Groups | Search | Server Info | Keyboard shortcuts | Login | Register [http] [https] [nntp] [nntps]

Groups > linux.kernel > #1656349 > unrolled thread

[PATCH 0/5] Extends block2mtd and ubi drivers

Back to article view | Back to linux.kernel

Contents

  [PATCH 0/5] Extends block2mtd and ubi drivers Pali Rohár <pali.rohar@gmail.com> - 2017-06-02 17:50 +0200
    [PATCH 1/5] mtd: block2mtd: Check for valid user supplied erase size Pali Rohár <pali.rohar@gmail.com> - 2017-06-02 17:50 +0200
    [PATCH 2/5] mtd: block2mtd: Add support for specifying MTD write size and subpage shift Pali Rohár <pali.rohar@gmail.com> - 2017-06-02 17:50 +0200
      Re: [PATCH 2/5] mtd: block2mtd: Add support for specifying MTD write  size and subpage shift Richard Weinberger <richard@nod.at> - 2017-06-02 18:20 +0200
        Re: [PATCH 2/5] mtd: block2mtd: Add support for specifying MTD write  size and subpage shift Pali Rohár <pali.rohar@gmail.com> - 2017-06-05 13:30 +0200
          Re: [PATCH 2/5] mtd: block2mtd: Add support for specifying MTD write  size and subpage shift Pali Rohár <pali.rohar@gmail.com> - 2017-06-05 13:30 +0200
            Re: [PATCH 2/5] mtd: block2mtd: Add support for specifying MTD write  size and subpage shift Richard Weinberger <richard@nod.at> - 2017-06-05 13:30 +0200
              Re: [PATCH 2/5] mtd: block2mtd: Add support for specifying MTD write  size and subpage shift Pali Rohár <pali.rohar@gmail.com> - 2017-06-07 10:50 +0200
          Re: [PATCH 2/5] mtd: block2mtd: Add support for specifying MTD write  size and subpage shift Richard Weinberger <richard@nod.at> - 2017-06-05 13:30 +0200
    [PATCH 4/5] mtd: block2mtd: Add support for deleting block2mtd mapping Pali Rohár <pali.rohar@gmail.com> - 2017-06-02 17:50 +0200
    Re: [PATCH 0/5] Extends block2mtd and ubi drivers Richard Weinberger <richard@nod.at> - 2017-06-02 18:20 +0200
      Re: [PATCH 0/5] Extends block2mtd and ubi drivers Pali Rohár <pali.rohar@gmail.com> - 2017-06-05 13:20 +0200

#1656349 — [PATCH 0/5] Extends block2mtd and ubi drivers

This patch series extends block2mtd and ubi drivers to better handle
read-only devices and allow to load UBI image from local file which was
created for nand device.

Tested for Nokia N900 with Maemo 5 rootfs ubifs image
(rootfs_RX-51_2009SE_21.2011.38-1_PR_MR0) which has erase size 128k,
write size 2k and nand subpage shift 2.

$ losetup -r /dev/loop0 rootfs_RX-51_2009SE_21.2011.38-1_PR_MR0.ubifs
$ echo -n /dev/loop0,131072,2048,2 > /sys/module/block2mtd/parameters/block2mtd
$ ubiattach -p /dev/mtd0
$ mount /dev/ubi0_0 /mnt/ubi -t ubifs
...
$ umount /dev/ubi0_0
$ ubidetach -p /dev/mtd0
$ echo -n del=/dev/loop0 > /sys/module/block2mtd/parameters/block2mtd
$ losetup -d /dev/loop0

Pali Rohár (5):
  mtd: block2mtd: Check for valid user supplied erase size
  mtd: block2mtd: Add support for specifying MTD write size and subpage
    shift
  mtd: block2mtd: Fallback to read-only mode
  mtd: block2mtd: Add support for deleting block2mtd mapping
  ubi: Allow to use read-only UBI volume with not enough PEBs

 drivers/mtd/devices/block2mtd.c |  129 ++++++++++++++++++++++++++++++---------
 drivers/mtd/ubi/vtbl.c          |   14 +++--
 2 files changed, 110 insertions(+), 33 deletions(-)

-- 
1.7.9.5

[toc] | [next] | [standalone]

#1656350 — [PATCH 1/5] mtd: block2mtd: Check for valid user supplied erase size

Erase size is limited to 32bit unsigned integer, but value parsed from user
is limited up to size_t C type.

