_write_obj_type(ctx, thread->tls_array,
-				  sizeof(thread->tls_array),
-				  CKPT_HDR_THREAD_TLS);
-
-	/* IGNORE RESTART BLOCKS FOR NOW ... */
-
-	return ret;
+	return ckpt_write_obj_type(ctx, thread->tls_array,
+				   sizeof(thread->tls_array),
+				   CKPT_HDR_THREAD_TLS);
 }
 
 #ifndef CONFIG_X86_64
diff --git a/checkpoint/checkpoint.c b/checkpoint/checkpoint.c
index 086f2d9..3999d80 100644
--- a/checkpoint/checkpoint.c
+++ b/checkpoint/checkpoint.c
@@ -23,6 +23,7 @@
 #include <linux/mount.h>
 #include <linux/utsname.h>
 #include <linux/magic.h>
+#include <linux/hrtimer.h>
 #include <linux/checkpoint.h>
 #include <linux/checkpoint_hdr.h>
 
diff --git a/checkpoint/process.c b/checkpoint/process.c
index 3ce82cb..876be3e 100644
--- a/checkpoint/process.c
+++ b/checkpoint/process.c
@@ -12,6 +12,9 @@
 #define CKPT_DFLAG  CKPT_DSYS
 
 #include <linux/sched.h>
+#include <linux/posix-timers.h>
+#include <linux/futex.h>
+#include <linux/poll.h>
 #include <linux/checkpoint.h>
 #include <linux/checkpoint_hdr.h>
 
@@ -80,6 +83,116 @@ static int checkpoint_task_objs(struct ckpt_ctx *ctx, struct task_struct *t)
 	return ret;
 }
 
+/* dump the task_struct of a given task */
+int checkpoint_restart_block(struct ckpt_ctx *ctx, struct task_struct *t)
+{
+	struct ckpt_hdr_restart_block *h;
+	struct restart_block *restart_block;
+	long (*fn)(struct restart_block *);
+	s64 base, expire = 0;
+	int ret;
+
+	h = ckpt_hdr_get_type(ctx, sizeof(*h), CKPT_HDR_RESTART_BLOCK);
+	if (!h)
+		return -ENOMEM;
+
+	base = ktime_to_ns(ctx->ktime_begin);
+	restart_block = &task_thread_info(t)->restart_block;
+	fn = restart_block->fn;
+
+	/* FIX: enumerate clockid_t so we're immune to changes */
+
+	if (fn == do_no_restart_syscall) {
+
+		h->function_type = CKPT_RESTART_BLOCK_NONE;
+		ckpt_debug("restart_block: non\n");
+
+	} else if (fn == hrtimer_nanosleep_restart) {
+
+		h->function_type = CKPT_RESTART_BLOCK_HRTIMER_NANOSLEEP;
+		h->arg_0 = restart_block->nanosleep.index;
+		h->arg_1 = (unsigned long) restart_block->nanosleep.rmtp;
+		expire = restart_block->nanosleep.expires;
+		ckpt_debug("restart_block: hrtimer expire %lld now %lld\n",
+			 expire, base);
+
+	} else if (fn == posix_cpu_nsleep_restart) {
+		struct timespec ts;
+
+		h->function_type = CKPT_RESTART_BLOCK_POSIX_CPU_NANOSLEEP;
+		h->arg_0 = restart_block->arg0;
+		h->arg_1 = restart_block->arg1;
+		ts.tv_sec = restart_block->arg2;
+		ts.tv_nsec = restart_block->arg3;
+		expire = timespec_to_ns(&ts);
+		ckpt_debug("restart_block: posix_cpu expire %lld now %lld\n",
+			 expire, base);
+
+#ifdef CONFIG_COMPAT
+	} else if (fn == compat_nanosleep_restart) {
+
+		h->function_type = CKPT_RESTART_BLOCK_NANOSLEEP;
+		h->arg_0 = restart_block->nanosleep.index;
+		h->arg_1 = (unsigned long)restart_block->nanosleep.rmtp;
+		h->arg_2 = (unsigned long)restart_block->nanosleep.compat_rmtp;
+		expire = restart_block->nanosleep.expires;
+		ckpt_debug("restart_block: compat expire %lld now %lld\n",
+			 expire, base);
+
+	} else if (fn == compat_clock_nanosleep_restart) {
+
+		h->function_type = CKPT_RESTART_BLOCK_COMPAT_CLOCK_NANOSLEEP;
+		h->arg_0 = restart_block->nanosleep.index;
+		h->arg_1 = (unsigned long)restart_block->nanosleep.rmtp;
+		h->arg_2 = (unsigned long)restart_block->nanosleep.compat_rmtp;
+		expire = restart_block->nanosleep.expires;
+		ckpt_debug("restart_block: compat_clock expire %lld now %lld\n",
+			 expire, base);
+
+#endif
+	} else if (fn == futex_wait_restart) {
+
+		h->function_type = CKPT_RESTART_BLOCK_FUTEX;
+		h->arg_0 = (unsigned long) restart_block->futex.uaddr;
+		h->arg_1 = restart_block->futex.val;
+		h->arg_2 = restart_block->futex.flags;
+		h->arg_3 = restart_block->futex.bitset;
+		expire = restart_block->futex.time;
+		ckpt_debug("restart_block: futex expire %lld now %lld\n",
+			 expire, base);
+
+	} else if (fn == do_restart_poll) {
+		struct timespec ts;
+
+		h->function_type = CKPT_RESTART_BLOCK_POLL;
+		h->arg_0 = (unsigned long) restart_block->poll.ufds;
+		h->arg_1 = restart_block->poll.nfds;
+		h->arg_2 = restart_block->poll.has_timeout;
+		ts.tv_sec = restart_block->poll.tv_sec;
+		ts.tv_nsec = restart_block->poll.tv_nsec;
+		expire = timespec_to_ns(&ts);
+		ckpt_debug("restart_block: poll expire %lld now %lld\n",
+			 expire, base);
+
+	} else {
+
+		BUG();
+
+	}
+
+	/* common to all restart blocks: */
+	h->arg_4 = (base < expire ? expire - base : 0);
+
+	ckpt_debug("restart_block: args %#llx %#llx %#llx %#llx %#llx\n",
+		 h->arg_0, h->arg_1, h->arg_2, h->arg_3, h->arg_4);
+
+	ret = ckpt_write_obj(ctx, &h->h);
+	ckpt_hdr_put(ctx, h);
+
+	ckpt_debug("restart_block ret %d\n", ret);
+	return ret;
+}
+
 /* dump the entire state of a given task */
 int checkpoint_task(struct ckpt_ctx *ctx, struct task_struct *t)
 {
@@ -97,6 +210,10 @@ int checkpoint_task(struct ckpt_ctx *ctx, struct task_struct *t)
 	ckpt_debug("thread %d\n", ret);
 	if (ret < 0)
 		goto out;
+	ret = checkpoint_restart_block(ctx, t);
+	ckpt_debug("restart-blocks %d\n", ret);
+	if (ret < 0)
+		goto out;
 	ret = checkpoint_cpu(ctx, t);
 	ckpt_debug("cpu %d\n", ret);
  out:
@@ -150,6 +267,111 @@ static int restore_task_objs(struct ckpt_ctx *ctx)
 	return ret;
 }
 
