mbox series

[v5,0/3] Add nvmem support for dynamic partitions

Message ID 20220606151417.19227-1-ansuelsmth@gmail.com
Headers show
Series Add nvmem support for dynamic partitions | expand

Message

Christian Marangi June 6, 2022, 3:14 p.m. UTC
This very small series comes to fix the very annyoing problem of
partitions declared by parser at runtime NOT supporting nvmem cells
definition.

The current implementation is very generic. The idea is to provide an of
node if defined for everyone and not strictly limit this to nvmem stuff.
But still the actual change is done only for nvmem-cells mtd. (just to
make sure) This can totally change by removing the compatible check.

The idea here is that a user can still use these dynamic parsers
instead of declaring a fixed-partition and also declare how nvmem-cells
are defined for the partition.
This live with the assumption that dynamic partition have always the
same name and they are known. (this is the case for smem-part partition
that would require a bootloader reflash to change and for parsers like
cmdlinepart where the name is always the same.)
With this assumption, it's easy to fix this problem. Just introduce a
new partition node that will declare just these special partition.
Mtdcore then will check if these special declaration are present and
connect the dynamic partition with the OF node present in the dts. Nvmem
will automagically find the OF node and cells will be works based on the
data provided by the parser.

The initial idea was to create a special nvmem driver with a special
compatible where a user would declare the mtd partition name and this
driver would search it and register the nvmem cells but that became
difficult really fast, mtd notifier system is problematic for this kind
of stuff. So here is the better implementation. A variant of this is
already tested on openwrt where we have devices that use cmdlinepart.
(that current variant have defined in the dts the exact copy of
cmdlinepart in the fixed-partition scheme and we patched the cmdlinepart
parser to scan this fixed-partition node (that is ignored as cmdlinepart
have priority) and connect the dynamic partition with the dts node)

I provided an example of this in the documentation commit.
In short it's needed to add to the partitions where the compatible parser
is declared, a partition with just the label declared (instead of the reg).
Then declare some nvmem-cells and it will all work at runtime.
Mtdcore will check if a node with the same label is present and assign an
OF node to the MTD.

I currently tested this on my device that have smem-part and the
gmac driver use nvmem to get the mac-address. This works correctly and
the same address is provided.

v5:
- Simplify Documentation and move it generic partition.yaml
- Split commit and move smem example to dedicated commit.
v4:
- Make it simple. No suffix. Make label mandatory again.
- Update Documentation with new implementation.
- Rename files to a better and correct name
v3:
- Fix warning from bot (function not declared as static)
- Updated code to support also node name
- Made partition label optional
v2:
- Simplify this. Drop dynamic-partition
- Fix problem with parser with ko
- Do not pollude mtd_get_of_node
- Fix problem with Documentation

Ansuel Smith (3):
  dt-bindings: mtd: partitions: Support label only partition
  dt-bindings: mtd: partitions: add additional example for
    qcom,smem-part
  mtd: core: introduce of support for dynamic partitions

 .../bindings/mtd/partitions/partition.yaml    | 16 +++++-
 .../mtd/partitions/qcom,smem-part.yaml        | 27 ++++++++++
 drivers/mtd/mtdcore.c                         | 49 +++++++++++++++++++
 3 files changed, 90 insertions(+), 2 deletions(-)

