From: Patrick Farrell <pfarrell@whamcloud.com>
Date: Thu, 28 Sep 2023 00:09:03 +0000 (-0400)
Subject: LU-10026 ptlrpc: verify large allocations are aligned
X-Git-Tag: 2.16.56~22
X-Git-Url: https://git.whamcloud.com/?a=commitdiff_plain;h=4c972aa6696a50ddfbc0dcb3e8db36f4b435fa4e;p=fs%2Flustre-release.git

LU-10026 ptlrpc: verify large allocations are aligned

If we were ever to do an allocation with kmalloc(), we could
get non-page-aligned memory.

We haven't seen any problem yet, but we are trying to be ready for
this in advance. The arguments from https://lwn.net/Articles/787740/
also look strong.

Let's add an assertion to illuminate this dangerous behaviour.

EX-bug-id: EX-8245
Signed-off-by: Patrick Farrell <pfarrell@whamcloud.com>
Signed-off-by: Artem Blagodarenko <ablagodarenko@ddn.com>
Change-Id: Id20898065b516d363d9dc280e71be1b5cfb6f4a7
Reviewed-on: https://review.whamcloud.com/c/fs/lustre-release/+/57844
Reviewed-by: Andreas Dilger <adilger@whamcloud.com>
Reviewed-by: Oleg Drokin <green@whamcloud.com>
Reviewed-by: Sergey Cheremencev <scherementsev@ddn.com>
Tested-by: Maloo <maloo@whamcloud.com>
Tested-by: jenkins <devops@whamcloud.com>
---

diff --git a/lustre/obdclass/page_pools.c b/lustre/obdclass/page_pools.c
index edc4b0b..34d8695 100644
--- a/lustre/obdclass/page_pools.c
+++ b/lustre/obdclass/page_pools.c
@@ -569,6 +569,18 @@ static int pool_add_objects(int nobjects, struct obd_page_pool *page_pool)
 			else {
 				OBD_ALLOC_LARGE(ptr_pages[i][j],
 					object_size(page_pool));
+				/*
+				 * It is possible that at some moment kmalloc
+				 * will start to return non-page aligned memory,
+				 * so leave this assort for quicker problem
+				 * detection
+				 */
+				LASSERTF(IS_ALIGNED((unsigned long)
+						    (ptr_pages[i][j]),
+						    PAGE_SIZE),
+				    "Page %p (order %i) is not aligned to PAGE_SIZE",
+				    ptr_pages[i][j], page_pool->opp_order);
+
 			}
 			if (ptr_pages[i][j] == NULL)
 				goto out_ptr_pages;