[PATCH v5 0/6] Introduce NVIDIA Tegra Partition Table

[PATCH v5 0/6] Introduce NVIDIA Tegra Partition Table

[Dmitry Osipenko]

Hello,

This series adds support for the NVIDIA Tegra Partition Table format,
which is needed in order to support consumer-grade Tegra-based devices
by the upstream kernel.  These devices have Secure Boot enabled and it
can't be disabled, and thus, it's not possible to easily modify bootloader
and eMMC storage partitioning.

Big thanks to everyone who helped with figuring out the TegraPT format
and with testing!

Changelog:

v5: - Corrected Kconfig dependency, like it was suggested by Randy Dunlap
      in the review comment to v4.

    - Corrected a typo in the commit message, which was spotted by
      Steve McIntyre in v4.

    - The new force_gpt_sector variable has been moved out to the common
      parser state because I realized that multiple drives could be scanned
      simultaneously, and thus, a global shared variable isn't a suitable
      variant.

    - The Tegra's enforced GPT-scan now supports older bootloader versions,
      which do not provide the gpt_sector= command line option.  The gpt_sector
      should be calculated using NVIDIA's algorithm in this case.

v4: - Scanning of the eMMC boot partitions has been dropped because it
      requires a bit too messy hacks in the kernel.  We can live without
      this feature, at least for now it's not really needed.

    - Instead of the dropped boot partitions scanning, the "gpt_sector="
      command line option has been brought back [1].  But now it's done
      in a different way, not bothering platforms other than Tegra and
      not touching block devices other than eMMC, which was requested during
      of the [1] review.  The "gpt_sector=" usage is needed when partition
      table is inaccessible by kernel, which is the case for the Ouya game
      console device for example.  Please note that "gpt_sector=" is not
      available on all devices, such devices will fall back to the TegraPT,
      Samsung Galaxy Tab 10.1 is an example of a such device.

      [1] https://patchwork.ozlabs.org/project/linux-tegra/patch/20200219162339.16192-1-digetx@gmail.com/

    - We got a bit more broad testing and discovered that Samsung Galaxy
      Tab 10.1 uses 2K logical sectors instead of 4K.  So the 2K sectors
      are supported now by the TegraPT parser.  Thanks to Nils Östlund for
      the testing!

    - TegraPT parser now utilizes "tegraboot=" cmdline option which is
      passed to kernel by the NVIDIA's bootloader.  It's used for verifying
      that SDMMC device is *the* boot source.

    - Added patch that exposes Boot Configuration Table to userspace via
      sysfs.  Thanks to Michał Mirosław for the suggestion!

    - Misc changes:
        - The EB2 (second bootloader) partition is added to the list of
          known partitions.  Used by the Galaxy Tab 10.

        - The blkdev logical sector size is checked now, for consistency.
          It always should be 512 bytes on the supported/tested devices.

        - Verbose error messages are replaced with pr_debug().

v3: - Fixed "BUG: KASAN: slab-out-of-bounds in tegra_partition".  Thanks to
      Peter Geis for noticing the problem.

    - The MMC boot partitions scanning is now opt-in. See this patch:

        mmc: block: Support partition-table scanning on boot partitions

    - The found MMC boot partitions won't be assigned to the MMC boot
      block device ever due to the new GENHD_FL_PART_SCAN_ONCE flag.

      This makes us to ensure that the old behavior of the MMC core is
      preserved for a non-Tegra MMC-block users.

    New patches in v3:

        block: Introduce GENHD_FL_PART_SCAN_ONCE
        mmc: sdhci-tegra: Enable boot partitions scanning on Tegra20 and Tegra30

v2: - Addressed v1 review comments from Stephen Warren by using BIT for
      locating BCT position in IRAM.

    - Added more validations to the TegraPT parser: partition type is
      verified, eMMC instance ID is verified.

    - TegraPT parser now doesn't touch any devices other than eMMC.

    - EKS (encrypted keys) partition is blacklisted now.

