mirror of https://lore.kernel.org/lkml/
 help / color / mirror / Atom feed
From: Mingming Cao <cmm@us.ibm.com>
To: akpm@digeo.com
Cc: linux-kernel@vger.kernel.org, paul.mckenny@us.ibm.com,
	dipankar@in.ibm.com, rusty@rustcorp.com.au
Subject: [PATCH] Memory barriers in IPC RCU
Date: 26 Nov 2002 11:42:46 -0800	[thread overview]
Message-ID: <1038339769.1026.79.camel@w-ming.beaverton.ibm.com> (raw)

[-- Attachment #1: Type: text/plain, Size: 866 bytes --]

Hi Andrew,

Attached patch includes the following changes:

- ipc_lock() need a read_barrier_depends() to prevent indexing
uninitialized new array on the read side. This is corresponding to the
write memory barrier added in grow_ary() from Dipankar's patch to
prevent indexing uninitialized array.

-   Replaced "wmb()" in IPC code with "smp_wmb()"."wmb()" produces a
full write memory barrier in both UP and SMP kernels, while "smp_wmb()"
provides a full write memory barrier in an SMP kernel, but only a
compiler directive in a UP kernel. The same change are made for "rmb()".

- Removed rmb() in ipc_get(). We do not need a read memory barrier there
since ipc_get() is protected by ipc_ids.sem semaphore.

- Added more comments about why write barriers and read barriers are
needed (or not needed) here or there.

Patch against 2.5.49 kernel. 


Thanks,
Mingming

[-- Attachment #2: ipc_barriers.patch --]
[-- Type: text/x-patch, Size: 4117 bytes --]

diff -urN /home/ming/views/base/linux-2.5.49/ipc/util.c 2549-ipc/ipc/util.c
--- /home/ming/views/base/linux-2.5.49/ipc/util.c	2002-11-22 13:40:39.000000000 -0800
+++ 2549-ipc/ipc/util.c	2002-11-26 10:37:35.000000000 -0800
@@ -98,6 +98,10 @@
 	struct kern_ipc_perm* p;
 	int max_id = ids->max_id;
 
+	/*
+	 * read_barrier_depends is not needed here
+	 * since ipc_ids.sem is held
+	 */
 	for (id = 0; id <= max_id; id++) {
 		p = ids->entries[id].p;
 		if(p==NULL)
@@ -134,12 +138,12 @@
 
 	/*
 	 * before setting the ids->entries to the new array, there must be a
-	 * wmb() to make sure that the memcpyed contents of the new array are
+	 * smp_wmb() to make sure the memcpyed contents of the new array are
 	 * visible before the new array becomes visible.
 	 */
-	wmb();
+	smp_wmb();	/* prevent seeing new array uninitialized. */
 	ids->entries = new;
-	wmb();
+	smp_wmb();	/* prevent indexing into old array based on new size. */
 	ids->size = newsize;
 
 	ipc_rcu_free(old, sizeof(struct ipc_id)*i);
@@ -156,6 +160,8 @@
  *	initialised and the first free entry is set up and the id assigned
  *	is returned. The list is returned in a locked state on success.
  *	On failure the list is not locked and -1 is returned.
+ *
+ *	Called with ipc_ids.sem held.
  */
  
 int ipc_addid(struct ipc_ids* ids, struct kern_ipc_perm* new, int size)
@@ -163,6 +169,11 @@
 	int id;
 
 	size = grow_ary(ids,size);
+
+	/*
+	 * read_barrier_depends() is not needed here since
+	 * ipc_ids.sem is held
+	 */
 	for (id = 0; id < size; id++) {
 		if(ids->entries[id].p == NULL)
 			goto found;
@@ -207,7 +218,11 @@
 	int lid = id % SEQ_MULTIPLIER;
 	if(lid >= ids->size)
 		BUG();
-	
+
+	/* 
+	 * do not need a read_barrier_depends() here to force ordering
+	 * on Alpha, since the ipc_ids.sem is held.
+	 */	
 	p = ids->entries[lid].p;
 	ids->entries[lid].p = NULL;
 	if(p==NULL)
@@ -414,15 +429,15 @@
 }
 
 /*
- * ipc_get() requires ipc_ids.sem down, otherwise we need a rmb() here
- * to sync with grow_ary();
- *
- * So far only shm_get_stat() uses ipc_get() via shm_get().  So ipc_get()
- * is called with shm_ids.sem locked.  Thus a rmb() is not needed here,
- * as grow_ary() also requires shm_ids.sem down(for shm).
- *
- * But if ipc_get() is used in the future without ipc_ids.sem down,
- * we need to add a rmb() before accessing the entries array
+ * So far only shm_get_stat() calls ipc_get() via shm_get(), so ipc_get()
+ * is called with shm_ids.sem locked.  Since grow_ary() is also called with
+ * shm_ids.sem down(for Shared Memory), there is no need to add read 
+ * barriers here to gurantee the writes in grow_ary() are seen in order 
+ * here (for Alpha).
+ *
+ * However ipc_get() itself does not necessary require ipc_ids.sem down. So
+ * if in the future ipc_get() is used by other places without ipc_ids.sem
+ * down, then ipc_get() needs read memery barriers as ipc_lock() does.
  */
 struct kern_ipc_perm* ipc_get(struct ipc_ids* ids, int id)
 {
@@ -430,7 +445,6 @@
 	int lid = id % SEQ_MULTIPLIER;
 	if(lid >= ids->size)
 		return NULL;
-	rmb();
 	out = ids->entries[lid].p;
 	return out;
 }
@@ -439,6 +453,7 @@
 {
 	struct kern_ipc_perm* out;
 	int lid = id % SEQ_MULTIPLIER;
+	struct ipc_id* entries;
 
 	rcu_read_lock();
 	if(lid >= ids->size) {
@@ -446,9 +461,18 @@
 		return NULL;
 	}
 
-	/* we need a barrier here to sync with grow_ary() */
-	rmb();
-	out = ids->entries[lid].p;
+	/* 
+	 * Note: The following two read barriers are corresponding
+	 * to the two write barriers in grow_ary(). They gurantee 
+	 * the writes are seen in the same order on the read side. 
+	 * smp_rmb() has effect on all CPUs.  read_barrier_depends() 
+	 * is used if there are data dependency between two reads, and 
+	 * has effect only on Alpha.
+	 */
+	smp_rmb(); /* prevent indexing old array with new size */
+	entries = ids->entries;
+	read_barrier_depends(); /*prevent seeing new array unitialized */
+	out = entries[lid].p;
 	if(out == NULL) {
 		rcu_read_unlock();
 		return NULL;

                 reply	other threads:[~2002-11-26 19:36 UTC|newest]

Thread overview: [no followups] expand[flat|nested]  mbox.gz  Atom feed

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=1038339769.1026.79.camel@w-ming.beaverton.ibm.com \
    --to=cmm@us.ibm.com \
    --cc=akpm@digeo.com \
    --cc=dipankar@in.ibm.com \
    --cc=linux-kernel@vger.kernel.org \
    --cc=paul.mckenny@us.ibm.com \
    --cc=rusty@rustcorp.com.au \
    /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

all inboxes | Powered by JetHome®