Skip to contentSkip to conversation

Fix various issues with RFD and FTLs

PATCH Applied Sean Young 2 participants 11 messages Aug 7, 2021 – Aug 16, 2021 Memory technology devices (MTD)

lore
Share & tools
lore

Sean Young

author
#0 lore raw

v2 cover Fix various issues with RFD and FTLs

Changes since v2:
- Review by Miquel Raynal
- Improved commit message for discard and mtdram
- Split use-after-free commit into two commits; see container-of()

Sean Young (5):
mtd: rfd_ftl: allow use of MTD_RAM for testing purposes
mtd: rfd_ftl: add discard support
mtd: blk_devs: make discard work on FTLs
mtd: rfd_ftl: fix use-after-free
mtd: rfd_ftl: use container_of() rather than cast

 drivers/mtd/mtd_blkdevs.c |  1 +
 drivers/mtd/rfd_ftl.c     | 44 ++++++++++++++++++++++++++++++++-------
 2 files changed, 38 insertions(+), 7 deletions(-)
Signature · 6 lines
2.31.1


______________________________________________________
Linux MTD discussion mailing list
http://lists.infradead.org/mailman/listinfo/linux-mtd/
author
#1 lore raw

PATCH v2 1/5 mtd: rfd_ftl: allow use of MTD_RAM for testing purposes

This allows the rfd_ftl to be used with the mtdram module, so we can test different mtd sizes and test the rfd_ftl on machines without a physical nor flash device.

---
 drivers/mtd/rfd_ftl.c | 3 ++-
 1 file changed, 2 insertions(+), 1 deletion(-)
drivers/mtd/rfd_ftl.c +2 -1
index 6e0d5ce9b010..7b243f2b2fa3 100644
@@ -720,7 +720,8 @@ static void rfd_ftl_add_mtd(struct mtd_blktrans_ops *tr, struct mtd_info *mtd)
 {
 	struct partition *part;
 
-	if (mtd->type != MTD_NORFLASH || mtd->size > UINT_MAX)
+	if ((mtd->type != MTD_NORFLASH && mtd->type != MTD_RAM) ||
+	    mtd->size > UINT_MAX)
 		return;
 
 	part = kzalloc(sizeof(struct partition), GFP_KERNEL);
Signature · 6 lines
2.31.1


______________________________________________________
Linux MTD discussion mailing list
http://lists.infradead.org/mailman/listinfo/linux-mtd/
maintainer Applied
#2 lore raw

Re: [PATCH v2 1/5] mtd: rfd_ftl: allow use of MTD_RAM for testing purposes

Quoted 5 lines · Sean Young

This allows the rfd_ftl to be used with the mtdram module, so we can test different mtd sizes and test the rfd_ftl on machines without a physical nor flash device.

Applied to https://git.kernel.org/pub/scm/linux/kernel/git/mtd/linux.git mtd/next, thanks.

Miquel

______________________________________________________
Linux MTD discussion mailing list
http://lists.infradead.org/mailman/listinfo/linux-mtd/

author
#3 lore raw

PATCH v2 2/5 mtd: rfd_ftl: add discard support

In reply to Sean Young

I proposed this change 16 years ago before discard was a feature in the block layer: https://lwn.net/Articles/162776/

Now that the block layer has discard, we can finally merge this change.

Discard is also known as trim. By implementing discard, both fstrim and the discard filesystem option can be used.

Implementing discard in the ftl means that when files are removed, there is less data in the ftl mapping. This means less stuff to move around for erasing and also less erasing to do; this means improved wear levelling and improved performance.

---
 drivers/mtd/rfd_ftl.c | 29 +++++++++++++++++++++++++++++
 1 file changed, 29 insertions(+)
drivers/mtd/rfd_ftl.c +29 -0
index 7b243f2b2fa3..7f5f6d247cae 100644
@@ -705,6 +705,34 @@ static int rfd_ftl_writesect(struct mtd_blktrans_dev *dev, u_long sector, char *
 	return rc;
 }
 
+static int rfd_ftl_discardsect(struct mtd_blktrans_dev *dev,
+			       unsigned long sector, unsigned int nr_sects)
+{
+	struct partition *part = (struct partition *)dev;
+	u_long addr;
+	int rc;
+
+	while (nr_sects) {
+		if (sector >= part->sector_count)
+			return -EIO;
+
+		addr = part->sector_map[sector];
+
+		if (addr != -1) {
+			rc = mark_sector_deleted(part, addr);
+			if (rc)
+				return rc;
+
+			part->sector_map[sector] = -1;
+		}
+
+		sector++;
+		nr_sects--;
+	}
+
+	return 0;
+}
+
 static int rfd_ftl_getgeo(struct mtd_blktrans_dev *dev, struct hd_geometry *geo)
 {
 	struct partition *part = (struct partition*)dev;
@@ -786,6 +814,7 @@ static struct mtd_blktrans_ops rfd_ftl_tr = {
 