    - Implemented eMMC boot partitions scanning.  These new patches are
      added in a result:

        mmc: block: Add mmc_bdev_to_part_type() helper
        mmc: block: Add mmc_bdev_to_area_type() helper
        mmc: block: Add MMC_QUIRK_RESCAN_MAIN_BLKDEV
        mmc: block: Enable partition-table scanning for boot partitions
        partitions/tegra: Implement eMMC boot partitions scanning

Dmitry Osipenko (6):
  mmc: core: Add raw_boot_mult field to mmc_ext_csd
  mmc: block: Add mmc_bdev_to_card() helper
  partitions: Introduce NVIDIA Tegra Partition Table
  partitions/efi: Support GPT entry lookup at a non-standard location
  partitions/tegra: Support enforced GPT scanning
  soc/tegra: Expose Boot Configuration Table via sysfs

 arch/arm/mach-tegra/tegra.c   |  58 ++++
 block/partitions/Kconfig      |   8 +
 block/partitions/Makefile     |   1 +
 block/partitions/check.h      |   3 +
 block/partitions/core.c       |   4 +
 block/partitions/efi.c        |  14 +
 block/partitions/tegra.c      | 624 ++++++++++++++++++++++++++++++++++
 drivers/mmc/core/block.c      |  15 +
 drivers/mmc/core/mmc.c        |   2 +
 drivers/soc/tegra/Makefile    |   1 +
 drivers/soc/tegra/bootdata.c  |  51 +++
 drivers/soc/tegra/common.c    |  17 +
 include/linux/mmc/blkdev.h    |  13 +
 include/linux/mmc/card.h      |   1 +
 include/soc/tegra/bootdata.h  |  48 +++
 include/soc/tegra/common.h    |  12 +
 include/soc/tegra/partition.h |  84 +++++
 17 files changed, 956 insertions(+)
 create mode 100644 block/partitions/tegra.c
 create mode 100644 drivers/soc/tegra/bootdata.c
 create mode 100644 include/linux/mmc/blkdev.h
 create mode 100644 include/soc/tegra/bootdata.h
 create mode 100644 include/soc/tegra/partition.h

-- 
2.26.0

[PATCH v5 1/6] mmc: core: Add raw_boot_mult field to mmc_ext_csd

[Dmitry Osipenko]

In order to support parsing of NVIDIA Tegra Partition Table format, we
need to know the BOOT_SIZE_MULT value of the Extended CSD register because
NVIDIA's bootloader linearizes the boot0/boot1/main partitions into a
single virtual space, and thus, all partition addresses are shifted by
the size of boot0 + boot1 partitions.

Signed-off-by: Dmitry Osipenko <digetx@gmail.com>
---
 drivers/mmc/core/mmc.c   | 2 ++
 include/linux/mmc/card.h | 1 +
 2 files changed, 3 insertions(+)

diff --git a/drivers/mmc/core/mmc.c b/drivers/mmc/core/mmc.c
index 4203303f946a..112edfb1eb1d 100644
--- a/drivers/mmc/core/mmc.c
+++ b/drivers/mmc/core/mmc.c
@@ -417,6 +417,8 @@ static int mmc_decode_ext_csd(struct mmc_card *card, u8 *ext_csd)
 		ext_csd[EXT_CSD_ERASE_TIMEOUT_MULT];
 	card->ext_csd.raw_hc_erase_grp_size =
 		ext_csd[EXT_CSD_HC_ERASE_GRP_SIZE];
+	card->ext_csd.raw_boot_mult =
+		ext_csd[EXT_CSD_BOOT_MULT];
 	if (card->ext_csd.rev >= 3) {
 		u8 sa_shift = ext_csd[EXT_CSD_S_A_TIMEOUT];
 		card->ext_csd.part_config = ext_csd[EXT_CSD_PART_CONFIG];
diff --git a/include/linux/mmc/card.h b/include/linux/mmc/card.h
index 7d46411ffaa2..cd6b58b66010 100644
--- a/include/linux/mmc/card.h
+++ b/include/linux/mmc/card.h
@@ -109,6 +109,7 @@ struct mmc_ext_csd {
 	u8			raw_hc_erase_gap_size;	/* 221 */
 	u8			raw_erase_timeout_mult;	/* 223 */
 	u8			raw_hc_erase_grp_size;	/* 224 */
+	u8			raw_boot_mult;		/* 226 */
 	u8			raw_sec_trim_mult;	/* 229 */
 	u8			raw_sec_erase_mult;	/* 230 */
 	u8			raw_sec_feature_support;/* 231 */
-- 
2.26.0