Signed-off-by: Pali Rohár <pali.rohar@gmail.com>
---
 drivers/mtd/devices/block2mtd.c |    2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/drivers/mtd/devices/block2mtd.c b/drivers/mtd/devices/block2mtd.c
index 7c887f1..ee47cdd 100644
--- a/drivers/mtd/devices/block2mtd.c
+++ b/drivers/mtd/devices/block2mtd.c
@@ -419,7 +419,7 @@ static int block2mtd_setup2(const char *val)
 
 	if (token[1]) {
 		ret = parse_num(&erase_size, token[1]);
-		if (ret) {
+		if (ret || erase_size > U32_MAX) {
 			pr_err("illegal erase size\n");
 			return 0;
 		}
-- 
1.7.9.5

[toc] | [prev] | [next] | [standalone]

#1656352 — [PATCH 2/5] mtd: block2mtd: Add support for specifying MTD write size and subpage shift

It is needed for creating emulated devices suitable for using in UBI layer
and with UBIFS.

Signed-off-by: Pali Rohár <pali.rohar@gmail.com>
---
 drivers/mtd/devices/block2mtd.c |   55 ++++++++++++++++++++++++++++++---------
 1 file changed, 42 insertions(+), 13 deletions(-)

diff --git a/drivers/mtd/devices/block2mtd.c b/drivers/mtd/devices/block2mtd.c
index ee47cdd..5ba5fad 100644
--- a/drivers/mtd/devices/block2mtd.c
+++ b/drivers/mtd/devices/block2mtd.c
@@ -3,6 +3,7 @@
  *
  * Copyright (C) 2001,2002	Simon Evans <spse@secret.org.uk>
  * Copyright (C) 2004-2006	Joern Engel <joern@wh.fh-wedel.de>
+ * Copyright (C) 2012-2017	Pali Rohár <pali.rohar@gmail.com>
  *
  * Licence: GPL
  */
@@ -218,8 +219,8 @@ static void block2mtd_free_device(struct block2mtd_dev *dev)
 }
 
 
-static struct block2mtd_dev *add_device(char *devname, int erase_size,
-		int timeout)
+static struct block2mtd_dev *add_device(char *devname, uint32_t erase_size,
+		uint32_t write_size, int subpage_sft, int timeout)
 {
 #ifndef MODULE
 	int i;
@@ -279,6 +280,11 @@ static struct block2mtd_dev *add_device(char *devname, int erase_size,
 		goto err_free_block2mtd;
 	}
 
+	if ((long)dev->blkdev->bd_inode->i_size % write_size) {
+		pr_err("writesize must be divisor of device size\n");
+		goto err_free_block2mtd;
+	}
+
 	mutex_init(&dev->write_mutex);
 
 	/* Setup the MTD structure */
@@ -291,7 +297,8 @@ static struct block2mtd_dev *add_device(char *devname, int erase_size,
 
 	dev->mtd.size = dev->blkdev->bd_inode->i_size & PAGE_MASK;
 	dev->mtd.erasesize = erase_size;
-	dev->mtd.writesize = 1;
+	dev->mtd.writesize = write_size;
+	dev->mtd.subpage_sft = subpage_sft;
 	dev->mtd.writebufsize = PAGE_SIZE;
 	dev->mtd.type = MTD_RAM;
 	dev->mtd.flags = MTD_CAP_RAM;
@@ -308,10 +315,12 @@ static struct block2mtd_dev *add_device(char *devname, int erase_size,
 	}
 
 	list_add(&dev->list, &blkmtd_device_list);
-	pr_info("mtd%d: [%s] erase_size = %dKiB [%d]\n",
+	pr_info("mtd%d: [%s] erase_size = %dKiB [%d], write_size = %dKiB [%d], subpage_sft = %d\n",
 		dev->mtd.index,
 		dev->mtd.name + strlen("block2mtd: "),
-		dev->mtd.erasesize >> 10, dev->mtd.erasesize);
+		dev->mtd.erasesize >> 10, dev->mtd.erasesize,
+		dev->mtd.writesize >> 10, dev->mtd.writesize,
+		dev->mtd.subpage_sft);
 	return dev;
 
 err_destroy_mutex:
@@ -375,18 +384,20 @@ static inline void kill_final_newline(char *str)
 
