linux-mm.kvack.org archive mirror
 help / color / mirror / Atom feed
From: Srikar Dronamraju <srikar@linux.vnet.ibm.com>
To: Peter Zijlstra <peterz@infradead.org>
Cc: Ingo Molnar <mingo@elte.hu>, Steven Rostedt <rostedt@goodmis.org>,
	Linux-mm <linux-mm@kvack.org>,
	Arnaldo Carvalho de Melo <acme@infradead.org>,
	Linus Torvalds <torvalds@linux-foundation.org>,
	Andi Kleen <andi@firstfloor.org>, Hugh Dickins <hughd@google.com>,
	Christoph Hellwig <hch@infradead.org>,
	Jonathan Corbet <corbet@lwn.net>,
	Thomas Gleixner <tglx@linutronix.de>,
	Masami Hiramatsu <masami.hiramatsu.pt@hitachi.com>,
	Oleg Nesterov <oleg@redhat.com>,
	LKML <linux-kernel@vger.kernel.org>,
	Jim Keniston <jkenisto@linux.vnet.ibm.com>,
	Roland McGrath <roland@hack.frob.com>,
	Ananth N Mavinakayanahalli <ananth@in.ibm.com>,
	Andrew Morton <akpm@linux-foundation.org>
Subject: Re: [PATCH v4 3.0-rc2-tip 7/22]  7: uprobes: mmap and fork hooks.
Date: Wed, 22 Jun 2011 20:09:06 +0530	[thread overview]
Message-ID: <20110622143906.GF16471@linux.vnet.ibm.com> (raw)
In-Reply-To: <1308662243.26237.144.camel@twins>

* Peter Zijlstra <peterz@infradead.org> [2011-06-21 15:17:23]:

> On Fri, 2011-06-17 at 11:41 +0200, Peter Zijlstra wrote:
> > 
> > On thing I was thinking of to fix that initial problem of spurious traps
> > was to leave the uprobe in the tree but skip all probes without
> > consumers in mmap_uprobe().
> 
> Can you find fault with using __unregister_uprobe() as a cleanup path
> for __register_uprobe() so that we do a second vma-rmap walk, and
> ignoring empty probes on uprobe_mmap()?

It gets a little complicated to handle simultaneous mmaps of the same
inode/file on different processes. 

- Same uprobe cannot be in two different temporary lists at the same
  time. So we have to serialize the mmap_uprobe hook.
  
- If we use auxillary structures that refers to uprobes as nodes of
  tmplist, we dont know how many of them to preallocate. We cannot allocate
  on demand since we traverse RB tree with uprobes_treelock.

> 
> We won't get spurious traps because the empty (no consumers) uprobe is
> still in the tree, we won't get any 'lost' probe insn because the
> cleanup does a second vma-rmap walk which will include the new mmap().
> And double probe insertion is harmless.
> 

so I am thinking of a solution that includes most of your ideas along
with using i_mmap_mutex in mmap_uprobe path.

/*
Changes:
1. Uses inode->i_mutex instead of uprobes_mutex. (This is optional).
2. Now along with vma rma walk, i_mmap_mutex is even held when we do deletion of uprobes into RB tree.
3. mmap_uprobe takes i_mmap_mutex.
4. inode->uprobes_count ( Again this is optional.)


Advantages:
1. No need to drop mmap_sem.
2. Now register/unregister can run in parallel. (iff we use i_mutex);
3. No need to take extra reference to uprobe in mmap_uprobe().
*/

void _unregister_uprobe(...)
{
	if (!del_consumer(...)) {	// includes tree removal on last consumer
		return;
	}
	if (uprobe->consumers)
		return;

	mutex_lock(&inode->i_map_mutex);	//sync with mmap.
	vma_prio_tree_foreach() {
		// create list
	}

	mutex_unlock(&inode->i_map_mutex);

	list_for_each_entry_safe() {
		// remove from list
		down_read(&mm->mmap_sem);
		remove_breakpoint();	// unconditional, if it wasn't there
		up_read(&mm->mmap_sem);
	}

	mutex_lock(&inode->i_mmap_mutex);
	delete_uprobe(uprobe);
	mutex_unlock(&inode->i_mmap_mutex);

	inode->uprobes_count --;
	mutex_unlock(&inode->i_mutex);
}

int register_uprobe(...)
{
	uprobe = alloc_uprobe(...);	// find or insert in tree

	mutex_lock(&inode->i_mutex);	// sync with register/unregister
	if (uprobe->consumers) {
		add_consumer();
		goto put_unlock;
	}
	add_consumer();
	inode->uprobes_count ++;
	mutex_lock(&inode->i_map_mutex);	//sync with mmap.
	vma_prio_tree_foreach(..) {
		// get mm ref, add to list blah blah
	}