[PATCH v5 2/6] mmc: block: Add mmc_bdev_to_card() helper

[Dmitry Osipenko]

NVIDIA Tegra Partition Table takes into account MMC card's BOOT_SIZE_MULT
parameter, and thus, the partition parser needs to retrieve that EXT_CSD
value from the block device.  There are also some other parts of struct
mmc_card that are needed for the partition parser in order to calculate
the eMMC offset and verify different things.  This patch introduces new
helper which takes block device for the input argument and returns the
corresponding MMC card.

Signed-off-by: Dmitry Osipenko <digetx@gmail.com>
---
 drivers/mmc/core/block.c   | 15 +++++++++++++++
 include/linux/mmc/blkdev.h | 13 +++++++++++++
 2 files changed, 28 insertions(+)
 create mode 100644 include/linux/mmc/blkdev.h

diff --git a/drivers/mmc/core/block.c b/drivers/mmc/core/block.c
index c5367e2c8487..99298e888381 100644
--- a/drivers/mmc/core/block.c
+++ b/drivers/mmc/core/block.c
@@ -40,6 +40,7 @@
 #include <linux/debugfs.h>
 
 #include <linux/mmc/ioctl.h>
+#include <linux/mmc/blkdev.h>
 #include <linux/mmc/card.h>
 #include <linux/mmc/host.h>
 #include <linux/mmc/mmc.h>
@@ -305,6 +306,20 @@ static ssize_t force_ro_store(struct device *dev, struct device_attribute *attr,
 	return ret;
 }
 
+struct mmc_card *mmc_bdev_to_card(struct block_device *bdev)
+{
+	struct mmc_blk_data *md;
+
+	if (bdev->bd_disk->major != MMC_BLOCK_MAJOR)
+		return NULL;
+
+	md = mmc_blk_get(bdev->bd_disk);
+	if (!md)
+		return NULL;
+
+	return md->queue.card;
+}
+
 static int mmc_blk_open(struct block_device *bdev, fmode_t mode)
 {
 	struct mmc_blk_data *md = mmc_blk_get(bdev->bd_disk);
diff --git a/include/linux/mmc/blkdev.h b/include/linux/mmc/blkdev.h
new file mode 100644
index 000000000000..67608c58de70
--- /dev/null
+++ b/include/linux/mmc/blkdev.h
@@ -0,0 +1,13 @@
+/* SPDX-License-Identifier: GPL-2.0-only */
+/*
+ *  linux/include/linux/mmc/blkdev.h
+ */
+#ifndef LINUX_MMC_BLOCK_DEVICE_H
+#define LINUX_MMC_BLOCK_DEVICE_H
+
+struct block_device;
+struct mmc_card;
+
+struct mmc_card *mmc_bdev_to_card(struct block_device *bdev);
+
+#endif /* LINUX_MMC_BLOCK_DEVICE_H */
-- 
2.26.0

[PATCH v5 3/6] partitions: Introduce NVIDIA Tegra Partition Table

[Dmitry Osipenko]

All NVIDIA Tegra devices use a special partition table format for the
internal storage partitioning.  Most of Tegra devices have GPT partition
in addition to TegraPT, but some older Android consumer-grade devices do
not or GPT is placed in a wrong sector, and thus, the TegraPT is needed
in order to support these devices properly by the upstream kernel. This
patch adds support for NVIDIA Tegra Partition Table format that is used
at least by all NVIDIA Tegra20 and Tegra30 devices.

