linux-kernel - [PATCH] MN10300: Fix flush_icache

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [day] [month] [year] [list]

Message-ID: <20101001093103.10891.25807.stgit@warthog.procyon.org.uk>
Date:	Fri, 01 Oct 2010 10:31:03 +0100
From:	David Howells <dhowells@...hat.com>
To:	torvalds@...l.org, akpm@...ux-foundation.org
Cc:	linux-am33-list@...hat.com, linux-kernel@...r.kernel.org,
	David Howells <dhowells@...hat.com>,
	Akira Takeuchi <takeuchi.akr@...panasonic.com>
Subject: [PATCH] MN10300: Fix flush_icache_range()

flush_icache_range() is given virtual addresses to describe the region.  It
deals with these by attempting to translate them through the current set of
page tables.

This is fine for userspace memory and vmalloc()'d areas as they are governed by
page tables.  However, since the regions above 0x80000000 aren't translated
through the page tables by the MMU, the kernel doesn't bother to set up page
tables for them (see paging_init()).

This means flush_icache_range() as it stands cannot be used to flush regions of
the VM area between 0x80000000 and 0x9fffffff where the kernel resides if the
data cache is operating in WriteBack mode.

To fix this, make flush_icache_range() first check for addresses in the upper
half of VM space and deal with them appropriately, before dealing with any
range in the page table mapped area.

Ordinarily, this is not a problem, but it has the capacity to make kprobes and
kgdb malfunction.  It should not affect gdbstub, signal frame setup or module
loading as gdb has its own flush functions, and the others take place in the
page table mapped area only.

Signed-off-by: David Howells <dhowells@...hat.com>
Acked-by: Akira Takeuchi <takeuchi.akr@...panasonic.com>
---

 arch/mn10300/mm/cache.c |   20 +++++++++++++++++++-
 1 files changed, 19 insertions(+), 1 deletions(-)

diff --git a/arch/mn10300/mm/cache.c b/arch/mn10300/mm/cache.c
index 1b76719..9261217 100644
--- a/arch/mn10300/mm/cache.c
+++ b/arch/mn10300/mm/cache.c
@@ -54,13 +54,30 @@ EXPORT_SYMBOL(flush_icache_page);
 void flush_icache_range(unsigned long start, unsigned long end)
 {
 #ifdef CONFIG_MN10300_CACHE_WBACK
-	unsigned long addr, size, off;
+	unsigned long addr, size, base, off;
 	struct page *page;
 	pgd_t *pgd;
 	pud_t *pud;
 	pmd_t *pmd;
 	pte_t *ppte, pte;

+	if (end > 0x80000000UL) {
+		/* addresses above 0xa0000000 do not go through the cache */
+		if (end > 0xa0000000UL) {
+			end = 0xa0000000UL;
+			if (start >= end)
+				return;
+		}
+
+		/* kernel addresses between 0x80000000 and 0x9fffffff do not
+		 * require page tables, so we just map such addresses directly */
+		base = (start >= 0x80000000UL) ? start : 0x80000000UL;
+		mn10300_dcache_flush_range(base, end);
+		if (base == start)
+			goto invalidate;
+		end = base;
+	}
+
 	for (; start < end; start += size) {
 		/* work out how much of the page to flush */
 		off = start & (PAGE_SIZE - 1);
@@ -104,6 +121,7 @@ void flush_icache_range(unsigned long start, unsigned long end)
 	}
 #endif

+invalidate:
 	mn10300_icache_inv();
 }
 EXPORT_SYMBOL(flush_icache_range);

--
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/