Comments

Christian Marangi June 9, 2022, 7:57 p.m. UTC | #1
On Thu, Jun 09, 2022 at 12:32:52PM -0600, Rob Herring wrote:
> On Mon, Jun 06, 2022 at 05:14:15PM +0200, Ansuel Smith wrote:
> > Document new partition nodes that declare only the label instead of the
> > reg used to provide an OF node for partition registred at runtime by
> > parsers. This is required for nvmem system to declare and detect
> > nvmem-cells.
> > 
> > With these special partitions, the reg / offset is not required.
> > The label binding is used to match the partition allocated by the
> > parser at runtime and the parser will provide reg and offset of the mtd.
> > 
> > NVMEM will use the data from the parser and provide the NVMEM cells
> > declared in the DTS, "connecting" the dynamic partition with a
> > static declaration of cells in them.
> > 
> > Signed-off-by: Ansuel Smith <ansuelsmth@gmail.com>
> > ---
> >  .../bindings/mtd/partitions/partition.yaml       | 16 ++++++++++++++--
> >  1 file changed, 14 insertions(+), 2 deletions(-)
> > 
> > diff --git a/Documentation/devicetree/bindings/mtd/partitions/partition.yaml b/Documentation/devicetree/bindings/mtd/partitions/partition.yaml
> > index e1ac08064425..bff6fb980e6b 100644
> > --- a/Documentation/devicetree/bindings/mtd/partitions/partition.yaml
> > +++ b/Documentation/devicetree/bindings/mtd/partitions/partition.yaml
> > @@ -11,6 +11,13 @@ description: |
> >    relative offset and size specified. Depending on partition function extra
> >    properties can be used.
> >  
> > +  A partition may be dynamically allocated by a specific parser at runtime.
> > +  In this specific case, the label is required instead of the reg.
> > +  This is used to assign an OF node to the dynamiccally allocated partition
> > +  so that subsystem like NVMEM can provide an OF node and declare NVMEM cells.
> > +  The OF node will be assigned only if the partition label declared match the
> > +  one assigned by the parser at runtime.
> > +
> >  maintainers:
> >    - Rafał Miłecki <rafal@milecki.pl>
> >  
> > @@ -22,6 +29,8 @@ properties:
> >    label:
> >      description: The label / name for this partition. If omitted, the label
> >        is taken from the node name (excluding the unit address).
> > +      With dynamically allocated partition the label is required and won't
> > +      fallback to the node name.
> 
> Generally, label is never required being something for humans rather 
> than the s/w to consume. I don't see any reason why we can't still use 
> the node name (with 'partition-' stripped off).
> 

How to enforce the use of 'partition-'? Should the driver then check the
node name and reject any wrong node name (and return error)?


> If the purpose is to define what the partition contains, then 
> 'compatible' is the right thing for that.
>

Introducing a compatible means creating another scheme I think or we can
add that special compatible in the partition scheme?

> >  
> >    read-only:
> >      description: This parameter, if present, is a hint that this partition
> > @@ -41,7 +50,10 @@ properties:
> >        immune to paired-pages corruptions
> >      type: boolean
> >  
> > -required:
> > -  - reg
> > +if:
> > +  not:
> > +    required: [ reg ]
> > +then:
> > +  required: [ label ]
> >  
> >  additionalProperties: true
> > -- 
> > 2.36.1
> > 
> >
Christian Marangi June 11, 2022, 12:36 p.m. UTC | #2
On Fri, Jun 10, 2022 at 11:02:05AM -0600, Rob Herring wrote:
> On Thu, Jun 09, 2022 at 09:57:52PM +0200, Ansuel Smith wrote:
> > On Thu, Jun 09, 2022 at 12:32:52PM -0600, Rob Herring wrote:
> > > On Mon, Jun 06, 2022 at 05:14:15PM +0200, Ansuel Smith wrote:
> > > > Document new partition nodes that declare only the label instead of the
> > > > reg used to provide an OF node for partition registred at runtime by
> > > > parsers. This is required for nvmem system to declare and detect
> > > > nvmem-cells.
> > > > 
> > > > With these special partitions, the reg / offset is not required.
> > > > The label binding is used to match the partition allocated by the
> > > > parser at runtime and the parser will provide reg and offset of the mtd.
> > > > 
> > > > NVMEM will use the data from the parser and provide the NVMEM cells
> > > > declared in the DTS, "connecting" the dynamic partition with a
> > > > static declaration of cells in them.
> > > > 
> > > > Signed-off-by: Ansuel Smith <ansuelsmth@gmail.com>
> > > > ---
> > > >  .../bindings/mtd/partitions/partition.yaml       | 16 ++++++++++++++--
> > > >  1 file changed, 14 insertions(+), 2 deletions(-)
> > > > 
> > > > diff --git a/Documentation/devicetree/bindings/mtd/partitions/partition.yaml b/Documentation/devicetree/bindings/mtd/partitions/partition.yaml
> > > > index e1ac08064425..bff6fb980e6b 100644
> > > > --- a/Documentation/devicetree/bindings/mtd/partitions/partition.yaml
> > > > +++ b/Documentation/devicetree/bindings/mtd/partitions/partition.yaml
> > > > @@ -11,6 +11,13 @@ description: |
> > > >    relative offset and size specified. Depending on partition function extra
> > > >    properties can be used.
> > > >  
> > > > +  A partition may be dynamically allocated by a specific parser at runtime.
> > > > +  In this specific case, the label is required instead of the reg.
> > > > +  This is used to assign an OF node to the dynamiccally allocated partition
> > > > +  so that subsystem like NVMEM can provide an OF node and declare NVMEM cells.
> > > > +  The OF node will be assigned only if the partition label declared match the
> > > > +  one assigned by the parser at runtime.
> > > > +
> > > >  maintainers:
> > > >    - Rafał Miłecki <rafal@milecki.pl>
> > > >  
> > > > @@ -22,6 +29,8 @@ properties:
> > > >    label:
> > > >      description: The label / name for this partition. If omitted, the label
> > > >        is taken from the node name (excluding the unit address).
> > > > +      With dynamically allocated partition the label is required and won't
> > > > +      fallback to the node name.
> > > 
> > > Generally, label is never required being something for humans rather 
> > > than the s/w to consume. I don't see any reason why we can't still use 
> > > the node name (with 'partition-' stripped off).
> > > 
> > 
> > How to enforce the use of 'partition-'? Should the driver then check the
> > node name and reject any wrong node name (and return error)?
> 
> The schema can do it either in the parent (of partition nodes) schema or 
> with $nodename 'property'.
> 
> $nodename:
>   oneOf:
>     - pattern: '^.*@.*$'
>     - pattern: '^partition-.*$'
> 
> or:
> 
> if:
>   not:
>     required:
>       - reg
> then:
>   properties:
>     $nodename:
>       pattern: '^partition-.*$'
> 
> 
> The latter is a bit clearer on the intent I think.
> 
> > > If the purpose is to define what the partition contains, then 
> > > 'compatible' is the right thing for that.
> > >
> > 
> > Introducing a compatible means creating another scheme I think or we can
> > add that special compatible in the partition scheme?
> 
> It would be another schema. You could make 'compatible' required here 
> perhaps, but maybe there's a use for an empty node?
> 
> Rob