Tested-by: Nils Östlund <nils@naltan.com>
Signed-off-by: Dmitry Osipenko <digetx@gmail.com>
---
 arch/arm/mach-tegra/tegra.c   |  54 ++++
 block/partitions/Kconfig      |   8 +
 block/partitions/Makefile     |   1 +
 block/partitions/check.h      |   1 +
 block/partitions/core.c       |   3 +
 block/partitions/tegra.c      | 567 ++++++++++++++++++++++++++++++++++
 include/soc/tegra/bootdata.h  |  46 +++
 include/soc/tegra/common.h    |   9 +
 include/soc/tegra/partition.h |  84 +++++
 9 files changed, 773 insertions(+)
 create mode 100644 block/partitions/tegra.c
 create mode 100644 include/soc/tegra/bootdata.h
 create mode 100644 include/soc/tegra/partition.h

diff --git a/arch/arm/mach-tegra/tegra.c b/arch/arm/mach-tegra/tegra.c
index c011359bcdb4..da6bcd85398b 100644
--- a/arch/arm/mach-tegra/tegra.c
+++ b/arch/arm/mach-tegra/tegra.c
@@ -28,7 +28,9 @@
 
 #include <linux/firmware/trusted_foundations.h>
 
+#include <soc/tegra/bootdata.h>
 #include <soc/tegra/fuse.h>
+#include <soc/tegra/partition.h>
 #include <soc/tegra/pmc.h>
 
 #include <asm/firmware.h>
@@ -62,9 +64,61 @@ u32 tegra_uart_config[3] = {
 	0,
 };
 
