Message-ID: <20210401101812.GE14052@work>
Date:   Thu, 1 Apr 2021 15:48:12 +0530
From:   Manivannan Sadhasivam <manivannan.sadhasivam@...aro.org>
To:     Miquel Raynal <miquel.raynal@...tlin.com>
Cc:     richard@....at, vigneshr@...com, robh+dt@...nel.org,
        linux-arm-msm@...r.kernel.org, devicetree@...r.kernel.org,
        linux-mtd@...ts.infradead.org, linux-kernel@...r.kernel.org,
        boris.brezillon@...labora.com, Daniele.Palmas@...it.com,
        bjorn.andersson@...aro.org
Subject: Re: [PATCH v8 3/3] mtd: rawnand: Add support for secure regions in
 NAND memory

Hi Miquel,

On Tue, Mar 23, 2021 at 05:57:15PM +0100, Miquel Raynal wrote:
> Hi Manivannan,
> 
> Manivannan Sadhasivam <manivannan.sadhasivam@...aro.org> wrote on Tue,
> 23 Mar 2021 13:09:30 +0530:
> 
> > On a typical end product, a vendor may choose to secure some regions in
> > the NAND memory which are supposed to stay intact between FW upgrades.
> > The access to those regions will be blocked by a secure element like
> > Trustzone. So the normal world software like Linux kernel should not
> > touch these regions (including reading).
> > 
> > The regions are declared using a NAND chip DT property,
> > "secure-regions". So let's make use of this property in the raw NAND
> > core and skip access to the secure regions present in a system.
> > 
> > Signed-off-by: Manivannan Sadhasivam <manivannan.sadhasivam@...aro.org>
> > ---
> >  drivers/mtd/nand/raw/nand_base.c | 105 +++++++++++++++++++++++++++++++
> >  include/linux/mtd/rawnand.h      |  14 +++++
> >  2 files changed, 119 insertions(+)
> > 
> > diff --git a/drivers/mtd/nand/raw/nand_base.c b/drivers/mtd/nand/raw/nand_base.c
> > index c33fa1b1847f..2a990219f498 100644
> > --- a/drivers/mtd/nand/raw/nand_base.c
> > +++ b/drivers/mtd/nand/raw/nand_base.c
> > @@ -278,11 +278,46 @@ static int nand_block_bad(struct nand_chip *chip, loff_t ofs)
> >  	return 0;
> >  }
> >  
> > +/**
> > + * nand_check_secure_region() - Check if the region is secured
> > + * @chip: NAND chip object
> > + * @offset: Offset of the region to check
> > + * @size: Size of the region to check
> > + *
> > + * Checks if the region is secured by comparing the offset and size with the
> > + * list of secure regions obtained from DT. Returns -EIO if the region is
> > + * secured else 0.
> > + */
> > +static int nand_check_secure_region(struct nand_chip *chip, loff_t offset, u64 size)
> 
> I think I would prefer a boolean return value here, with a rename:
> 
> static bool nand_region_is_secured() or
> nand_region_is_accessible/reachable/whatever()
> 
> then something lik:
> 
> 	if (nand_region_is_secured())
> 		return -EIO;
>

Okay

> > +{
> > +	int i;
> > +
> > +	/* Skip touching the secure regions if present */
> > +	for (i = 0; i < chip->nr_secure_regions; i++) {
> > +		const struct nand_secure_region *region = &chip->secure_regions[i];
> > +
> > +		if (offset + size < region->offset ||
> > +		    offset >= region->offset + region->size)
> 
> I think as-is the condition does not work.
> 
> Let's assume we want to check the region { .offset = 1, size = 1 } and
> the region { .offset = 2, size = 1 } is reserved. This is:
> 
> 		if ((1 + 1 < 2) /* false */ ||
> 		    (1 >= 2 + 1) /* false */)
> 			continue;
> 		return -EIO; /* EIO is returned while the area is valid

I made a mistake. I should've used "offset + size <= region->offset" as
suggested by Boris.

The reason why I didn't go for it because the SoC was still accessing
the secure region with (>=). So I went with just (>) blindly :/

The actual issue was with the check at nand_isbad_bbm(), where I didn't
pass the size of the region to check, instead just offset as below:

	nand_check_secure_region(chip, ofs, 0);

Because of this, the check went fine but since the block_bad() function
reads the blocks starting from the offset, the secure region was
accessed.

For fixing this, I'm going to use below diff:

diff --git a/drivers/mtd/nand/raw/nand_base.c b/drivers/mtd/nand/raw/nand_base.c
index 2a990219f498..53589c835f66 100644
--- a/drivers/mtd/nand/raw/nand_base.c
+++ b/drivers/mtd/nand/raw/nand_base.c
@@ -296,7 +296,7 @@ static int nand_check_secure_region(struct nand_chip *chip, loff_t offset, u64 s
        for (i = 0; i < chip->nr_secure_regions; i++) {
                const struct nand_secure_region *region = &chip->secure_regions[i];
 
-               if (offset + size < region->offset ||
+               if (offset + size <= region->offset ||
                    offset >= region->offset + region->size)
                        continue;
 
@@ -308,13 +308,16 @@ static int nand_check_secure_region(struct nand_chip *chip, loff_t offset, u64 s
 
 static int nand_isbad_bbm(struct nand_chip *chip, loff_t ofs)
 {
+       struct mtd_info *mtd = nand_to_mtd(chip);
+       int last_page = ((mtd->erasesize - mtd->writesize) >>
+                        chip->page_shift) & chip->pagemask;
        int ret;
 
        if (chip->options & NAND_NO_BBM_QUIRK)
                return 0;
 
        /* Check if the region is secured */
-       ret = nand_check_secure_region(chip, ofs, 0);
+       ret = nand_check_secure_region(chip, ofs, last_page);
        if (ret)
                return ret;