 #ifndef MODULE
 static int block2mtd_init_called = 0;
-/* 80 for device, 12 for erase size */
-static char block2mtd_paramline[80 + 12];
+/* 80 for device, 12 for erase size, 12 for write size, 3 for subpage_sft */
+static char block2mtd_paramline[80 + 12 + 12 + 3];
 #endif
 
 static int block2mtd_setup2(const char *val)
 {
-	/* 80 for device, 12 for erase size, 80 for name, 8 for timeout */
-	char buf[80 + 12 + 80 + 8];
+	/* 80 for name, 80 for device, 12 for erase size, 12 for write size, 3 for subpage_sft */
+	char buf[80 + 12 + 80 + 12 + 3];
 	char *str = buf;
-	char *token[2];
+	char *token[4];
 	char *name;
 	size_t erase_size = PAGE_SIZE;
+	size_t write_size = 1;
+	size_t subpage_sft = 0;
 	unsigned long timeout = MTD_DEFAULT_TIMEOUT;
 	int i, ret;
 
@@ -398,7 +409,7 @@ static int block2mtd_setup2(const char *val)
 	strcpy(str, val);
 	kill_final_newline(str);
 
-	for (i = 0; i < 2; i++)
+	for (i = 0; i < 4; i++)
 		token[i] = strsep(&str, ",");
 
 	if (str) {
@@ -425,7 +436,23 @@ static int block2mtd_setup2(const char *val)
 		}
 	}
 
-	add_device(name, erase_size, timeout);
+	if (token[2]) {
+		ret = parse_num(&write_size, token[2]);
+		if (ret || write_size > U32_MAX) {
+			pr_err("illegal write size");
+			return 0;
+		}
+	}
+
+	if (token[3]) {
+		ret = parse_num(&subpage_sft, token[3]);
+		if (ret || subpage_sft > INT_MAX) {
+			pr_err("illegal subpage_sft");
+			return 0;
+		}
+	}
+
+	add_device(name, erase_size, write_size, subpage_sft, timeout);
 
 	return 0;
 }
@@ -459,7 +486,9 @@ static int block2mtd_setup(const char *val, struct kernel_param *kp)
 
 
 module_param_call(block2mtd, block2mtd_setup, NULL, NULL, 0200);
-MODULE_PARM_DESC(block2mtd, "Device to use. \"block2mtd=<dev>[,<erasesize>]\"");
+MODULE_PARM_DESC(block2mtd, "Block device, erase size (default page size), "
+	"write size (default 1), nand subpage shift (default 0) "
+	"\"block2mtd=<dev>[,<erasesize>[,<writesize>[,<subpage_sft>]]]\"");
 
 static int __init block2mtd_init(void)
 {
-- 
1.7.9.5

[toc] | [prev] | [next] | [standalone]

#1656376 — Re: [PATCH 2/5] mtd: block2mtd: Add support for specifying MTD write size and subpage shift

Pali,

Am 02.06.2017 um 17:43 schrieb Pali Rohár:
> It is needed for creating emulated devices suitable for using in UBI layer
> and with UBIFS.

Why?

It is not clear to me why kind of MTD you are constructing.
subpages are NAND specific you create RAM and ROM MTDs.
Sure, the MTD allows such constructs but I'm not sure whether this is a good idea.