+static void __init tegra_boot_config_table_init(void)
+{
+	struct tegra30_boot_config_table __iomem *t30_bct;
+	struct tegra20_boot_config_table __iomem *t20_bct;
+	struct tegra20_boot_info_table   __iomem *t20_bit;
+	u32 iram_end   = TEGRA_IRAM_BASE + TEGRA_IRAM_SIZE;
+	u32 iram_start = TEGRA_IRAM_BASE;
+	u32 pt_addr, pt_size, bct_size;
+
+	t20_bit = IO_ADDRESS(TEGRA_IRAM_BASE);
+
+	if (of_machine_is_compatible("nvidia,tegra20")) {
+		bct_size = sizeof(*t20_bct);
+
+		if (t20_bit->bct_size != bct_size ||
+		    t20_bit->bct_ptr < iram_start ||
+		    t20_bit->bct_ptr > iram_end - bct_size)
+			return;
+
+		t20_bct = IO_ADDRESS(t20_bit->bct_ptr);
+
+		if (t20_bct->boot_data_version != TEGRA_BOOTDATA_VERSION_T20)
+			return;
+
+		pt_addr = t20_bct->partition_table_logical_sector_address;
+		pt_size = t20_bct->partition_table_num_logical_sectors;
+
+	} else if (of_machine_is_compatible("nvidia,tegra30")) {
+		bct_size = sizeof(*t30_bct);
+
+		if (t20_bit->bct_size != bct_size ||
+		    t20_bit->bct_ptr < iram_start ||
+		    t20_bit->bct_ptr > iram_end - bct_size)
+			return;
+
+		t30_bct = IO_ADDRESS(t20_bit->bct_ptr);
+
+		if (t30_bct->boot_data_version != TEGRA_BOOTDATA_VERSION_T30)
+			return;
+
+		pt_addr = t30_bct->partition_table_logical_sector_address;
+		pt_size = t30_bct->partition_table_num_logical_sectors;
+	} else {
+		return;
+	}
+
+	pr_info("%s: BCT found in IRAM\n", __func__);
+
+	tegra_partition_table_setup(pt_addr, pt_size);
+}
+
 static void __init tegra_init_early(void)
 {
 	of_register_trusted_foundations();
+	tegra_boot_config_table_init();
 	tegra_cpu_reset_handler_init();
 	call_firmware_op(l2x0_init);
 }
diff --git a/block/partitions/Kconfig b/block/partitions/Kconfig
index 702689a628f0..97cd4a030f98 100644
--- a/block/partitions/Kconfig
+++ b/block/partitions/Kconfig
@@ -268,3 +268,11 @@ config CMDLINE_PARTITION
 	help
 	  Say Y here if you want to read the partition table from bootargs.
 	  The format for the command line is just like mtdparts.
+
+config TEGRA_PARTITION
+	bool "NVIDIA Tegra Partition support" if PARTITION_ADVANCED
+	default y if ARCH_TEGRA
+	depends on MMC_BLOCK && (ARCH_TEGRA || COMPILE_TEST)
+	help
+	  Say Y here if you would like to be able to read the hard disk
+	  partition table format used by NVIDIA Tegra machines.
diff --git a/block/partitions/Makefile b/block/partitions/Makefile
index a7f05cdb02a8..83cb70c6d08d 100644
--- a/block/partitions/Makefile
+++ b/block/partitions/Makefile
@@ -20,3 +20,4 @@ obj-$(CONFIG_IBM_PARTITION) += ibm.o
 obj-$(CONFIG_EFI_PARTITION) += efi.o
 obj-$(CONFIG_KARMA_PARTITION) += karma.o
 obj-$(CONFIG_SYSV68_PARTITION) += sysv68.o
+obj-$(CONFIG_TEGRA_PARTITION) += tegra.o
diff --git a/block/partitions/check.h b/block/partitions/check.h
index c577e9ee67f0..ffa01cc4b0b0 100644
--- a/block/partitions/check.h
+++ b/block/partitions/check.h
@@ -67,4 +67,5 @@ int osf_partition(struct parsed_partitions *state);
 int sgi_partition(struct parsed_partitions *state);
 int sun_partition(struct parsed_partitions *state);
 int sysv68_partition(struct parsed_partitions *state);
+int tegra_partition(struct parsed_partitions *state);
 int ultrix_partition(struct parsed_partitions *state);
diff --git a/block/partitions/core.c b/block/partitions/core.c
index 297004fd2264..0b4720372f07 100644
--- a/block/partitions/core.c
+++ b/block/partitions/core.c
@@ -81,6 +81,9 @@ static int (*check_part[])(struct parsed_partitions *) = {
 #endif
 #ifdef CONFIG_SYSV68_PARTITION
 	sysv68_partition,
+#endif
+#ifdef CONFIG_TEGRA_PARTITION
+	tegra_partition,
 #endif
 	NULL
 };