 	.readsect	= rfd_ftl_readsect,
 	.writesect	= rfd_ftl_writesect,
+	.discard	= rfd_ftl_discardsect,
 	.getgeo		= rfd_ftl_getgeo,
 	.add_mtd	= rfd_ftl_add_mtd,
 	.remove_dev	= rfd_ftl_remove_dev,
Signature · 6 lines
2.31.1


______________________________________________________
Linux MTD discussion mailing list
http://lists.infradead.org/mailman/listinfo/linux-mtd/
maintainer Applied
#4 lore raw

Re: [PATCH v2 2/5] mtd: rfd_ftl: add discard support

Quoted 8 lines · Sean Young

I proposed this change 16 years ago before discard was a feature in the block layer: https://lwn.net/Articles/162776/

Now that the block layer has discard, we can finally merge this change.

Discard is also known as trim. By implementing discard, both fstrim and the discard filesystem option can be used.

Quoted 6 lines

Implementing discard in the ftl means that when files are removed, there is less data in the ftl mapping. This means less stuff to move around for erasing and also less erasing to do; this means improved wear levelling and improved performance.

Applied to https://git.kernel.org/pub/scm/linux/kernel/git/mtd/linux.git mtd/next, thanks.

Miquel

______________________________________________________
Linux MTD discussion mailing list
http://lists.infradead.org/mailman/listinfo/linux-mtd/

author
#5 lore raw

PATCH v2 3/5 mtd: blk_devs: make discard work on FTLs

In reply to Sean Young

If the discard_granularity is not set, discard will not work and the following error is logged:

rfda: Error: discard_granularity is 0.

Since all the FTLs use a sector size of 512, this can be hardcoded.

---
 drivers/mtd/mtd_blkdevs.c | 1 +
 1 file changed, 1 insertion(+)
drivers/mtd/mtd_blkdevs.c +1 -0
index 6ce4bc57f919..486251e058a6 100644
@@ -419,6 +419,7 @@ int add_mtd_blktrans_dev(struct mtd_blktrans_dev *new)
 	if (tr->discard) {
 		blk_queue_flag_set(QUEUE_FLAG_DISCARD, new->rq);
 		blk_queue_max_discard_sectors(new->rq, UINT_MAX);
+		new->rq->limits.discard_granularity = 512;
 	}
 
 	gd->queue = new->rq;
Signature · 6 lines
2.31.1


______________________________________________________
Linux MTD discussion mailing list
http://lists.infradead.org/mailman/listinfo/linux-mtd/
maintainer Applied
#6 lore raw

Re: [PATCH v2 3/5] mtd: blk_devs: make discard work on FTLs

Quoted 8 lines · Sean Young

If the discard_granularity is not set, discard will not work and the following error is logged:

rfda: Error: discard_granularity is 0.

Since all the FTLs use a sector size of 512, this can be hardcoded.

Applied to https://git.kernel.org/pub/scm/linux/kernel/git/mtd/linux.git mtd/next, thanks.

Miquel

______________________________________________________
Linux MTD discussion mailing list
http://lists.infradead.org/mailman/listinfo/linux-mtd/

author
#7 lore raw

PATCH v2 4/5 mtd: rfd_ftl: fix use-after-free

In reply to Sean Young

del_mtd_blktrans_dev() will kfree part, so after this call both part and dev point to freed memory. Move the call to avoid use-after-free.

---
 drivers/mtd/rfd_ftl.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)
drivers/mtd/rfd_ftl.c +1 -1
index 7f5f6d247cae..52be9f1fa9a2 100644
@@ -800,10 +800,10 @@ static void rfd_ftl_remove_dev(struct mtd_blktrans_dev *dev)
 			part->mbd.mtd->name, i, part->blocks[i].erases);
 	}
 
-	del_mtd_blktrans_dev(dev);
 	vfree(part->sector_map);
 	kfree(part->header_cache);
 	kfree(part->blocks);
+	del_mtd_blktrans_dev(dev);
 }
 
 static struct mtd_blktrans_ops rfd_ftl_tr = {
Signature · 6 lines
2.31.1


______________________________________________________
Linux MTD discussion mailing list
http://lists.infradead.org/mailman/listinfo/linux-mtd/
maintainer Applied
#8 lore raw

Re: [PATCH v2 4/5] mtd: rfd_ftl: fix use-after-free

Quoted 4 lines · Sean Young

del_mtd_blktrans_dev() will kfree part, so after this call both part and dev point to freed memory. Move the call to avoid use-after-free.