	mutex_unlock(&inode->i_map_mutex);
	list_for_each_entry_safe() {
		if (ret) {
			// del from list etc..
			//
			continue;
		}
		down_read(mm->mmap_sem);
		ret = install_breakpoint();
		up_read(..);
		// del from list etc..
		//
		if (ret && (ret == -ESRCH || ret == -EEXIST))
			ret = 0;
	}

	if (ret) {
		_unregister_uprobe();

put_unlock:
	mutex_unlock(&inode->i_mutex);
	put_uprobe(uprobe);
	return ret;
}

void unregister_uprobe(...)
{
	mutex_lock(&inode->i_mutex);	// sync with register/unregister
	uprobe = find_uprobe();	// ref++
	_unregister_uprobe();
	mutex_unlock(&inode->i_mutex);
	put_uprobe(uprobe);
}


int mmap_uprobe(struct vm_area_struct *vma)
{
	struct list_head tmp_list;
	struct uprobe *uprobe, *u;
	struct mm_struct *mm;
	struct inode *inode;
	int ret = 0;

	if (!valid_vma(vma))
		return ret;	/* Bail-out */

	mm = vma->vm_mm;
	inode = vma->vm_file->f_mapping->host;
	if (inode->uprobes_count)
		return ret;
	__iget(inode);

	INIT_LIST_HEAD(&tmp_list);

	mutex_lock(&inode->i_map_mutex);
	add_to_temp_list(vma, inode, &tmp_list);
	list_for_each_entry_safe(uprobe, u, &tmp_list, pending_list) {
		loff_t vaddr;

		list_del(&uprobe->pending_list);
		if (ret)
			continue;

		vaddr = vma->vm_start + uprobe->offset;
		vaddr -= vma->vm_pgoff << PAGE_SHIFT;
		ret = install_breakpoint(mm, uprobe, vaddr);

		if (ret && (ret == -ESRCH || ret == -EEXIST))
			ret = 0;
	}

	mutex_unlock(&inode->i_map_mutex);
	iput(inode);
	return ret;
}

int munmap_uprobe(struct vm_area_struct *vma)
{
	struct list_head tmp_list;
	struct uprobe *uprobe, *u;
	struct mm_struct *mm;
	struct inode *inode;
	int ret = 0;

	if (!valid_vma(vma))
		return ret;	/* Bail-out */

	mm = vma->vm_mm;
	inode = vma->vm_file->f_mapping->host;
	if (inode->uprobes_count)
		return ret;


//	walk thro RB tree and decrement mm->uprobes_count
	walk_rbtree_and_dec_uprobes_count(); //hold treelock.

	return ret;
}

-- 
Thanks and Regards
Srikar

--
To unsubscribe, send a message with 'unsubscribe linux-mm' in
the body to majordomo@kvack.org.  For more info on Linux MM,
see: http://www.linux-mm.org/ .
Fight unfair telecom internet charges in Canada: sign http://stopthemeter.ca/
Don't email: <a href=mailto:"dont@kvack.org"> email@kvack.org </a>

  reply	other threads:[~2011-06-22 14:48 UTC|newest]