diff --git a/block/partitions/tegra.c b/block/partitions/tegra.c
new file mode 100644
index 000000000000..d3a00ade145a
--- /dev/null
+++ b/block/partitions/tegra.c
@@ -0,0 +1,567 @@
+// SPDX-License-Identifier: GPL-2.0
+/*
+ * NVIDIA Tegra Partition Table
+ *
+ * Copyright (C) 2020 GRATE-DRIVER project
+ * Copyright (C) 2020 Dmitry Osipenko <digetx@gmail.com>
+ *
+ * Credits for the partition table format:
+ *
+ *   Andrey Danin <danindrey@mail.ru>       (Toshiba AC100 TegraPT format)
+ *   Gilles Grandou <gilles@grandou.net>    (Toshiba AC100 TegraPT format)
+ *   Ryan Grachek <ryan@edited.us>          (Google TV "Molly" TegraPT format)
+ *   Stephen Warren <swarren@wwwdotorg.org> (Useful suggestions about eMMC/etc)
+ */
+
+#define pr_fmt(fmt) "tegra-partition: " fmt
+
+#include <linux/blkdev.h>
+#include <linux/init.h>
+#include <linux/kernel.h>
+#include <linux/of.h>
+#include <linux/sizes.h>
+#include <linux/slab.h>
+#include <linux/string.h>
+
+#include <linux/mmc/blkdev.h>
+#include <linux/mmc/card.h>
+#include <linux/mmc/host.h>
+
+#include <soc/tegra/common.h>
+#include <soc/tegra/partition.h>
+
+#include "check.h"
+
+#define TEGRA_PT_SECTOR_SIZE(ptp)	((ptp)->logical_sector_size / SZ_512)
+#define TEGRA_PT_SECTOR(ptp, s)		((s) * TEGRA_PT_SECTOR_SIZE(ptp))
+
+#define TEGRA_PT_HEADER_SIZE						\
+	(sizeof(struct tegra_partition_header_insecure) +		\
+	 sizeof(struct tegra_partition_header_secure))			\
+
+#define TEGRA_PT_MAX_PARTITIONS(ptp)					\
+	(((ptp)->logical_sector_size - TEGRA_PT_HEADER_SIZE) /		\
+	 sizeof(struct tegra_partition))
+
+#define TEGRA_PT_ERR(ptp, fmt, ...)					\
+	pr_debug("%s: " fmt,						\
+		 (ptp)->state->bdev->bd_disk->disk_name, ##__VA_ARGS__)	\
+
+#define TEGRA_PT_PARSE_ERR(ptp, fmt, ...)				\
+	TEGRA_PT_ERR(ptp, "sector %llu: invalid " fmt,			\
+		     (ptp)->sector, ##__VA_ARGS__)
+
+struct tegra_partition_table_parser {
+	struct tegra_partition_table *pt;
+	unsigned int logical_sector_size;
+	struct parsed_partitions *state;
+	bool pt_entry_checked;
+	sector_t sector;
+	int boot_offset;
+	u32 dev_instance;
+	u32 dev_id;
+};
+
+union tegra_partition_table_u {
+	struct tegra_partition_table pt;
+	u8 pt_parts[SZ_4K / SZ_512][SZ_512];
+};
+
+struct tegra_partition_type {
+	unsigned int type;
+	char *name;
+};
+
+static sector_t tegra_pt_logical_sector_address;
+static sector_t tegra_pt_logical_sectors_num;
+
+void tegra_partition_table_setup(unsigned int logical_sector_address,
+				 unsigned int logical_sectors_num)
+{
+	tegra_pt_logical_sector_address = logical_sector_address;
+	tegra_pt_logical_sectors_num    = logical_sectors_num;
+
+	pr_info("initialized to logical sector = %llu sectors_num = %llu\n",
+		tegra_pt_logical_sector_address, tegra_pt_logical_sectors_num);
+}
+
+/*
+ * Some partitions are very sensitive, changing data on them may brick device.
+ *
+ * For more details about partitions see:
+ *
+ *  "https://docs.nvidia.com/jetson/l4t/Tegra Linux Driver Package Development Guide/part_config.html"
+ */
+static const char * const partitions_blacklist[] = {
+	"BCT", "EBT", "EB2", "EKS", "GP1", "GPT", "MBR", "PT",
+};
+
+static bool tegra_partition_name_match(struct tegra_partition *p,
+				       const char *name)
+{
+	return !strncmp(p->partition_name, name, TEGRA_PT_NAME_SIZE);
+}
+
+static bool tegra_partition_skip(struct tegra_partition *p,
+				 struct tegra_partition_table_parser *ptp,
+				 sector_t sector)
+{
+	unsigned int i;
+
+	/* skip eMMC boot partitions */
+	if (sector < ptp->boot_offset)
+		return true;
+
+	for (i = 0; i < ARRAY_SIZE(partitions_blacklist); i++) {
+		if (tegra_partition_name_match(p, partitions_blacklist[i]))