Think I will just add the $nodename properties just to now complicate
things with another scheme.
Han Xu Oct. 17, 2022, 7:59 p.m. UTC | #3
On 22/06/06 05:14PM, Ansuel Smith wrote:
> We have many parser that register mtd partitions at runtime. One example
> is the cmdlinepart or the smem-part parser where the compatible is defined
> in the dts and the partitions gets detected and registered by the
> parser. This is problematic for the NVMEM subsystem that requires an OF node
> to detect NVMEM cells.
> 
> To fix this problem, introduce an additional logic that will try to
> assign an OF node to the MTD if declared.
> 
> On MTD addition, it will be checked if the MTD has an OF node and if
> not declared will check if a partition with the same label is
> declared in DTS. If an exact match is found, the partition dynamically
> allocated by the parser will have a connected OF node.
> 
> The NVMEM subsystem will detect the OF node and register any NVMEM cells
> declared statically in the DTS.
> 
> Signed-off-by: Ansuel Smith <ansuelsmth@gmail.com>
> ---
>  drivers/mtd/mtdcore.c | 49 +++++++++++++++++++++++++++++++++++++++++++
>  1 file changed, 49 insertions(+)
> 
> diff --git a/drivers/mtd/mtdcore.c b/drivers/mtd/mtdcore.c
> index 7731796024e0..807194efb580 100644
> --- a/drivers/mtd/mtdcore.c
> +++ b/drivers/mtd/mtdcore.c
> @@ -546,6 +546,54 @@ static int mtd_nvmem_add(struct mtd_info *mtd)
>  	return 0;
>  }
>  
> +static void mtd_check_of_node(struct mtd_info *mtd)
> +{
> +	struct device_node *partitions, *parent_dn, *mtd_dn = NULL;
> +	struct mtd_info *parent;
> +	const char *mtd_name;
> +	bool found = false;
> +	int plen;
> +
> +	/* Check if MTD already has a device node */
> +	if (dev_of_node(&mtd->dev))
> +		return;
> +
> +	/* Check if a partitions node exist */
> +	parent = mtd->parent;
> +	parent_dn = dev_of_node(&parent->dev);
> +	if (!parent_dn)
> +		return;
> +
> +	partitions = of_get_child_by_name(parent_dn, "partitions");
> +	if (!partitions)
> +		goto exit_parent;
> +
> +	/* Search if a partition is defined with the same name */
> +	for_each_child_of_node(partitions, mtd_dn) {
> +		/* Skip partition with no label */
> +		mtd_name = of_get_property(mtd_dn, "label", &plen);
> +		if (!mtd_name)
> +			continue;
> +
> +		if (!strncmp(mtd->name, mtd_name, plen)) {
> +			found = true;
> +			break;
> +		}
> +	}
> +
> +	if (!found)
> +		goto exit_partitions;
> +
> +	/* Set of_node only for nvmem */
> +	if (of_device_is_compatible(mtd_dn, "nvmem-cells"))
> +		mtd_set_of_node(mtd, mtd_dn);
> +
> +exit_partitions:
> +	of_node_put(partitions);
> +exit_parent:
> +	of_node_put(parent_dn);
> +}
> +
>  /**
>   *	add_mtd_device - register an MTD device
>   *	@mtd: pointer to new MTD device info structure
> @@ -651,6 +699,7 @@ int add_mtd_device(struct mtd_info *mtd)
>  	mtd->dev.devt = MTD_DEVT(i);
>  	dev_set_name(&mtd->dev, "mtd%d", i);
>  	dev_set_drvdata(&mtd->dev, mtd);
> +	mtd_check_of_node(mtd);
>  	of_node_get(mtd_get_of_node(mtd));
>  	error = device_register(&mtd->dev);
>  	if (error)