Thread overview: 120+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2011-06-07 12:58 [PATCH v4 3.0-rc2-tip 0/22] 0: Uprobes patchset with perf probe support Srikar Dronamraju
2011-06-07 12:58 ` [PATCH v4 3.0-rc2-tip 1/22] 1: X86 specific breakpoint definitions Srikar Dronamraju
2011-06-07 12:58 ` [PATCH v4 3.0-rc2-tip 2/22] 2: uprobes: Breakground page replacement Srikar Dronamraju
2011-06-09 23:03   ` Peter Zijlstra
2011-06-13  8:48     ` Srikar Dronamraju
2011-06-09 23:03   ` Peter Zijlstra
2011-06-13  8:50     ` Srikar Dronamraju
2011-06-09 23:03   ` Peter Zijlstra
2011-06-09 23:03   ` Peter Zijlstra
2011-06-13  8:59     ` Srikar Dronamraju
2011-06-14 12:57       ` Peter Zijlstra
2011-06-14 14:57         ` Srikar Dronamraju
2011-06-09 23:03   ` Peter Zijlstra
2011-06-13  9:14     ` Srikar Dronamraju
2011-06-13 13:46   ` Oleg Nesterov
2011-06-13 17:00   ` Oleg Nesterov
2011-06-14 12:35     ` Srikar Dronamraju
2011-06-14 14:20       ` Oleg Nesterov
2011-06-15  8:55         ` Srikar Dronamraju
2011-06-15 17:54           ` Oleg Nesterov
2011-06-14 13:01     ` Peter Zijlstra
2011-06-14 14:27       ` Oleg Nesterov
2011-06-14 15:07         ` Peter Zijlstra
2011-06-14 15:40           ` Oleg Nesterov
2011-06-14 18:22             ` Peter Zijlstra
2011-06-16 12:48   ` Peter Zijlstra
2011-06-07 12:58 ` [PATCH v4 3.0-rc2-tip 3/22] 3: uprobes: Adding and remove a uprobe in a rb tree Srikar Dronamraju
2011-06-08  4:12   ` Stephen Wilson
2011-06-08  7:04     ` Josh Stone
2011-06-08 10:29       ` Srikar Dronamraju
2011-06-08 10:30   ` Srikar Dronamraju
2011-06-07 12:59 ` [PATCH v4 3.0-rc2-tip 4/22] 4: Uprobes: register/unregister probes Srikar Dronamraju
2011-06-08 22:10   ` Stephen Wilson
2011-06-09  5:43     ` Srikar Dronamraju
2011-06-09 23:03   ` Peter Zijlstra
2011-06-16  5:40     ` Srikar Dronamraju
2011-06-09 23:03   ` Peter Zijlstra
2011-06-13 16:50     ` Steven Rostedt
2011-06-16  5:26     ` Srikar Dronamraju
2011-06-16  9:42       ` Peter Zijlstra
2011-06-13 19:57   ` Oleg Nesterov
2011-06-14 12:00     ` Srikar Dronamraju
2011-06-14 14:29       ` Oleg Nesterov
2011-06-15 17:30   ` Oleg Nesterov
2011-06-16  5:09     ` Srikar Dronamraju
2011-06-15 17:41   ` Peter Zijlstra
2011-06-16  4:11     ` Srikar Dronamraju
2011-06-16  9:46       ` Peter Zijlstra
2011-06-16  9:54         ` Srikar Dronamraju
2011-06-16 10:09           ` Peter Zijlstra
2011-06-16 13:51           ` Oleg Nesterov
2011-06-17  9:29             ` Srikar Dronamraju
2011-06-15 18:01   ` Peter Zijlstra
2011-07-24 18:07   ` Oleg Nesterov
2011-07-25 12:17     ` Srikar Dronamraju
2011-06-07 12:59 ` [PATCH v4 3.0-rc2-tip 5/22] 5: x86: analyze instruction and determine fixups Srikar Dronamraju
2011-06-09 23:03   ` Peter Zijlstra
2011-06-07 12:59 ` [PATCH v4 3.0-rc2-tip 6/22] 6: uprobes: store/restore original instruction Srikar Dronamraju
2011-06-07 12:59 ` [PATCH v4 3.0-rc2-tip 7/22] 7: uprobes: mmap and fork hooks Srikar Dronamraju
2011-06-08 22:12   ` Stephen Wilson
2011-06-09  5:50     ` Srikar Dronamraju
2011-06-15 18:11   ` Peter Zijlstra
2011-06-16  3:26     ` Srikar Dronamraju
2011-06-16 12:00       ` Peter Zijlstra
2011-06-16 13:00         ` Srikar Dronamraju
2011-06-16 18:23           ` Peter Zijlstra
2011-06-16 18:25             ` Peter Zijlstra
2011-06-17  4:50             ` Srikar Dronamraju
2011-06-17  8:03               ` Peter Zijlstra
2011-06-17  9:05                 ` Srikar Dronamraju
2011-06-17  9:41                   ` Peter Zijlstra
2011-06-21 13:17                     ` Peter Zijlstra
2011-06-22 14:39                       ` Srikar Dronamraju [this message]
2011-06-24  2:06                         ` Srikar Dronamraju
2011-06-24  7:42                           ` Peter Zijlstra
2011-06-27  6:45                             ` Srikar Dronamraju
2011-06-27  8:57                               ` Peter Zijlstra
2011-07-18  9:20                                 ` Srikar Dronamraju
2011-07-18 14:31                                   ` Peter Zijlstra
2011-07-19  6:53                                     ` Srikar Dronamraju
2011-07-20 11:56                                       ` Peter Zijlstra
2011-06-07 12:59 ` [PATCH v4 3.0-rc2-tip 8/22] 8: x86: architecture specific task information Srikar Dronamraju
2011-06-07 12:59 ` [PATCH v4 3.0-rc2-tip 9/22] 9: uprobes: task specific information Srikar Dronamraju
2011-06-07 13:00 ` [PATCH v4 3.0-rc2-tip 10/22] 10: uprobes: slot allocation for uprobes Srikar Dronamraju
2011-06-07 13:00 ` [PATCH v4 3.0-rc2-tip 11/22] 11: uprobes: get the breakpoint address Srikar Dronamraju
2011-06-07 13:00 ` [PATCH v4 3.0-rc2-tip 12/22] 12: x86: x86 specific probe handling Srikar Dronamraju
2011-06-07 13:00 ` [PATCH v4 3.0-rc2-tip 13/22] 13: uprobes: Handing int3 and singlestep exception Srikar Dronamraju
2011-06-08 22:11   ` Stephen Wilson
2011-06-09  5:47     ` Srikar Dronamraju
2011-06-16 11:52   ` Peter Zijlstra
2011-06-16 12:04     ` Srikar Dronamraju
2011-06-16 12:35       ` Peter Zijlstra
2011-06-07 13:01 ` [PATCH v4 3.0-rc2-tip 14/22] 14: x86: uprobes exception notifier for x86 Srikar Dronamraju
2011-06-21 13:31   ` Peter Zijlstra
2011-06-21 13:32     ` Peter Zijlstra
2011-06-22 14:54       ` Srikar Dronamraju
2011-06-22 16:40         ` Roland McGrath
2011-06-07 13:01 ` [PATCH v4 3.0-rc2-tip 15/22] 15: uprobes: register a notifier for uprobes Srikar Dronamraju
2011-06-07 13:01 ` [PATCH v4 3.0-rc2-tip 16/22] 16: tracing: Extract out common code for kprobes/uprobes traceevents Srikar Dronamraju
2011-06-07 13:01 ` [PATCH v4 3.0-rc2-tip 17/22] 17: tracing: uprobes trace_event interface Srikar Dronamraju
2011-06-07 13:01 ` [PATCH v4 3.0-rc2-tip 18/22] 18: tracing: Uprobe tracer documentation Srikar Dronamraju
2011-06-07 13:02 ` [PATCH v4 3.0-rc2-tip 19/22] 19: perf: rename target_module to target Srikar Dronamraju
2011-06-07 13:02 ` [PATCH v4 3.0-rc2-tip 20/22] 20: perf: perf interface for uprobes Srikar Dronamraju
2011-06-07 13:30   ` Christoph Hellwig
2011-06-07 13:38     ` Ananth N Mavinakayanahalli
2011-06-07 14:21       ` Arnaldo Carvalho de Melo
2011-06-07 16:06         ` Srikar Dronamraju
2011-06-08  3:41       ` Masami Hiramatsu
2011-06-07 19:59   ` Josh Stone
2011-06-08  3:44     ` Srikar Dronamraju
2011-06-10 11:50   ` Masami Hiramatsu
2011-06-07 13:02 ` [PATCH v4 3.0-rc2-tip 21/22] 21: perf: show possible probes in a given executable file or library Srikar Dronamraju
2011-06-07 13:02 ` [PATCH v4 3.0-rc2-tip 22/22] 22: perf: Documentation for perf uprobes Srikar Dronamraju
2011-06-10 12:03   ` Masami Hiramatsu
2011-06-09 18:42 ` [PATCH v4 3.0-rc2-tip 0/22] 0: Uprobes patchset with perf probe support Peter Zijlstra
2011-06-10  5:56   ` Ananth N Mavinakayanahalli
2011-06-13  9:23   ` Srikar Dronamraju
2011-06-09 23:03 ` Peter Zijlstra
2011-06-10  6:15   ` Masami Hiramatsu
2011-06-13 10:08   ` Srikar Dronamraju