Thanks,
//richard

[toc] | [prev] | [next] | [standalone]

#1657524 — Re: [PATCH 2/5] mtd: block2mtd: Add support for specifying MTD write size and subpage shift

On Friday 02 June 2017 18:13:02 Richard Weinberger wrote:
> Pali,
> 
> Am 02.06.2017 um 17:43 schrieb Pali Rohár:
> > It is needed for creating emulated devices suitable for using in UBI layer
> > and with UBIFS.
> 
> Why?

ubifs depends on write size of nand. And without those parameters as
specified in cover letter I'm unable to mount N900 rootfs image exported
via block2mtd. ubifs reject such image.

-- 
Pali Rohár
pali.rohar@gmail.com

[toc] | [prev] | [next] | [standalone]

#1657525 — Re: [PATCH 2/5] mtd: block2mtd: Add support for specifying MTD write size and subpage shift

On Monday 05 June 2017 13:23:22 Richard Weinberger wrote:
> Pali,
> 
> Am 05.06.2017 um 13:21 schrieb Pali Rohár:
> > On Friday 02 June 2017 18:13:02 Richard Weinberger wrote:
> >> Pali,
> >>
> >> Am 02.06.2017 um 17:43 schrieb Pali Rohár:
> >>> It is needed for creating emulated devices suitable for using in UBI layer
> >>> and with UBIFS.
> >>
> >> Why?
> > 
> > ubifs depends on write size of nand. And without those parameters as
> > specified in cover letter I'm unable to mount N900 rootfs image exported
> > via block2mtd. ubifs reject such image.
> 
> Hmm, so you render block2mtd into a semi-NAND chip? :)

Probably you can call it like that. But it is still MTD device...

-- 
Pali Rohár
pali.rohar@gmail.com

[toc] | [prev] | [next] | [standalone]

#1657526 — Re: [PATCH 2/5] mtd: block2mtd: Add support for specifying MTD write size and subpage shift

Pali,

Am 05.06.2017 um 13:25 schrieb Pali Rohár:
> On Monday 05 June 2017 13:23:22 Richard Weinberger wrote:
>> Pali,
>>
>> Am 05.06.2017 um 13:21 schrieb Pali Rohár:
>>> On Friday 02 June 2017 18:13:02 Richard Weinberger wrote:
>>>> Pali,
>>>>
>>>> Am 02.06.2017 um 17:43 schrieb Pali Rohár:
>>>>> It is needed for creating emulated devices suitable for using in UBI layer
>>>>> and with UBIFS.
>>>>
>>>> Why?
>>>
>>> ubifs depends on write size of nand. And without those parameters as
>>> specified in cover letter I'm unable to mount N900 rootfs image exported
>>> via block2mtd. ubifs reject such image.
>>
>> Hmm, so you render block2mtd into a semi-NAND chip? :)
> 
> Probably you can call it like that. But it is still MTD device...

This is what I meant in my other mail.
You add NAND specific properties but still denote it as MTD_RAM/ROM.
I'm not sure whether this is a good idea.

Thanks,
//richard

[toc] | [prev] | [next] | [standalone]

#1659532 — Re: [PATCH 2/5] mtd: block2mtd: Add support for specifying MTD write size and subpage shift

On Monday 05 June 2017 13:27:18 Richard Weinberger wrote:
> Pali,
> 
> Am 05.06.2017 um 13:25 schrieb Pali Rohár:
> > On Monday 05 June 2017 13:23:22 Richard Weinberger wrote:
> >> Pali,
> >>
> >> Am 05.06.2017 um 13:21 schrieb Pali Rohár:
> >>> On Friday 02 June 2017 18:13:02 Richard Weinberger wrote:
> >>>> Pali,
> >>>>
> >>>> Am 02.06.2017 um 17:43 schrieb Pali Rohár:
> >>>>> It is needed for creating emulated devices suitable for using in UBI layer
> >>>>> and with UBIFS.
> >>>>
> >>>> Why?
> >>>
> >>> ubifs depends on write size of nand. And without those parameters as
> >>> specified in cover letter I'm unable to mount N900 rootfs image exported
> >>> via block2mtd. ubifs reject such image.
> >>
> >> Hmm, so you render block2mtd into a semi-NAND chip? :)
> > 
> > Probably you can call it like that. But it is still MTD device...
> 
> This is what I meant in my other mail.
> You add NAND specific properties but still denote it as MTD_RAM/ROM.
> I'm not sure whether this is a good idea.

Ok, lets wait what other people think.

At least patches like fallback or check should be less problematic and
could be applied separately.

-- 
Pali Rohár
pali.rohar@gmail.com

[toc] | [prev] | [next] | [standalone]

#1657527 — Re: [PATCH 2/5] mtd: block2mtd: Add support for specifying MTD write size and subpage shift

Pali,

Am 05.06.2017 um 13:21 schrieb Pali Rohár:
> On Friday 02 June 2017 18:13:02 Richard Weinberger wrote:
>> Pali,
>>
>> Am 02.06.2017 um 17:43 schrieb Pali Rohár:
>>> It is needed for creating emulated devices suitable for using in UBI layer
>>> and with UBIFS.
>>
>> Why?
> 
> ubifs depends on write size of nand. And without those parameters as
> specified in cover letter I'm unable to mount N900 rootfs image exported
> via block2mtd. ubifs reject such image.

Hmm, so you render block2mtd into a semi-NAND chip? :)

Thanks,
//richard

[toc] | [prev] | [next] | [standalone]

#1656353 — [PATCH 4/5] mtd: block2mtd: Add support for deleting block2mtd mapping

This patch allows user to delete block2mtd mapping via parameters file
/sys/module/block2mtd/parameters/block2mtd

Syntax is "del=device", e.g.:

$ echo -n del=/dev/loop0 > /sys/module/block2mtd/parameters/block2mtd

Signed-off-by: Pali Rohár <pali.rohar@gmail.com>
---
 drivers/mtd/devices/block2mtd.c |   35 +++++++++++++++++++++++++++--------
 1 file changed, 27 insertions(+), 8 deletions(-)

diff --git a/drivers/mtd/devices/block2mtd.c b/drivers/mtd/devices/block2mtd.c
index ad96937..9b6f7b5 100644
--- a/drivers/mtd/devices/block2mtd.c
+++ b/drivers/mtd/devices/block2mtd.c
@@ -355,6 +355,19 @@ static struct block2mtd_dev *add_device(char *devname, uint32_t erase_size,
 	return NULL;
 }
 
+static void del_device(struct block2mtd_dev *dev)
+{
+	if (!dev->ro_mode)
+		block2mtd_sync(&dev->mtd);
+	mtd_device_unregister(&dev->mtd);
+	mutex_destroy(&dev->write_mutex);
+	pr_info("mtd%d: [%s] removed\n",
+		dev->mtd.index,
+		dev->mtd.name + strlen("block2mtd: "));
+	list_del(&dev->list);
+	block2mtd_free_device(dev);
+}
+
 
 /* This function works similar to reguler strtoul.  In addition, it
  * allows some suffixes for a more human-readable number format:
@@ -477,6 +490,19 @@ static int block2mtd_setup2(const char *val)
 		}
 	}
 
+	if (strncmp(name, "del=", strlen("del=")) == 0) {
+		struct list_head *pos, *next;
+		list_for_each_safe(pos, next, &blkmtd_device_list) {
+			struct block2mtd_dev *dev =
+				list_entry(pos, typeof(*dev), list);
+			if (strcmp(dev->mtd.name + strlen("block2mtd: "),
+				name + strlen("del=")) != 0)
+				continue;
+			del_device(dev);
+			return 0;
+		}
+	}
+
 	add_device(name, erase_size, write_size, subpage_sft, timeout);
 
 	return 0;
@@ -536,14 +562,7 @@ static void block2mtd_exit(void)
 	/* Remove the MTD devices */
 	list_for_each_safe(pos, next, &blkmtd_device_list) {
 		struct block2mtd_dev *dev = list_entry(pos, typeof(*dev), list);
-		block2mtd_sync(&dev->mtd);
-		mtd_device_unregister(&dev->mtd);
-		mutex_destroy(&dev->write_mutex);
-		pr_info("mtd%d: [%s] removed\n",
-			dev->mtd.index,
-			dev->mtd.name + strlen("block2mtd: "));
-		list_del(&dev->list);
-		block2mtd_free_device(dev);
+		del_device(dev);
 	}
 }
 