+int restore_restart_block(struct ckpt_ctx *ctx)
+{
+	struct ckpt_hdr_restart_block *h;
+	struct restart_block restart_block;
+	struct timespec ts;
+	clockid_t clockid;
+	s64 expire;
+	int ret = 0;
+
+	h = ckpt_read_obj_type(ctx, sizeof(*h), CKPT_HDR_RESTART_BLOCK);
+	if (IS_ERR(h))
+		return PTR_ERR(h);
+
+	expire = ktime_to_ns(ctx->ktime_begin) + h->arg_4;
+	restart_block.fn = NULL;
+
+	ckpt_debug("restart_block: expire %lld begin %lld\n",
+		 expire, ktime_to_ns(ctx->ktime_begin));
+	ckpt_debug("restart_block: args %#llx %#llx %#llx %#llx %#llx\n",
+		 h->arg_0, h->arg_1, h->arg_2, h->arg_3, h->arg_4);
+
+	switch (h->function_type) {
+	case CKPT_RESTART_BLOCK_NONE:
+		restart_block.fn = do_no_restart_syscall;
+		break;
+	case CKPT_RESTART_BLOCK_HRTIMER_NANOSLEEP:
+		clockid = h->arg_0;
+		if (clockid < 0 || invalid_clockid(clockid))
+			break;
+		restart_block.fn = hrtimer_nanosleep_restart;
+		restart_block.nanosleep.index = clockid;
+		restart_block.nanosleep.rmtp =
+			(struct timespec __user *) (unsigned long) h->arg_1;
+		restart_block.nanosleep.expires = expire;
+		break;
+	case CKPT_RESTART_BLOCK_POSIX_CPU_NANOSLEEP:
+		clockid = h->arg_0;
+		if (clockid < 0 || invalid_clockid(clockid))
+			break;
+		restart_block.fn = posix_cpu_nsleep_restart;
+		restart_block.arg0 = clockid;
+		restart_block.arg1 = h->arg_1;
+		ts = ns_to_timespec(expire);
+		restart_block.arg2 = ts.tv_sec;
+		restart_block.arg3 = ts.tv_nsec;
+		break;
+#ifdef CONFIG_COMPAT
+	case CKPT_RESTART_BLOCK_COMPAT_NANOSLEEP:
+		clockid = h->arg_0;
+		if (clockid < 0 || invalid_clockid(clockid))
+			break;
+		restart_block.fn = compat_nanosleep_restart;
+		restart_block.nanosleep.index = clockid;
+		restart_block.nanosleep.rmtp =
+			(struct timespec __user *) (unsigned long) h->arg_1;
+		restart_block.nanosleep.compat_rmtp =
+			(struct compat_timespec __user *)
+				(unsigned long) h->arg_2;
+		resatrt_block.nanosleep.expires = expire;
+		break;
+	case CKPT_RESTART_BLOCK_COMPAT_CLOCK_NANOSLEEP:
+		clockid = h->arg_0;
+		if (clockid < 0 || invalid_clockid(clockid))
+			break;
+		restart_block.fn = compat_clock_nanosleep_restart;
+		restart_block.nanosleep.index = clockid;
+		restart_block.nanosleep.rmtp =
+			(struct timespec __user *) (unsigned long) h->arg_1;
+		restart_block.nanosleep.compat_rmtp =
+			(struct compat_timespec __user *)
+				(unsigned long) h->arg_2;
+		resatrt_block.nanosleep.expires = expire;
+		break;
+#endif
+	case CKPT_RESTART_BLOCK_FUTEX:
+		restart_block.fn = futex_wait_restart;
+		restart_block.futex.uaddr = (u32 *) (unsigned long) h->arg_0;
+		restart_block.futex.val = h->arg_1;
+		restart_block.futex.flags = h->arg_2;
+		restart_block.futex.bitset = h->arg_3;
+		restart_block.futex.time = expire;
+		break;
+	case CKPT_RESTART_BLOCK_POLL:
+		restart_block.fn = do_restart_poll;
+		restart_block.poll.ufds =
+			(struct pollfd __user *) (unsigned long) h->arg_0;
+		restart_block.poll.nfds = h->arg_1;
+		restart_block.poll.has_timeout = h->arg_2;
+		ts = ns_to_timespec(expire);
+		restart_block.poll.tv_sec = ts.tv_sec;
+		restart_block.poll.tv_nsec = ts.tv_nsec;
+		break;
+	default:
+		break;
+	}
+
+	if (restart_block.fn)
+		task_thread_info(current)->restart_block = restart_block;
+	else
+		ret = -EINVAL;
+
+	ckpt_hdr_put(ctx, h);
+	return ret;
+}
+
 /* read the entire state of the current task */
 int restore_task(struct ckpt_ctx *ctx)
 {
@@ -167,6 +389,10 @@ int restore_task(struct ckpt_ctx *ctx)
 	ckpt_debug("thread %d\n", ret);
 	if (ret < 0)
 		goto out;
+	ret = restore_restart_block(ctx);
+	ckpt_debug("restart-blocks %d\n", ret);
+	if (ret < 0)
+		goto out;
 	ret = restore_cpu(ctx);
 	ckpt_debug("cpu %d\n", ret);
  out:
diff --git a/checkpoint/restart.c b/checkpoint/restart.c
index ca33539..e5067a9 100644
--- a/checkpoint/restart.c
+++ b/checkpoint/restart.c
@@ -16,6 +16,8 @@
 #include <linux/file.h>
 #include <linux/magic.h>
 #include <linux/utsname.h>
+#include <asm/syscall.h>
+#include <linux/elf.h>
 #include <linux/checkpoint.h>
 #include <linux/checkpoint_hdr.h>
 
