Message ID | 1459546164-6269-2-git-send-email-ezequiel@vanguardiasur.com.ar |
---|---|
State | Accepted |
Commit | 20c07a5bf094198ff2382aa5e7c930b3c9807792 |
Headers | show |
Am 01.04.2016 um 23:29 schrieb Ezequiel Garcia: > Since commit 807f16d4db95 ("mtd: core: set some defaults > when dev.parent is set"), it's now legal for drivers > to call nand_scan and nand_scan_ident without setting > mtd.owner. > > Drop the check and while at it remove the BUG() abuse. > > Signed-off-by: Ezequiel Garcia <ezequiel@vanguardiasur.com.ar> > --- > drivers/mtd/nand/nand_base.c | 10 +--------- > 1 file changed, 1 insertion(+), 9 deletions(-) > > diff --git a/drivers/mtd/nand/nand_base.c b/drivers/mtd/nand/nand_base.c > index c3733a10a6e7..befa04ef4a04 100644 > --- a/drivers/mtd/nand/nand_base.c > +++ b/drivers/mtd/nand/nand_base.c > @@ -4013,7 +4013,6 @@ static int nand_dt_init(struct nand_chip *chip) > * This is the first phase of the normal nand_scan() function. It reads the > * flash ID and sets up MTD fields accordingly. > * > - * The mtd->owner field must be set to the module of the caller. > */ > int nand_scan_ident(struct mtd_info *mtd, int maxchips, > struct nand_flash_dev *table) > @@ -4433,19 +4432,12 @@ EXPORT_SYMBOL(nand_scan_tail); > * > * This fills out all the uninitialized function pointers with the defaults. > * The flash ID is read and the mtd/chip structures are filled with the > - * appropriate values. The mtd->owner field must be set to the module of the > - * caller. > + * appropriate values. > */ > int nand_scan(struct mtd_info *mtd, int maxchips) > { > int ret; > > - /* Many callers got this wrong, so check for it for a while... */ > - if (!mtd->owner && caller_is_module()) { > - pr_crit("%s called with NULL mtd->owner!\n", __func__); > - BUG(); > - } Look okay to me. Do we have an in-kernel user which currently hits this condition? Thanks, //richard
On 1 April 2016 at 18:57, Richard Weinberger <richard@nod.at> wrote: > Am 01.04.2016 um 23:29 schrieb Ezequiel Garcia: >> Since commit 807f16d4db95 ("mtd: core: set some defaults >> when dev.parent is set"), it's now legal for drivers >> to call nand_scan and nand_scan_ident without setting >> mtd.owner. >> >> Drop the check and while at it remove the BUG() abuse. >> >> Signed-off-by: Ezequiel Garcia <ezequiel@vanguardiasur.com.ar> >> --- >> drivers/mtd/nand/nand_base.c | 10 +--------- >> 1 file changed, 1 insertion(+), 9 deletions(-) >> >> diff --git a/drivers/mtd/nand/nand_base.c b/drivers/mtd/nand/nand_base.c >> index c3733a10a6e7..befa04ef4a04 100644 >> --- a/drivers/mtd/nand/nand_base.c >> +++ b/drivers/mtd/nand/nand_base.c >> @@ -4013,7 +4013,6 @@ static int nand_dt_init(struct nand_chip *chip) >> * This is the first phase of the normal nand_scan() function. It reads the >> * flash ID and sets up MTD fields accordingly. >> * >> - * The mtd->owner field must be set to the module of the caller. >> */ >> int nand_scan_ident(struct mtd_info *mtd, int maxchips, >> struct nand_flash_dev *table) >> @@ -4433,19 +4432,12 @@ EXPORT_SYMBOL(nand_scan_tail); >> * >> * This fills out all the uninitialized function pointers with the defaults. >> * The flash ID is read and the mtd/chip structures are filled with the >> - * appropriate values. The mtd->owner field must be set to the module of the >> - * caller. >> + * appropriate values. >> */ >> int nand_scan(struct mtd_info *mtd, int maxchips) >> { >> int ret; >> >> - /* Many callers got this wrong, so check for it for a while... */ >> - if (!mtd->owner && caller_is_module()) { >> - pr_crit("%s called with NULL mtd->owner!\n", __func__); >> - BUG(); >> - } > > Look okay to me. > Do we have an in-kernel user which currently hits this condition? > I think that (after the recent mtd.owner swept) all the drivers that use nand_scan, will hit this condition if built as modules.