> 		*/
> 
> > +			continue;
> > +
> 
> Perhaps a dev_dbg() entry here would make sense.
> 

Okay

> > +		return -EIO;
> > +	}
> > +
> > +	return 0;
> > +}
> > +
> 
> [...]
> 
> > +static int of_get_nand_secure_regions(struct nand_chip *chip)
> > +{
> > +	struct device_node *dn = nand_get_flash_node(chip);
> > +	struct property *prop;
> > +	int length, nr_elem, i, j;
> > +
> > +	prop = of_find_property(dn, "secure-regions", &length);
> > +	if (prop) {
> 
> I generally prefer the below logic:
> 
> 	if (!prop)
> 		return 0;
> 
> Then you earn an indentation level.
> 
> > +		nr_elem = length / sizeof(u64);
> 
> of_property_count_elems_of_size() ?
> 

Okay

> > +		chip->nr_secure_regions = nr_elem / 2;
> > +
> > +		chip->secure_regions = kcalloc(nr_elem, sizeof(*chip->secure_regions), GFP_KERNEL);
> 
> IIRC ->secure_regions is a structure with lengths and offset, so you
> don't want to allocate nr_elem but nr_secure_regions number of
> items here.
> 

Oh yeah, I missed it.

> > +		if (!chip->secure_regions)
> > +			return -ENOMEM;
> > +
> > +		for (i = 0, j = 0; i < chip->nr_secure_regions; i++, j += 2) {
> > +			of_property_read_u64_index(dn, "secure-regions", j,
> > +						   &chip->secure_regions[i].offset);
> > +			of_property_read_u64_index(dn, "secure-regions", j + 1,
> > +						   &chip->secure_regions[i].size);
> > +		}
> > +	}
> > +
> > +	return 0;
> > +}
> > +
> >  static int rawnand_dt_init(struct nand_chip *chip)
> >  {
> >  	struct nand_device *nand = mtd_to_nanddev(nand_to_mtd(chip));
> >  	struct device_node *dn = nand_get_flash_node(chip);
> > +	int ret;
> >  
> >  	if (!dn)
> >  		return 0;
> > @@ -5015,6 +5107,16 @@ static int rawnand_dt_init(struct nand_chip *chip)
> >  	of_get_nand_ecc_user_config(nand);
> >  	of_get_nand_ecc_legacy_user_config(chip);
> >  
> > +	/*
> > +	 * Look for secure regions in the NAND chip. These regions are supposed
> > +	 * to be protected by a secure element like Trustzone. So the read/write
> > +	 * accesses to these regions will be blocked in the runtime by this
> > +	 * driver.
> > +	 */
> > +	ret = of_get_nand_secure_regions(chip);
> > +	if (!ret)
> > +		return ret;
> 
> I think we can do this initialization pretty much when we want in the
> init process as long as it is done before the BBT parsing logic.
> 
> Here, besides the fact the memory will not be freed from
> rawnand_dt_init()'s caller if something goes wrong, we are at a point
> where nand_cleanup will not be called. nand_cleanup() will only be
> called if the controller driver encounters an error *after* a
> successful nand_scan().
> 
> We could perhaps move this call to nand_scan() which would simply solve
> the situation. We don't need it in rawnand_dt_init() as this won't be
> rawnand specific anyway...
> 

Okay, will do.

Thanks,
Mani

> > +
> >  	/*
> >  	 * If neither the user nor the NAND controller have
> > requested a specific
> >  	 * ECC engine type, we will default to
> > NAND_ECC_ENGINE_TYPE_ON_HOST. @@ -6068,6 +6170,9 @@ void
> > nand_cleanup(struct nand_chip *chip) /* Free manufacturer priv data.
> > */ nand_manufacturer_cleanup(chip);
> >  
> > +	/* Free secure regions data */
> > +	kfree(chip->secure_regions);
> > +
> >  	/* Free controller specific allocations after chip
> > identification */ nand_detach(chip);
> >  
> > diff --git a/include/linux/mtd/rawnand.h b/include/linux/mtd/rawnand.h
> > index 6b3240e44310..17ddc900a1dc 100644
> > --- a/include/linux/mtd/rawnand.h
> > +++ b/include/linux/mtd/rawnand.h
> > @@ -1036,6 +1036,16 @@ struct nand_manufacturer {
> >  	void *priv;
> >  };
> >  
> > +/**
> > + * struct nand_secure_region - NAND secure region structure
> > + * @offset: Offset of the start of the secure region
> > + * @size: Size of the secure region
> > + */
> > +struct nand_secure_region {
> > +	u64 offset;
> > +	u64 size;
> > +};
> > +
> >  /**
> >   * struct nand_chip - NAND Private Flash Chip Data
> >   * @base: Inherit from the generic NAND device
> > @@ -1086,6 +1096,8 @@ struct nand_manufacturer {
> >   *          NAND Controller drivers should not modify this value,
> > but they're
> >   *          allowed to read it.
> >   * @read_retries: The number of read retry modes supported
> > + * @secure_regions: Structure containing the secure regions info
> > + * @nr_secure_regions: Number of secure regions
> >   * @controller: The hardware controller	structure which is
> > shared among multiple
> >   *              independent devices
> >   * @ecc: The ECC controller structure
> > @@ -1135,6 +1147,8 @@ struct nand_chip {
> >  	unsigned int suspended : 1;
> >  	int cur_cs;
> >  	int read_retries;
> > +	struct nand_secure_region *secure_regions;
> > +	u8 nr_secure_regions;
> >  
> >  	/* Externals */
> >  	struct nand_controller *controller;
> 
> Thanks,
> Miquèl

Powered by blists - more mailing lists