NXP GPMI NAND controller with 6 cmdline partitions meets refcount underflow
with this patch. The of_node_put(parent_dn) doesn't work with cmdline parser and
causes the issue.

[    1.078910] 6 cmdlinepart partitions found on MTD device gpmi-nand
[    1.085116] Creating 6 MTD partitions on "gpmi-nand":
[    1.090181] 0x000000000000-0x000008000000 : "nandboot"
[    1.096952] 0x000008000000-0x000009000000 : "nandfit"
[    1.103547] 0x000009000000-0x00000b000000 : "nandkernel"
[    1.110317] 0x00000b000000-0x00000c000000 : "nanddtb"
[    1.115525] ------------[ cut here ]------------
[    1.120141] refcount_t: addition on 0; use-after-free.
[    1.125328] WARNING: CPU: 0 PID: 1 at lib/refcount.c:25 refcount_warn_saturate+0xdc/0x148
[    1.133528] Modules linked in:
[    1.136589] CPU: 0 PID: 1 Comm: swapper/0 Not tainted 6.0.0-rc7-next-20220930-04543-g8cf3f7
[    1.146342] Hardware name: Freescale i.MX8DXL DDR3L EVK (DT)
[    1.151999] pstate: 600000c5 (nZCv daIF -PAN -UAO -TCO -DIT -SSBS BTYPE=--)
[    1.158965] pc : refcount_warn_saturate+0xdc/0x148
[    1.163760] lr : refcount_warn_saturate+0xdc/0x148
[    1.168556] sp : ffff800009ddb080
[    1.171866] x29: ffff800009ddb080 x28: ffff800009ddb35a x27: 0000000000000002
[    1.179015] x26: ffff8000098b06ad x25: ffffffffffffffff x24: ffff0a00ffffff05
[    1.186165] x23: ffff00001fdf6470 x22: ffff800009ddb367 x21: 0000000000000000
[    1.193314] x20: ffff00001fdfebe8 x19: ffff00001fdfec50 x18: ffffffffffffffff
[    1.200464] x17: 0000000000000000 x16: 0000000000000118 x15: 0000000000000004
[    1.207614] x14: 0000000000000fff x13: ffff800009bca248 x12: 0000000000000003
[    1.214764] x11: 00000000ffffefff x10: c0000000ffffefff x9 : 4762cb2ccb52de00
[    1.221914] x8 : 4762cb2ccb52de00 x7 : 205d313431303231 x6 : 312e31202020205b
[    1.229063] x5 : ffff800009d55c1f x4 : 0000000000000001 x3 : 0000000000000000
[    1.236213] x2 : 0000000000000000 x1 : ffff800009954be6 x0 : 000000000000002a
[    1.243365] Call trace:
[    1.245806]  refcount_warn_saturate+0xdc/0x148
[    1.250253]  kobject_get+0x98/0x9c
[    1.253658]  of_node_get+0x20/0x34
[    1.257072]  of_fwnode_get+0x3c/0x54
[    1.260652]  fwnode_get_nth_parent+0xd8/0xf4
[    1.264926]  fwnode_full_name_string+0x3c/0xb4
[    1.269373]  device_node_string+0x498/0x5b4
[    1.273561]  pointer+0x41c/0x5d0
[    1.276793]  vsnprintf+0x4d8/0x694
[    1.280198]  vprintk_store+0x164/0x528
[    1.283951]  vprintk_emit+0x98/0x164
[    1.287530]  vprintk_default+0x44/0x6c
[    1.291284]  vprintk+0xf0/0x134
[    1.294428]  _printk+0x54/0x7c
[    1.297486]  of_node_release+0xe8/0x128
[    1.301326]  kobject_put+0x98/0xfc
[    1.304732]  of_node_put+0x1c/0x28
[    1.308137]  add_mtd_device+0x484/0x6d4
[    1.311977]  add_mtd_partitions+0xf0/0x1d0
[    1.316078]  parse_mtd_partitions+0x45c/0x518
[    1.320439]  mtd_device_parse_register+0xb0/0x274
[    1.325147]  gpmi_nand_probe+0x51c/0x650
[    1.329074]  platform_probe+0xa8/0xd0
[    1.332740]  really_probe+0x130/0x334
[    1.336406]  __driver_probe_device+0xb4/0xe0
[    1.340681]  driver_probe_device+0x3c/0x1f8
[    1.344869]  __driver_attach+0xdc/0x1a4
[    1.348708]  bus_for_each_dev+0x80/0xcc
[    1.352548]  driver_attach+0x24/0x30
[    1.356127]  bus_add_driver+0x108/0x1f4
[    1.359967]  driver_register+0x78/0x114
[    1.363807]  __platform_driver_register+0x24/0x30
[    1.368515]  gpmi_nand_driver_init+0x1c/0x28
[    1.372798]  do_one_initcall+0xbc/0x238
[    1.376638]  do_initcall_level+0x94/0xb4
[    1.380565]  do_initcalls+0x54/0x94
[    1.384058]  do_basic_setup+0x1c/0x28
[    1.387724]  kernel_init_freeable+0x110/0x188
[    1.392084]  kernel_init+0x20/0x1a0
[    1.395578]  ret_from_fork+0x10/0x20
[    1.399157] ---[ end trace 0000000000000000 ]---
[    1.403782] ------------[ cut here ]------------