On Fri, Apr 01, 2016 at 06:29:23PM -0300, Ezequiel Garcia wrote: > Since commit 807f16d4db95 ("mtd: core: set some defaults > when dev.parent is set"), it's now legal for drivers > to call nand_scan and nand_scan_ident without setting > mtd.owner. > > Drop the check and while at it remove the BUG() abuse. > > Signed-off-by: Ezequiel Garcia <ezequiel@vanguardiasur.com.ar> > --- > drivers/mtd/nand/nand_base.c | 10 +--------- > 1 file changed, 1 insertion(+), 9 deletions(-) > > diff --git a/drivers/mtd/nand/nand_base.c b/drivers/mtd/nand/nand_base.c > index c3733a10a6e7..befa04ef4a04 100644 > --- a/drivers/mtd/nand/nand_base.c > +++ b/drivers/mtd/nand/nand_base.c > @@ -4013,7 +4013,6 @@ static int nand_dt_init(struct nand_chip *chip) > * This is the first phase of the normal nand_scan() function. It reads the > * flash ID and sets up MTD fields accordingly. > * > - * The mtd->owner field must be set to the module of the caller. > */ > int nand_scan_ident(struct mtd_info *mtd, int maxchips, > struct nand_flash_dev *table) > @@ -4433,19 +4432,12 @@ EXPORT_SYMBOL(nand_scan_tail); > * > * This fills out all the uninitialized function pointers with the defaults. > * The flash ID is read and the mtd/chip structures are filled with the > - * appropriate values. The mtd->owner field must be set to the module of the > - * caller. > + * appropriate values. > */ > int nand_scan(struct mtd_info *mtd, int maxchips) > { > int ret; > > - /* Many callers got this wrong, so check for it for a while... */ > - if (!mtd->owner && caller_is_module()) { > - pr_crit("%s called with NULL mtd->owner!\n", __func__); > - BUG(); > - } Ooh, yikes! Forgot this was there. I guess no one noticed, because fewer drivers are using plain nand_scan() these days (instead of splitting up nand_scan_ident() and nand_scan_tail()), and also, many NAND users don't run their drivers as modules. Anyway, this is probably worth -stable, right? (i.e., "Fixes: 807f16d4db95 ..." and "Cc: <stable@vger.kernel.org>") I can take this directly, or Boris, if you feel like there will be other for-v4.5 NAND material, you can queue this up instead. Brian > - > ret = nand_scan_ident(mtd, maxchips, NULL); > if (!ret) > ret = nand_scan_tail(mtd); > -- > 2.7.0 >
On Fri, 1 Apr 2016 15:26:49 -0700 Brian Norris <computersforpeace@gmail.com> wrote: > On Fri, Apr 01, 2016 at 06:29:23PM -0300, Ezequiel Garcia wrote: > > Since commit 807f16d4db95 ("mtd: core: set some defaults > > when dev.parent is set"), it's now legal for drivers > > to call nand_scan and nand_scan_ident without setting > > mtd.owner. > > > > Drop the check and while at it remove the BUG() abuse. > > > > Signed-off-by: Ezequiel Garcia <ezequiel@vanguardiasur.com.ar> Acked-by: Boris Brezillon <boris.brezillon@free-electrons.com> > > --- > > drivers/mtd/nand/nand_base.c | 10 +--------- > > 1 file changed, 1 insertion(+), 9 deletions(-) > > > > diff --git a/drivers/mtd/nand/nand_base.c b/drivers/mtd/nand/nand_base.c > > index c3733a10a6e7..befa04ef4a04 100644 > > --- a/drivers/mtd/nand/nand_base.c > > +++ b/drivers/mtd/nand/nand_base.c > > @@ -4013,7 +4013,6 @@ static int nand_dt_init(struct nand_chip *chip) > > * This is the first phase of the normal nand_scan() function. It reads the > > * flash ID and sets up MTD fields accordingly. > > * > > - * The mtd->owner field must be set to the module of the caller. > > */ > > int nand_scan_ident(struct mtd_info *mtd, int maxchips, > > struct nand_flash_dev *table) > > @@ -4433,19 +4432,12 @@ EXPORT_SYMBOL(nand_scan_tail); > > * > > * This fills out all the uninitialized function pointers with the defaults. > > * The flash ID is read and the mtd/chip structures are filled with the > > - * appropriate values. The mtd->owner field must be set to the module of the > > - * caller. > > + * appropriate values. > > */ > > int nand_scan(struct mtd_info *mtd, int maxchips) > > { > > int ret; > > > > - /* Many callers got this wrong, so check for it for a while... */ > > - if (!mtd->owner && caller_is_module()) { > > - pr_crit("%s called with NULL mtd->owner!\n", __func__); > > - BUG(); > > - } > > Ooh, yikes! Forgot this was there. I guess no one noticed, because fewer > drivers are using plain nand_scan() these days (instead of splitting up > nand_scan_ident() and nand_scan_tail()), and also, many NAND users don't > run their drivers as modules. > > Anyway, this is probably worth -stable, right? (i.e., "Fixes: > 807f16d4db95 ..." and "Cc: <stable@vger.kernel.org>") > > I can take this directly, or Boris, if you feel like there will be other > for-v4.5 NAND material, you can queue this up instead. No, take it directly. On a more general note, not sure creating a nand/fixes branch and sending you PRs after each -rc (if fixes are available of course) is really efficient. Maybe I'm wrong, but I don't expect to see more than a couple of fixes per release, and it's probably better if you keep taking them directly (with my acks). What do you think?