+			return true;
+	}
+
+	return false;
+}
+
+static const struct tegra_partition_type tegra_partition_expected_types[] = {
+	{ .type = TEGRA_PT_PART_TYPE_BCT,	.name = "BCT", },
+	{ .type = TEGRA_PT_PART_TYPE_EBT,	.name = "EBT", },
+	{ .type = TEGRA_PT_PART_TYPE_EBT,	.name = "EB2", },
+	{ .type = TEGRA_PT_PART_TYPE_PT,	.name = "PT",  },
+	{ .type = TEGRA_PT_PART_TYPE_GP1,	.name = "GP1", },
+	{ .type = TEGRA_PT_PART_TYPE_GPT,	.name = "GPT", },
+	{ .type = TEGRA_PT_PART_TYPE_GENERIC,	.name = NULL,  },
+};
+
+static int tegra_partition_type_valid(struct tegra_partition_table_parser *ptp,
+				      struct tegra_partition *p)
+{
+	const struct tegra_partition_type *ptype;
+	unsigned int i;
+
+	for (i = 0; i < ARRAY_SIZE(tegra_partition_expected_types); i++) {
+		ptype = &tegra_partition_expected_types[i];
+
+		if (ptype->name && !tegra_partition_name_match(p, ptype->name))
+			continue;
+
+		if (p->part_info.partition_type == ptype->type)
+			return 0;
+
+		/*
+		 * Unsure about all possible types, let's emit error and
+		 * allow to continue for now.
+		 */
+		if (!ptype->name)
+			return 1;
+	}
+
+	return -1;
+}
+
+static bool tegra_partition_valid(struct tegra_partition_table_parser *ptp,
+				  struct tegra_partition *p,
+				  struct tegra_partition *prev,
+				  sector_t sector,
+				  sector_t size)
+{
+	struct tegra_partition_info *prev_pi = &prev->part_info;
+	sector_t sect_end = TEGRA_PT_SECTOR(ptp,
+					    prev_pi->logical_sector_address +
+					    prev_pi->logical_sectors_num);
+	char *type, name[2][TEGRA_PT_NAME_SIZE + 1];
+	int err;
+
+	strscpy(name[0], p->partition_name,    sizeof(name[0]));
+	strscpy(name[1], prev->partition_name, sizeof(name[1]));
+
+	/* validate expected partition name/type */
+	err = tegra_partition_type_valid(ptp, p);
+	if (err) {
+		TEGRA_PT_PARSE_ERR(ptp, "partition_type: [%s] partition_type=%u\n",
+				   name[0], p->part_info.partition_type);
+		if (err < 0)
+			return false;
+
+		TEGRA_PT_ERR(ptp, "continuing, please update list of expected types\n");
+	}
+
+	/* validate partition table BCT addresses */
+	if (tegra_partition_name_match(p, "PT")) {
+		if (sector != TEGRA_PT_SECTOR(ptp, tegra_pt_logical_sector_address) &&
+		    size   != TEGRA_PT_SECTOR(ptp, tegra_pt_logical_sectors_num)) {
+			TEGRA_PT_PARSE_ERR(ptp, "PT location: sector=%llu size=%llu\n",
+					   sector, size);
+			return false;
+		}
+
+		if (ptp->pt_entry_checked) {
+			TEGRA_PT_PARSE_ERR(ptp, "(duplicated) PT\n");
+			return false;
+		}
+
+		ptp->pt_entry_checked = true;
+	}
+
+	if (sector + size < sector) {
+		TEGRA_PT_PARSE_ERR(ptp, "size: [%s] integer overflow sector=%llu size=%llu\n",
+				   name[0], sector, size);
+		return false;
+	}
+
+	/* validate allocation_policy=sequential (absolute unsupported) */
+	if (p != prev && sect_end > sector) {
+		TEGRA_PT_PARSE_ERR(ptp, "allocation_policy: [%s] end=%llu [%s] sector=%llu size=%llu\n",
+				   name[1], sect_end, name[0], sector, size);
+		return false;
+	}
+
+	if (ptp->dev_instance != p->mount_info.device_instance) {
+		TEGRA_PT_PARSE_ERR(ptp, "device_instance: [%s] device_instance=%u|%u\n",
+				   name[0], ptp->dev_instance,
+				   p->mount_info.device_instance);
+		return false;
+	}
+
+	if (ptp->dev_id != p->mount_info.device_id) {
+		TEGRA_PT_PARSE_ERR(ptp, "device_id: [%s] device_id=%u|%u\n",
+				   name[0], ptp->dev_id,
+				   p->mount_info.device_id);
+		return false;
+	}
+
+	if (p->partition_id > 127) {
+		TEGRA_PT_PARSE_ERR(ptp, "partition_id: [%s] partition_id=%u\n",
+				   name[0], p->partition_id);
+		return false;
+	}
+