Applied to https://git.kernel.org/pub/scm/linux/kernel/git/mtd/linux.git mtd/next, thanks.

Miquel

______________________________________________________
Linux MTD discussion mailing list
http://lists.infradead.org/mailman/listinfo/linux-mtd/

author
#9 lore raw

PATCH v2 5/5 mtd: rfd_ftl: use container_of() rather than cast

In reply to Sean Young

The container_of() is much more readable and also safer.

---
 drivers/mtd/rfd_ftl.c | 14 +++++++-------
 1 file changed, 7 insertions(+), 7 deletions(-)
drivers/mtd/rfd_ftl.c +7 -7
index 52be9f1fa9a2..af20a0a71108 100644
@@ -239,7 +239,7 @@ static int scan_header(struct partition *part)
 
 static int rfd_ftl_readsect(struct mtd_blktrans_dev *dev, u_long sector, char *buf)
 {
-	struct partition *part = (struct partition*)dev;
+	struct partition *part = container_of(dev, struct partition, mbd);
 	u_long addr;
 	size_t retlen;
 	int rc;
@@ -600,7 +600,7 @@ static int find_free_sector(const struct partition *part, const struct block *bl
 
 static int do_writesect(struct mtd_blktrans_dev *dev, u_long sector, char *buf, ulong *old_addr)
 {
-	struct partition *part = (struct partition*)dev;
+	struct partition *part = container_of(dev, struct partition, mbd);
 	struct block *block;
 	u_long addr;
 	int i;
@@ -666,7 +666,7 @@ static int do_writesect(struct mtd_blktrans_dev *dev, u_long sector, char *buf,
 
 static int rfd_ftl_writesect(struct mtd_blktrans_dev *dev, u_long sector, char *buf)
 {
-	struct partition *part = (struct partition*)dev;
+	struct partition *part = container_of(dev, struct partition, mbd);
 	u_long old_addr;
 	int i;
 	int rc = 0;
@@ -708,7 +708,7 @@ static int rfd_ftl_writesect(struct mtd_blktrans_dev *dev, u_long sector, char *
 static int rfd_ftl_discardsect(struct mtd_blktrans_dev *dev,
 			       unsigned long sector, unsigned int nr_sects)
 {
-	struct partition *part = (struct partition *)dev;
+	struct partition *part = container_of(dev, struct partition, mbd);
 	u_long addr;
 	int rc;
 
@@ -735,7 +735,7 @@ static int rfd_ftl_discardsect(struct mtd_blktrans_dev *dev,
 
 static int rfd_ftl_getgeo(struct mtd_blktrans_dev *dev, struct hd_geometry *geo)
 {
-	struct partition *part = (struct partition*)dev;
+	struct partition *part = container_of(dev, struct partition, mbd);
 
 	geo->heads = 1;
 	geo->sectors = SECTORS_PER_TRACK;
@@ -792,7 +792,7 @@ static void rfd_ftl_add_mtd(struct mtd_blktrans_ops *tr, struct mtd_info *mtd)
 
 static void rfd_ftl_remove_dev(struct mtd_blktrans_dev *dev)
 {
-	struct partition *part = (struct partition*)dev;
+	struct partition *part = container_of(dev, struct partition, mbd);
 	int i;
 
 	for (i=0; i<part->total_blocks; i++) {
@@ -803,7 +803,7 @@ static void rfd_ftl_remove_dev(struct mtd_blktrans_dev *dev)
 	vfree(part->sector_map);
 	kfree(part->header_cache);
 	kfree(part->blocks);
-	del_mtd_blktrans_dev(dev);
+	del_mtd_blktrans_dev(&part->mbd);
 }
 
 static struct mtd_blktrans_ops rfd_ftl_tr = {
Signature · 6 lines
2.31.1


______________________________________________________
Linux MTD discussion mailing list
http://lists.infradead.org/mailman/listinfo/linux-mtd/
maintainer Applied
#10 lore raw

Re: [PATCH v2 5/5] mtd: rfd_ftl: use container_of() rather than cast

Quoted 3 lines · Sean Young

The container_of() is much more readable and also safer.

Applied to https://git.kernel.org/pub/scm/linux/kernel/git/mtd/linux.git mtd/next, thanks.

Miquel

______________________________________________________
Linux MTD discussion mailing list
http://lists.infradead.org/mailman/listinfo/linux-mtd/