BPF List
 help / color / mirror / Atom feed
From: Peter Zijlstra <peterz@infradead.org>
To: Matthew Wilcox <willy@infradead.org>
Cc: "Paul E. McKenney" <paulmck@kernel.org>,
	Andrii Nakryiko <andrii.nakryiko@gmail.com>,
	Masami Hiramatsu <mhiramat@kernel.org>,
	mingo@kernel.org, andrii@kernel.org,
	linux-kernel@vger.kernel.org, rostedt@goodmis.org,
	oleg@redhat.com, jolsa@kernel.org, clm@meta.com,
	bpf <bpf@vger.kernel.org>
Subject: Re: [PATCH 00/10] perf/uprobe: Optimize uprobes
Date: Wed, 10 Jul 2024 11:40:13 +0200	[thread overview]
Message-ID: <20240710094013.GF28838@noisy.programming.kicks-ass.net> (raw)
In-Reply-To: <20240710091631.GT27299@noisy.programming.kicks-ass.net>

On Wed, Jul 10, 2024 at 11:16:31AM +0200, Peter Zijlstra wrote:

> If it were an actual sequence count, I could make it work, but sadly,
> not. Also, vma_end_write() seems to be missing :-( If anything it could
> be used to lockdep annotate the thing.
> 
> Mooo.. I need to stare more at this to see if perhaps it can be made to
> work, but so far, no joy :/

See, this is what I want, except I can't close the race against VMA
modification because of that crazy locking scheme :/


--- a/kernel/events/uprobes.c
+++ b/kernel/events/uprobes.c
@@ -2146,11 +2146,58 @@ static int is_trap_at_addr(struct mm_str
 	return is_trap_insn(&opcode);
 }
 
-static struct uprobe *find_active_uprobe(unsigned long bp_vaddr, int *is_swbp)
+#ifndef CONFIG_PER_VMA_LOCK
+static struct uprobe *__find_active_uprobe(unsigned long bp_vaddr)
+{
+	return NULL;
+}
+#else
+static struct uprobe *__find_active_uprobe(unsigned long bp_vaddr)
 {
 	struct mm_struct *mm = current->mm;
 	struct uprobe *uprobe = NULL;
 	struct vm_area_struct *vma;
+	MA_STATE(mas, &mm->mm_mt, bp_vaddr, bp_vaddr);
+
+	guard(rcu)();
+
+again:
+	vma = mas_walk(&mas);
+	if (!vma)
+		return NULL;
+
+	/* vma_write_start() -- in progress */
+	if (READ_ONCE(vma->vm_lock_seq) == READ_ONCE(vma->vm_mm->mm_lock_seq))
+		return NULL;
+
+	/* 
+	 * Completely broken, because of the crazy vma locking scheme you
+	 * cannot avoid the per-vma rwlock and doing so means you're racy
+	 * against modifications.
+	 *
+	 * A simple actual seqcount would'be been cheaper and more usefull.
+	 */
+
+	if (!valid_vma(vma, false))
+		return NULL;
+
+	struct inode = file_inode(vma->vm_file);
+	loff_t offset = vaddr_to_offset(vma, bp_vaddr);
+
+	// XXX: if (vma_seq_retry(...)) goto again;
+
+	return find_uprobe(inode, offset);
+}
+#endif
+
+static struct uprobe *find_active_uprobe(unsigned long bp_vaddr, int *is_swbp)
+{
+	struct uprobe *uprobe = __find_active_uprobe(bp_vaddr)
+	struct mm_struct *mm = current->mm;
+	struct vm_area_struct *vma;
+
+	if (uprobe)
+		return uprobe;
 
 	mmap_read_lock(mm);
 	vma = vma_lookup(mm, bp_vaddr);

  reply	other threads:[~2024-07-10  9:40 UTC|newest]