Reply instructions:

You may reply publicly to this message via plain-text email
using any one of the following methods:

* Save the following mbox file, import it into your mail client,
  and reply-to-all from there: mbox

  Avoid top-posting and favor interleaved quoting:
  https://en.wikipedia.org/wiki/Posting_style#Interleaved_style

* Reply using the --to, --cc, and --in-reply-to
  switches of git-send-email(1):

  git send-email \
    --in-reply-to=20110622143906.GF16471@linux.vnet.ibm.com \
    --to=srikar@linux.vnet.ibm.com \
    --cc=acme@infradead.org \
    --cc=akpm@linux-foundation.org \
    --cc=ananth@in.ibm.com \
    --cc=andi@firstfloor.org \
    --cc=corbet@lwn.net \
    --cc=hch@infradead.org \
    --cc=hughd@google.com \
    --cc=jkenisto@linux.vnet.ibm.com \
    --cc=linux-kernel@vger.kernel.org \
    --cc=linux-mm@kvack.org \
    --cc=masami.hiramatsu.pt@hitachi.com \
    --cc=mingo@elte.hu \
    --cc=oleg@redhat.com \
    --cc=peterz@infradead.org \
    --cc=roland@hack.frob.com \
    --cc=rostedt@goodmis.org \
    --cc=tglx@linutronix.de \
    --cc=torvalds@linux-foundation.org \
    /path/to/YOUR_REPLY

  https://kernel.org/pub/software/scm/git/docs/git-send-email.html

* If your mail client supports setting the In-Reply-To header
  via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox