| Message ID | 20140123205331.03314c6f82f24a6bb4b8ef53@skynet.be |
|---|---|
| State | Superseded |
| Headers | show |
Hi Fabian, I saw you sent a v2 for one of the patches on this series, but maybe this is worth considering too. On Thu, Jan 23, 2014 at 08:53:31PM +0100, Fabian Frederick wrote: [..] > -/* FIXME: ensure that mtd->size % erase_size == 0 */ > static struct block2mtd_dev *add_device(char *devname, int erase_size) > { > const fmode_t mode = FMODE_READ | FMODE_WRITE | FMODE_EXCL; > @@ -250,6 +249,11 @@ static struct block2mtd_dev *add_device(char *devname, int erase_size) > goto devinit_err1; > } > > + if ((long)dev->blkdev->bd_inode->i_size % erase_size) { > + pr_err("erasesize muse be a divisor of device size\n"); > + goto devinit_err1; > + } > + Brian: What do you think? Fabian: Have you tested this patch? Can you elaborate a bit more about the effect it would have, compared to the current behavior?
On Wed, 29 Jan 2014 07:04:25 -0300 Ezequiel Garcia <ezequiel.garcia@free-electrons.com> wrote: > Hi Fabian, > > I saw you sent a v2 for one of the patches on this series, > but maybe this is worth considering too. > > On Thu, Jan 23, 2014 at 08:53:31PM +0100, Fabian Frederick wrote: > [..] > > -/* FIXME: ensure that mtd->size % erase_size == 0 */ > > static struct block2mtd_dev *add_device(char *devname, int erase_size) > > { > > const fmode_t mode = FMODE_READ | FMODE_WRITE | FMODE_EXCL; > > @@ -250,6 +249,11 @@ static struct block2mtd_dev *add_device(char *devname, int erase_size) > > goto devinit_err1; > > } > > > > + if ((long)dev->blkdev->bd_inode->i_size % erase_size) { > > + pr_err("erasesize muse be a divisor of device size\n"); > > + goto devinit_err1; > > + } > > + > > Brian: What do you think? > > Fabian: Have you tested this patch? Can you elaborate a bit more about > the effect it would have, compared to the current behavior? Hi Ezequiel, This patch was tested with the following commands : rmmod block2mtd;modprobe block2mtd block2mtd=/dev/loop0,<erasesize>;dmesg with both correct and incorrect values.It tries to address the fixme comment above the function. (/* FIXME: ensure that mtd->size % erase_size == 0 */) From what I understand, global size must be a multiple of erase_size when it comes to any MTD I/O operations. If Brian finds this one interesting I can repost it with current error names against l2-mtd.git/next AFAICS, current behavior let any value work... PS : This would need further testing (eg boot command). Regards, Fabian > -- > Ezequiel García, Free Electrons > Embedded Linux, Kernel and Android Engineering > http://free-electrons.com
Hi, On Wed, Jan 29, 2014 at 08:51:05PM +0800, Fabian Frederick wrote: > On Wed, 29 Jan 2014 07:04:25 -0300 Ezequiel Garcia <ezequiel.garcia@free-electrons.com> wrote: > > I saw you sent a v2 for one of the patches on this series, > > but maybe this is worth considering too. > > > > On Thu, Jan 23, 2014 at 08:53:31PM +0100, Fabian Frederick wrote: > > [..] > > > -/* FIXME: ensure that mtd->size % erase_size == 0 */ > > > static struct block2mtd_dev *add_device(char *devname, int erase_size) > > > { > > > const fmode_t mode = FMODE_READ | FMODE_WRITE | FMODE_EXCL; > > > @@ -250,6 +249,11 @@ static struct block2mtd_dev *add_device(char *devname, int erase_size) > > > goto devinit_err1; > > > } > > > > > > + if ((long)dev->blkdev->bd_inode->i_size % erase_size) { > > > + pr_err("erasesize muse be a divisor of device size\n"); > > > + goto devinit_err1; > > > + } > > > + > > > > Brian: What do you think? > > > > Fabian: Have you tested this patch? Can you elaborate a bit more about > > the effect it would have, compared to the current behavior? > > Hi Ezequiel, > > This patch was tested with the following commands : > > rmmod block2mtd;modprobe block2mtd block2mtd=/dev/loop0,<erasesize>;dmesg > > with both correct and incorrect values.It tries to address the fixme > comment above the function. > (/* FIXME: ensure that mtd->size % erase_size == 0 */) > > From what I understand, global size must be a multiple of > erase_size when it comes to any MTD I/O operations. If Brian finds this one > interesting I can repost it with current error names against l2-mtd.git/next I'm not sure block2mtd would have many users, but the hunk above looks reasonable. Feel free to send a patch. > AFAICS, current behavior let any value work... I bet this doesn't work out too well in the end. Maybe I'll give this a whirl just to see. > PS : This would need further testing (eg boot command). Brian