-- 
1.7.9.5

[toc] | [prev] | [next] | [standalone]

#1656378

Pali,

Am 02.06.2017 um 17:43 schrieb Pali Rohár:
> This patch series extends block2mtd and ubi drivers to better handle
> read-only devices and allow to load UBI image from local file which was
> created for nand device.
> 
> Tested for Nokia N900 with Maemo 5 rootfs ubifs image
> (rootfs_RX-51_2009SE_21.2011.38-1_PR_MR0) which has erase size 128k,
> write size 2k and nand subpage shift 2.

What is the use case behind this series?

Did you see my nandsim rework some time ago?
http://lists.infradead.org/pipermail/linux-mtd/2016-September/069422.html
If you need a way to load files/nanddumps as NAND devices, this should be a good
starting point.
This reminds me that I need to revive that series. :-)

> $ losetup -r /dev/loop0 rootfs_RX-51_2009SE_21.2011.38-1_PR_MR0.ubifs
> $ echo -n /dev/loop0,131072,2048,2 > /sys/module/block2mtd/parameters/block2mtd
> $ ubiattach -p /dev/mtd0
> $ mount /dev/ubi0_0 /mnt/ubi -t ubifs
> ...
> $ umount /dev/ubi0_0
> $ ubidetach -p /dev/mtd0
> $ echo -n del=/dev/loop0 > /sys/module/block2mtd/parameters/block2mtd
> $ losetup -d /dev/loop0

The module-parameter interface is odd. IMHO we should not extend it.

Thanks,
//richard

[toc] | [prev] | [next] | [standalone]

#1657523

On Friday 02 June 2017 18:17:06 Richard Weinberger wrote:
> Pali,
> 
> Am 02.06.2017 um 17:43 schrieb Pali Rohár:
> > This patch series extends block2mtd and ubi drivers to better handle
> > read-only devices and allow to load UBI image from local file which was
> > created for nand device.
> > 
> > Tested for Nokia N900 with Maemo 5 rootfs ubifs image
> > (rootfs_RX-51_2009SE_21.2011.38-1_PR_MR0) which has erase size 128k,
> > write size 2k and nand subpage shift 2.
> 
> What is the use case behind this series?

Take existing ubi image (where is one ubifs volume) and unpack it. Or
rather unpack-modify-pack to do some small changes.

> Did you see my nandsim rework some time ago?
> http://lists.infradead.org/pipermail/linux-mtd/2016-September/069422.html

No yet. I have my patches since 2012, but I decided after cleaning up
them to send... As I think they could be useful for other people.

> If you need a way to load files/nanddumps as NAND devices, this should be a good
> starting point.
> This reminds me that I need to revive that series. :-)

nandsim.ko has problem that needs to be loaded with special parameters
compatible with characteristic of target nand for which is ubi image
prepared. Plus it is nand similator and not layer to translate arbitrary
file image on disk to mtd device. block2mtd is what is doing this part.

> > $ losetup -r /dev/loop0 rootfs_RX-51_2009SE_21.2011.38-1_PR_MR0.ubifs
> > $ echo -n /dev/loop0,131072,2048,2 > /sys/module/block2mtd/parameters/block2mtd
> > $ ubiattach -p /dev/mtd0
> > $ mount /dev/ubi0_0 /mnt/ubi -t ubifs
> > ...
> > $ umount /dev/ubi0_0
> > $ ubidetach -p /dev/mtd0
> > $ echo -n del=/dev/loop0 > /sys/module/block2mtd/parameters/block2mtd
> > $ losetup -d /dev/loop0
> 
> The module-parameter interface is odd. IMHO we should not extend it.

That file is used for adding new mapping from block device to mtd
device. Currently there is no other way how to specify that mapping or
removing mapping.

If you have better idea, let me know and I would try to implement it.

-- 
Pali Rohár
pali.rohar@gmail.com

[toc] | [prev] | [standalone]

Back to top | Article view | linux.kernel

csiph-web