> -- 
> 2.36.1
> 
> 
> ______________________________________________________
> Linux MTD discussion mailing list
> http://lists.infradead.org/mailman/listinfo/linux-mtd/
Rafał Miłecki Oct. 17, 2022, 9:52 p.m. UTC | #4
On 17.10.2022 21:59, han.xu wrote:
> On 22/06/06 05:14PM, Ansuel Smith wrote:
>> We have many parser that register mtd partitions at runtime. One example
>> is the cmdlinepart or the smem-part parser where the compatible is defined
>> in the dts and the partitions gets detected and registered by the
>> parser. This is problematic for the NVMEM subsystem that requires an OF node
>> to detect NVMEM cells.
>>
>> To fix this problem, introduce an additional logic that will try to
>> assign an OF node to the MTD if declared.
>>
>> On MTD addition, it will be checked if the MTD has an OF node and if
>> not declared will check if a partition with the same label is
>> declared in DTS. If an exact match is found, the partition dynamically
>> allocated by the parser will have a connected OF node.
>>
>> The NVMEM subsystem will detect the OF node and register any NVMEM cells
>> declared statically in the DTS.
>>
>> Signed-off-by: Ansuel Smith <ansuelsmth@gmail.com>
>> ---
>>   drivers/mtd/mtdcore.c | 49 +++++++++++++++++++++++++++++++++++++++++++
>>   1 file changed, 49 insertions(+)
>>
>> diff --git a/drivers/mtd/mtdcore.c b/drivers/mtd/mtdcore.c
>> index 7731796024e0..807194efb580 100644
>> --- a/drivers/mtd/mtdcore.c
>> +++ b/drivers/mtd/mtdcore.c
>> @@ -546,6 +546,54 @@ static int mtd_nvmem_add(struct mtd_info *mtd)
>>   	return 0;
>>   }
>>   
>> +static void mtd_check_of_node(struct mtd_info *mtd)
>> +{
>> +	struct device_node *partitions, *parent_dn, *mtd_dn = NULL;
>> +	struct mtd_info *parent;
>> +	const char *mtd_name;
>> +	bool found = false;
>> +	int plen;
>> +
>> +	/* Check if MTD already has a device node */
>> +	if (dev_of_node(&mtd->dev))
>> +		return;
>> +
>> +	/* Check if a partitions node exist */
>> +	parent = mtd->parent;
>> +	parent_dn = dev_of_node(&parent->dev);
>> +	if (!parent_dn)
>> +		return;
>> +
>> +	partitions = of_get_child_by_name(parent_dn, "partitions");
>> +	if (!partitions)
>> +		goto exit_parent;
>> +
>> +	/* Search if a partition is defined with the same name */
>> +	for_each_child_of_node(partitions, mtd_dn) {
>> +		/* Skip partition with no label */
>> +		mtd_name = of_get_property(mtd_dn, "label", &plen);
>> +		if (!mtd_name)
>> +			continue;
>> +
>> +		if (!strncmp(mtd->name, mtd_name, plen)) {
>> +			found = true;
>> +			break;
>> +		}
>> +	}
>> +
>> +	if (!found)
>> +		goto exit_partitions;
>> +
>> +	/* Set of_node only for nvmem */
>> +	if (of_device_is_compatible(mtd_dn, "nvmem-cells"))
>> +		mtd_set_of_node(mtd, mtd_dn);
>> +
>> +exit_partitions:
>> +	of_node_put(partitions);
>> +exit_parent:
>> +	of_node_put(parent_dn);
>> +}
>> +
>>   /**
>>    *	add_mtd_device - register an MTD device
>>    *	@mtd: pointer to new MTD device info structure
>> @@ -651,6 +699,7 @@ int add_mtd_device(struct mtd_info *mtd)
>>   	mtd->dev.devt = MTD_DEVT(i);
>>   	dev_set_name(&mtd->dev, "mtd%d", i);
>>   	dev_set_drvdata(&mtd->dev, mtd);
>> +	mtd_check_of_node(mtd);
>>   	of_node_get(mtd_get_of_node(mtd));
>>   	error = device_register(&mtd->dev);
>>   	if (error)
> 
> NXP GPMI NAND controller with 6 cmdline partitions meets refcount underflow
> with this patch. The of_node_put(parent_dn) doesn't work with cmdline parser and
> causes the issue.

Can you try:

diff --git a/drivers/mtd/mtdcore.c b/drivers/mtd/mtdcore.c
index 18aa54460d36..0b4ca0aa4132 100644
--- a/drivers/mtd/mtdcore.c
+++ b/drivers/mtd/mtdcore.c
@@ -562,7 +562,7 @@ static void mtd_check_of_node(struct mtd_info *mtd)
  	if (!mtd_is_partition(mtd))
  		return;
  	parent = mtd->parent;
-	parent_dn = dev_of_node(&parent->dev);
+	parent_dn = of_node_get(dev_of_node(&parent->dev));
  	if (!parent_dn)
  		return;
Han Xu Oct. 18, 2022, 2:48 a.m. UTC | #5
On 22/10/17 11:52PM, Rafał Miłecki wrote:
> On 17.10.2022 21:59, han.xu wrote:
> > On 22/06/06 05:14PM, Ansuel Smith wrote:
> > > We have many parser that register mtd partitions at runtime. One example
> > > is the cmdlinepart or the smem-part parser where the compatible is defined
> > > in the dts and the partitions gets detected and registered by the
> > > parser. This is problematic for the NVMEM subsystem that requires an OF node
> > > to detect NVMEM cells.
> > > 
> > > To fix this problem, introduce an additional logic that will try to
> > > assign an OF node to the MTD if declared.
> > > 
> > > On MTD addition, it will be checked if the MTD has an OF node and if
> > > not declared will check if a partition with the same label is
> > > declared in DTS. If an exact match is found, the partition dynamically
> > > allocated by the parser will have a connected OF node.
> > > 
> > > The NVMEM subsystem will detect the OF node and register any NVMEM cells
> > > declared statically in the DTS.
> > > 
> > > Signed-off-by: Ansuel Smith <ansuelsmth@gmail.com>
> > > ---
> > >   drivers/mtd/mtdcore.c | 49 +++++++++++++++++++++++++++++++++++++++++++
> > >   1 file changed, 49 insertions(+)
> > > 
> > > diff --git a/drivers/mtd/mtdcore.c b/drivers/mtd/mtdcore.c
> > > index 7731796024e0..807194efb580 100644
> > > --- a/drivers/mtd/mtdcore.c
> > > +++ b/drivers/mtd/mtdcore.c
> > > @@ -546,6 +546,54 @@ static int mtd_nvmem_add(struct mtd_info *mtd)
> > >   	return 0;
> > >   }
> > > +static void mtd_check_of_node(struct mtd_info *mtd)
> > > +{
> > > +	struct device_node *partitions, *parent_dn, *mtd_dn = NULL;
> > > +	struct mtd_info *parent;
> > > +	const char *mtd_name;
> > > +	bool found = false;
> > > +	int plen;
> > > +
> > > +	/* Check if MTD already has a device node */
> > > +	if (dev_of_node(&mtd->dev))
> > > +		return;
> > > +
> > > +	/* Check if a partitions node exist */
> > > +	parent = mtd->parent;
> > > +	parent_dn = dev_of_node(&parent->dev);
> > > +	if (!parent_dn)
> > > +		return;
> > > +
> > > +	partitions = of_get_child_by_name(parent_dn, "partitions");
> > > +	if (!partitions)
> > > +		goto exit_parent;
> > > +
> > > +	/* Search if a partition is defined with the same name */
> > > +	for_each_child_of_node(partitions, mtd_dn) {
> > > +		/* Skip partition with no label */
> > > +		mtd_name = of_get_property(mtd_dn, "label", &plen);
> > > +		if (!mtd_name)
> > > +			continue;
> > > +
> > > +		if (!strncmp(mtd->name, mtd_name, plen)) {
> > > +			found = true;
> > > +			break;
> > > +		}
> > > +	}
> > > +
> > > +	if (!found)
> > > +		goto exit_partitions;
> > > +
> > > +	/* Set of_node only for nvmem */
> > > +	if (of_device_is_compatible(mtd_dn, "nvmem-cells"))
> > > +		mtd_set_of_node(mtd, mtd_dn);
> > > +
> > > +exit_partitions:
> > > +	of_node_put(partitions);
> > > +exit_parent:
> > > +	of_node_put(parent_dn);
> > > +}
> > > +
> > >   /**
> > >    *	add_mtd_device - register an MTD device
> > >    *	@mtd: pointer to new MTD device info structure
> > > @@ -651,6 +699,7 @@ int add_mtd_device(struct mtd_info *mtd)
> > >   	mtd->dev.devt = MTD_DEVT(i);
> > >   	dev_set_name(&mtd->dev, "mtd%d", i);
> > >   	dev_set_drvdata(&mtd->dev, mtd);
> > > +	mtd_check_of_node(mtd);
> > >   	of_node_get(mtd_get_of_node(mtd));
> > >   	error = device_register(&mtd->dev);
> > >   	if (error)
> > 
> > NXP GPMI NAND controller with 6 cmdline partitions meets refcount underflow
> > with this patch. The of_node_put(parent_dn) doesn't work with cmdline parser and
> > causes the issue.
> 
> Can you try:
> 
> diff --git a/drivers/mtd/mtdcore.c b/drivers/mtd/mtdcore.c
> index 18aa54460d36..0b4ca0aa4132 100644
> --- a/drivers/mtd/mtdcore.c
> +++ b/drivers/mtd/mtdcore.c
> @@ -562,7 +562,7 @@ static void mtd_check_of_node(struct mtd_info *mtd)
>  	if (!mtd_is_partition(mtd))
>  		return;
>  	parent = mtd->parent;
> -	parent_dn = dev_of_node(&parent->dev);
> +	parent_dn = of_node_get(dev_of_node(&parent->dev));
>  	if (!parent_dn)
>  		return;

Yes, with of_node_get() the refcount issue gone.

> 
>