On Thu, 6 Mar 2014 00:54:22 -0800 Brian Norris <computersforpeace@gmail.com> wrote: > Hi, > > On Wed, Jan 29, 2014 at 08:51:05PM +0800, Fabian Frederick wrote: > > On Wed, 29 Jan 2014 07:04:25 -0300 Ezequiel Garcia <ezequiel.garcia@free-electrons.com> wrote: > > > I saw you sent a v2 for one of the patches on this series, > > > but maybe this is worth considering too. > > > > > > On Thu, Jan 23, 2014 at 08:53:31PM +0100, Fabian Frederick wrote: > > > [..] > > > > -/* FIXME: ensure that mtd->size % erase_size == 0 */ > > > > static struct block2mtd_dev *add_device(char *devname, int erase_size) > > > > { > > > > const fmode_t mode = FMODE_READ | FMODE_WRITE | FMODE_EXCL; > > > > @@ -250,6 +249,11 @@ static struct block2mtd_dev *add_device(char *devname, int erase_size) > > > > goto devinit_err1; > > > > } > > > > > > > > + if ((long)dev->blkdev->bd_inode->i_size % erase_size) { > > > > + pr_err("erasesize muse be a divisor of device size\n"); > > > > + goto devinit_err1; > > > > + } > > > > + > > > > > > Brian: What do you think? > > > > > > Fabian: Have you tested this patch? Can you elaborate a bit more about > > > the effect it would have, compared to the current behavior? > > > > Hi Ezequiel, > > > > This patch was tested with the following commands : > > > > rmmod block2mtd;modprobe block2mtd block2mtd=/dev/loop0,<erasesize>;dmesg > > > > with both correct and incorrect values.It tries to address the fixme > > comment above the function. > > (/* FIXME: ensure that mtd->size % erase_size == 0 */) > > > > From what I understand, global size must be a multiple of > > erase_size when it comes to any MTD I/O operations. If Brian finds this one > > interesting I can repost it with current error names against l2-mtd.git/next > > I'm not sure block2mtd would have many users, but the hunk above looks > reasonable. Feel free to send a patch. Hi Brian, I just send a patch v2 based on linux-next to avoid conflicts with my previous commit about mutex management in the same function and still staging there. Fabian > > > AFAICS, current behavior let any value work... > > I bet this doesn't work out too well in the end. Maybe I'll give this a > whirl just to see. > > > PS : This would need further testing (eg boot command). > > Brian
On Mar 06, Brian Norris wrote: [..] > > I'm not sure block2mtd would have many users, but the hunk above looks > reasonable. Feel free to send a patch. > I've bounced you the original patch from Fabian.
diff --git a/drivers/mtd/devices/block2mtd.c b/drivers/mtd/devices/block2mtd.c index 8071596..f0fd4fc 100644 --- a/drivers/mtd/devices/block2mtd.c +++ b/drivers/mtd/devices/block2mtd.c @@ -209,7 +209,6 @@ static void block2mtd_free_device(struct block2mtd_dev *dev) } -/* FIXME: ensure that mtd->size % erase_size == 0 */ static struct block2mtd_dev *add_device(char *devname, int erase_size) { const fmode_t mode = FMODE_READ | FMODE_WRITE | FMODE_EXCL; @@ -250,6 +249,11 @@ static struct block2mtd_dev *add_device(char *devname, int erase_size) goto devinit_err1; } + if ((long)dev->blkdev->bd_inode->i_size % erase_size) { + pr_err("erasesize muse be a divisor of device size\n"); + goto devinit_err1; + } + mutex_init(&dev->write_mutex); /* Setup the MTD structure */
fixme applied : check device size is a multiple of erasesize. Signed-off-by: Fabian Frederick <fabf@skynet.be> --- drivers/mtd/devices/block2mtd.c | 6 +++++- 1 file changed, 5 insertions(+), 1 deletion(-)