Thread overview: 52+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
     [not found] <20240708091241.544262971@infradead.org>
     [not found] ` <20240709075651.122204f1358f9f78d1e64b62@kernel.org>
2024-07-09  0:25   ` [PATCH 00/10] perf/uprobe: Optimize uprobes Andrii Nakryiko
2024-07-09  9:01     ` Peter Zijlstra
2024-07-09 14:11       ` Paul E. McKenney
2024-07-09 14:29         ` Peter Zijlstra
2024-07-09 14:36           ` Paul E. McKenney
2024-07-09 15:31             ` Peter Zijlstra
2024-07-09 15:56               ` Paul E. McKenney
2024-07-09 16:10           ` Matthew Wilcox
2024-07-09 16:30             ` Matthew Wilcox
2024-07-09 16:57               ` Paul E. McKenney
2024-07-10  9:16             ` Peter Zijlstra
2024-07-10  9:40               ` Peter Zijlstra [this message]
2024-07-22 19:09                 ` Suren Baghdasaryan
2024-07-27  0:20                   ` Andrii Nakryiko
2024-07-27  1:29                     ` Suren Baghdasaryan
2024-07-27  3:45                       ` Matthew Wilcox
2024-07-30  3:18                         ` Andrii Nakryiko
2024-07-30 13:10                         ` Peter Zijlstra
2024-07-30 18:10                           ` Suren Baghdasaryan
2024-08-03  5:47                             ` Andrii Nakryiko
2024-08-03  8:53                               ` Peter Zijlstra
2024-08-04 23:22                                 ` Andrii Nakryiko
2024-08-06  4:08                                   ` Andrii Nakryiko
2024-08-06 14:50                                     ` Suren Baghdasaryan
2024-08-06 17:40                                       ` Andrii Nakryiko
2024-08-06 17:44                                         ` Suren Baghdasaryan
2024-08-07  1:36                                     ` Suren Baghdasaryan
2024-08-07  5:13                                       ` Suren Baghdasaryan
2024-08-07 17:49                                         ` Andrii Nakryiko
2024-08-07 18:04                                           ` Suren Baghdasaryan
2024-08-07 18:30                                             ` Andrii Nakryiko
2024-08-07 18:33                                             ` Suren Baghdasaryan
2024-08-08  0:47                                               ` Andrii Nakryiko
2024-07-30 13:46                   ` Peter Zijlstra
2024-07-30 18:16                     ` Suren Baghdasaryan
2024-07-09 16:42         ` Andrii Nakryiko
2024-07-09  9:03     ` Peter Zijlstra
2024-07-09 10:01       ` Jiri Olsa
2024-07-09 10:16         ` Peter Zijlstra
2024-07-09 22:10           ` Masami Hiramatsu
2024-07-10 10:10             ` Peter Zijlstra
2024-07-10 14:56               ` Masami Hiramatsu
2024-07-10 18:40                 ` Andrii Nakryiko
2024-07-11  8:51                   ` Peter Zijlstra
2024-07-11 15:17                     ` Masami Hiramatsu
2024-07-11 15:22                       ` Peter Zijlstra
2024-07-11 17:47                         ` Steven Rostedt
2024-07-11 23:59                           ` Masami Hiramatsu
2024-07-10  0:55       ` Masami Hiramatsu
2024-07-09 21:47     ` Andrii Nakryiko
2024-07-10 10:12       ` Peter Zijlstra
2024-07-10 12:34         ` Peter Zijlstra

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=20240710094013.GF28838@noisy.programming.kicks-ass.net \
    --to=peterz@infradead.org \
    --cc=andrii.nakryiko@gmail.com \
    --cc=andrii@kernel.org \
    --cc=bpf@vger.kernel.org \
    --cc=clm@meta.com \
    --cc=jolsa@kernel.org \
    --cc=linux-kernel@vger.kernel.org \
    --cc=mhiramat@kernel.org \
    --cc=mingo@kernel.org \
    --cc=oleg@redhat.com \
    --cc=paulmck@kernel.org \
    --cc=rostedt@goodmis.org \
    --cc=willy@infradead.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