@@ -357,6 +359,34 @@ static int init_restart_ctx(struct ckpt_ctx *ctx, pid_t pid)
 	return 0;
 }
 
+static int restore_retval(void)
+{
+	struct pt_regs *regs = task_pt_regs(current);
+	int ret = 0;
+
+	/*
+	 * The retval should be either zero if the checkpointed task
+	 * had been in user-space when frozen, or the retval from the
+	 * syscall that had been interrupted then.
+	 *
+	 * In the latter, if the syscall succeeded (perhaps partially)
+	 * then the retval is non-negative. If it failed, the error
+	 * may be one of -ERESTART... gang, interpreted in the signal
+	 * handling code. In restart it must happen, too.
+	 *
+	 * To force execution of the signal handler now, too, we fake
+	 * a signal to ourselves (a la freeze/thaw) when ret < 0.
+	 */
+
+	/* were we from a system call?  if so, get old error/retval */
+	if (syscall_get_nr(current, regs) >= 0)
+		ret = syscall_get_error(current, regs);
+	/* old error ?  if so, make sure signal handling kicks in */
+	if (ret < 0)
+		set_tsk_thread_flag(current, TIF_SIGPENDING);
+	return ret;
+}
+
 int do_restart(struct ckpt_ctx *ctx, pid_t pid)
 {
 	int ret;
@@ -371,7 +401,8 @@ int do_restart(struct ckpt_ctx *ctx, pid_t pid)
 	if (ret < 0)
 		return ret;
 	ret = restore_read_tail(ctx);
+	if (ret < 0)
+		return ret;
 
