linux-kernel - [PATCH v3 4/4] mtd: phram: Allow cached mappings

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

Message-ID: <20220412135302.1682890-5-vincent.whitchurch@axis.com>
Date:   Tue, 12 Apr 2022 15:53:02 +0200
From:   Vincent Whitchurch <vincent.whitchurch@...s.com>
To:     <vigneshr@...com>, <richard@....at>, <miquel.raynal@...tlin.com>,
        <joern@...ybastard.org>
CC:     <kernel@...s.com>,
        Vincent Whitchurch <vincent.whitchurch@...s.com>,
        <linux-mtd@...ts.infradead.org>, <devicetree@...r.kernel.org>,
        <robh+dt@...nel.org>, <krzk+dt@...nel.org>,
        <frowand.list@...il.com>, <linux-kernel@...r.kernel.org>
Subject: [PATCH v3 4/4] mtd: phram: Allow cached mappings

Currently phram always uses ioremap(), but this is unnecessary when
normal memory is used.  If the reserved-memory node does not specify the
no-map property, indicating it should be mapped as system RAM and
ioremap() cannot be used on it, use a cached mapping using
memremap(MEMREMAP_WB) instead.

On one of my systems this improves read performance by ~70%.

Signed-off-by: Vincent Whitchurch <vincent.whitchurch@...s.com>
---
 drivers/mtd/devices/phram.c | 13 ++++++++++---
 1 file changed, 10 insertions(+), 3 deletions(-)

diff --git a/drivers/mtd/devices/phram.c b/drivers/mtd/devices/phram.c
index 506e9edf5c85..89d74a1eff4f 100644
--- a/drivers/mtd/devices/phram.c
+++ b/drivers/mtd/devices/phram.c
@@ -34,6 +34,7 @@
 struct phram_mtd_list {
 	struct mtd_info mtd;
 	struct list_head list;
+	bool cached;
 };
 
 static LIST_HEAD(phram_list);
@@ -96,6 +97,7 @@ static int register_device(struct platform_device *pdev, const char *name,
 			   phys_addr_t start, size_t len, uint32_t erasesize)
 {
 	struct device_node *np = pdev ? pdev->dev.of_node : NULL;
+	bool cached = np ? !of_property_read_bool(np, "no-map") : false;
 	struct phram_mtd_list *new;
 	int ret = -ENOMEM;
 
@@ -103,8 +105,13 @@ static int register_device(struct platform_device *pdev, const char *name,
 	if (!new)
 		goto out0;
 
+	new->cached = cached;
+
 	ret = -EIO;
-	new->mtd.priv = ioremap(start, len);
+	if (cached)
+		new->mtd.priv = memremap(start, len, MEMREMAP_WB);
+	else
+		new->mtd.priv = ioremap(start, len);
 	if (!new->mtd.priv) {
 		pr_err("ioremap failed\n");
 		goto out1;
@@ -140,7 +147,7 @@ static int register_device(struct platform_device *pdev, const char *name,
 	return 0;
 
 out2:
-	iounmap(new->mtd.priv);
+	cached ? memunmap(new->mtd.priv) : iounmap(new->mtd.priv);
 out1:
 	kfree(new);
 out0:
@@ -362,7 +369,7 @@ static int phram_remove(struct platform_device *pdev)
 	struct phram_mtd_list *phram = platform_get_drvdata(pdev);
 
 	mtd_device_unregister(&phram->mtd);
-	iounmap(phram->mtd.priv);
+	phram->cached ? memunmap(phram->mtd.priv) : iounmap(phram->mtd.priv);
 	kfree(phram);
 
 	return 0;
-- 
2.34.1