On Sat, Apr 02, 2016 at 03:52:16PM +0200, Boris Brezillon wrote: > On Fri, 1 Apr 2016 15:26:49 -0700 > Brian Norris <computersforpeace@gmail.com> wrote: > > On Fri, Apr 01, 2016 at 06:29:23PM -0300, Ezequiel Garcia wrote: > > > Since commit 807f16d4db95 ("mtd: core: set some defaults > > > when dev.parent is set"), it's now legal for drivers > > > to call nand_scan and nand_scan_ident without setting > > > mtd.owner. > > > > > > Drop the check and while at it remove the BUG() abuse. > > > > > > Signed-off-by: Ezequiel Garcia <ezequiel@vanguardiasur.com.ar> > > Acked-by: Boris Brezillon <boris.brezillon@free-electrons.com> > [...] > > > > Ooh, yikes! Forgot this was there. I guess no one noticed, because fewer > > drivers are using plain nand_scan() these days (instead of splitting up > > nand_scan_ident() and nand_scan_tail()), and also, many NAND users don't > > run their drivers as modules. > > > > Anyway, this is probably worth -stable, right? (i.e., "Fixes: > > 807f16d4db95 ..." and "Cc: <stable@vger.kernel.org>") > > > > I can take this directly, or Boris, if you feel like there will be other > > for-v4.5 NAND material, you can queue this up instead. > > No, take it directly. Applied patch 1 to linux-mtd.git. Thanks! > On a more general note, not sure creating a nand/fixes branch and > sending you PRs after each -rc (if fixes are available of course) is > really efficient. Maybe I'm wrong, but I don't expect to see more than a > couple of fixes per release, and it's probably better if you keep > taking them directly (with my acks). > > What do you think? That's probably fine. I'll still keep my eyes open for fixes like this, but ping me if you think I've missed anything. Brian
diff --git a/drivers/mtd/nand/nand_base.c b/drivers/mtd/nand/nand_base.c index c3733a10a6e7..befa04ef4a04 100644 --- a/drivers/mtd/nand/nand_base.c +++ b/drivers/mtd/nand/nand_base.c @@ -4013,7 +4013,6 @@ static int nand_dt_init(struct nand_chip *chip) * This is the first phase of the normal nand_scan() function. It reads the * flash ID and sets up MTD fields accordingly. * - * The mtd->owner field must be set to the module of the caller. */ int nand_scan_ident(struct mtd_info *mtd, int maxchips, struct nand_flash_dev *table) @@ -4433,19 +4432,12 @@ EXPORT_SYMBOL(nand_scan_tail); * * This fills out all the uninitialized function pointers with the defaults. * The flash ID is read and the mtd/chip structures are filled with the - * appropriate values. The mtd->owner field must be set to the module of the - * caller. + * appropriate values. */ int nand_scan(struct mtd_info *mtd, int maxchips) { int ret; - /* Many callers got this wrong, so check for it for a while... */ - if (!mtd->owner && caller_is_module()) { - pr_crit("%s called with NULL mtd->owner!\n", __func__); - BUG(); - } - ret = nand_scan_ident(mtd, maxchips, NULL); if (!ret) ret = nand_scan_tail(mtd);
Since commit 807f16d4db95 ("mtd: core: set some defaults when dev.parent is set"), it's now legal for drivers to call nand_scan and nand_scan_ident without setting mtd.owner. Drop the check and while at it remove the BUG() abuse. Signed-off-by: Ezequiel Garcia <ezequiel@vanguardiasur.com.ar> --- drivers/mtd/nand/nand_base.c | 10 +--------- 1 file changed, 1 insertion(+), 9 deletions(-)