-	/* on success, adjust the return value if needed [TODO] */
-	return ret;
+	return restore_retval();
 }
diff --git a/checkpoint/sys.c b/checkpoint/sys.c
index c809120..5b91a39 100644
--- a/checkpoint/sys.c
+++ b/checkpoint/sys.c
@@ -194,6 +194,7 @@ static struct ckpt_ctx *ckpt_ctx_alloc(int fd, unsigned long uflags,
 
 	ctx->uflags = uflags;
 	ctx->kflags = kflags;
+	ctx->ktime_begin = ktime_get();
 
 	INIT_LIST_HEAD(&ctx->pgarr_list);
 	INIT_LIST_HEAD(&ctx->pgarr_pool);
diff --git a/include/linux/checkpoint.h b/include/linux/checkpoint.h
index 169aa70..e1204cf 100644
--- a/include/linux/checkpoint.h
+++ b/include/linux/checkpoint.h
@@ -69,6 +69,10 @@ extern int restore_thread(struct ckpt_ctx *ctx);
 extern int restore_cpu(struct ckpt_ctx *ctx);
 extern int restore_mm_context(struct ckpt_ctx *ctx, struct mm_struct *mm);
 
+extern int checkpoint_restart_block(struct ckpt_ctx *ctx,
+				    struct task_struct *t);
+extern int restore_restart_block(struct ckpt_ctx *ctx);
+
 /* file table */
 extern int checkpoint_obj_file_table(struct ckpt_ctx *ctx,
 				     struct task_struct *t);
diff --git a/include/linux/checkpoint_hdr.h b/include/linux/checkpoint_hdr.h
index 6ab3c8b..939f4e3 100644
--- a/include/linux/checkpoint_hdr.h
+++ b/include/linux/checkpoint_hdr.h
@@ -48,6 +48,7 @@ enum {
 
 	CKPT_HDR_TASK = 101,
 	CKPT_HDR_TASK_OBJS,
+	CKPT_HDR_RESTART_BLOCK,
 	CKPT_HDR_THREAD,
 	CKPT_HDR_CPU,
 
@@ -144,6 +145,27 @@ struct ckpt_hdr_task_objs {
 	__s32 mm_objref;
 } __attribute__((aligned(8)));
 
+/* restart blocks */
+struct ckpt_hdr_restart_block {
+	struct ckpt_hdr h;
+	__u64 function_type;
+	__u64 arg_0;
+	__u64 arg_1;
+	__u64 arg_2;
+	__u64 arg_3;
+	__u64 arg_4;
+} __attribute__((aligned(8)));
+
+enum restart_block_type {
+	CKPT_RESTART_BLOCK_NONE = 1,
+	CKPT_RESTART_BLOCK_HRTIMER_NANOSLEEP,
+	CKPT_RESTART_BLOCK_POSIX_CPU_NANOSLEEP,
+	CKPT_RESTART_BLOCK_COMPAT_NANOSLEEP,
+	CKPT_RESTART_BLOCK_COMPAT_CLOCK_NANOSLEEP,
+	CKPT_RESTART_BLOCK_POLL,
+	CKPT_RESTART_BLOCK_FUTEX
+};
+
 /* file system */
 struct ckpt_hdr_file_table {
 	struct ckpt_hdr h;
diff --git a/include/linux/checkpoint_types.h b/include/linux/checkpoint_types.h
index 4369f90..72052ef 100644
--- a/include/linux/checkpoint_types.h
+++ b/include/linux/checkpoint_types.h
@@ -21,12 +21,15 @@ struct ckpt_hdr_vma;
 #include <linux/list.h>
 #include <linux/path.h>
 #include <linux/fs.h>
+#include <linux/ktime.h>
 
 #include <linux/sched.h>
 
 struct ckpt_ctx {
 	int crid;		/* unique checkpoint id */
 
+	ktime_t ktime_begin;	/* checkpoint start time */
+
 	pid_t root_pid;		/* container identifier */
 	struct task_struct *root_task;	/* container root task */
 	struct nsproxy *root_nsproxy;	/* container root nsproxy */
-- 
1.6.0.4

--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@...r.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/

Powered by blists - more mailing lists