+	sect_end = get_capacity(ptp->state->bdev->bd_disk);
+
+	/* eMMC boot partitions are below ptp->boot_offset */
+	if (sector < ptp->boot_offset) {
+		sect_end += ptp->boot_offset;
+		type = "boot";
+	} else {
+		sector -= ptp->boot_offset;
+		type = "main";
+	}
+
+	/* validate size */
+	if (!size || sector + size > sect_end) {
+		TEGRA_PT_PARSE_ERR(ptp, "size: [%s] %s partition boot_offt=%d end=%llu sector=%llu size=%llu\n",
+				   name[0], type, ptp->boot_offset, sect_end,
+				   sector, size);
+		return false;
+	}
+
+	return true;
+}
+
+static bool tegra_partitions_parsed(struct tegra_partition_table_parser *ptp,
+				    bool check_only)
+{
+	struct parsed_partitions *state = ptp->state;
+	struct tegra_partition_table *pt = ptp->pt;
+	sector_t sector, size;
+	int i, slot = 1;
+
+	ptp->pt_entry_checked = false;
+
+	for (i = 0; i < pt->secure.num_partitions; i++) {
+		struct tegra_partition *p = &pt->partitions[i];
+		struct tegra_partition *prev = &pt->partitions[max(i - 1, 0)];
+		struct tegra_partition_info *pi = &p->part_info;
+
+		if (slot == state->limit && !check_only)
+			break;
+
+		sector = TEGRA_PT_SECTOR(ptp, pi->logical_sector_address);
+		size   = TEGRA_PT_SECTOR(ptp, pi->logical_sectors_num);
+
+		if (check_only &&
+		    !tegra_partition_valid(ptp, p, prev, sector, size))
+			return false;
+
+		if (check_only ||
+		    tegra_partition_skip(p, ptp, sector))
+			continue;
+
+		put_partition(state, slot++, sector - ptp->boot_offset, size);
+	}
+
+	if (check_only && !ptp->pt_entry_checked) {
+		TEGRA_PT_PARSE_ERR(ptp, "PT: table entry not found\n");
+		return false;
+	}
+
+	return true;
+}
+
+static bool
+tegra_partition_table_parsed(struct tegra_partition_table_parser *ptp)
+{
+	if (ptp->pt->secure.num_partitions == 0 ||
+	    ptp->pt->secure.num_partitions > TEGRA_PT_MAX_PARTITIONS(ptp)) {
+		TEGRA_PT_PARSE_ERR(ptp, "num_partitions=%u\n",
+				   ptp->pt->secure.num_partitions);
+		return false;
+	}
+
+	return tegra_partitions_parsed(ptp, true) &&
+	       tegra_partitions_parsed(ptp, false);
+}
+
+static int
+tegra_partition_table_insec_hdr_valid(struct tegra_partition_table_parser *ptp)
+{
+	if (ptp->pt->insecure.magic   != TEGRA_PT_MAGIC ||
+	    ptp->pt->insecure.version != TEGRA_PT_VERSION) {
+		TEGRA_PT_PARSE_ERR(ptp, "insecure header: magic=0x%llx ver=0x%x\n",
+				   ptp->pt->insecure.magic,
+				   ptp->pt->insecure.version);
+		return 0;
+	}
+
+	return 1;
+}
+
+static int
+tegra_partition_table_sec_hdr_valid(struct tegra_partition_table_parser *ptp)
+{
+	size_t pt_size = ptp->pt->secure.num_partitions;
+
+	pt_size *= sizeof(ptp->pt->partitions[0]);
+	pt_size += TEGRA_PT_HEADER_SIZE;
+
+	if (ptp->pt->secure.magic   != TEGRA_PT_MAGIC ||
+	    ptp->pt->secure.version != TEGRA_PT_VERSION ||
+	    ptp->pt->secure.length  != ptp->pt->insecure.length ||
+	    ptp->pt->secure.length  < pt_size) {
+		TEGRA_PT_PARSE_ERR(ptp, "secure header: magic=0x%llx ver=0x%x length=%u|%u|%zu\n",
+				   ptp->pt->secure.magic,
+				   ptp->pt->secure.version,
+				   ptp->pt->secure.length,
+				   ptp->pt->insecure.length,
+				   pt_size);
+		return 0;
+	}
+
+	return 1;
+}
+
+static int
+tegra_partition_table_unencrypted(struct tegra_partition_table_parser *ptp)
+{
+	/* AES IV, all zeros if unencrypted */
+	if (ptp->pt->secure.random_data[0] || ptp->pt->secure.random_data[1] ||
+	    ptp->pt->secure.random_data[2] || ptp->pt->secure.random_data[3]) {
+		pr_err_once("encrypted partition table unsupported\n");
+		return 0;
+	}
+
+	return 1;
+}
+
+static int tegra_read_partition_table(struct tegra_partition_table_parser *ptp)
+{
+	union tegra_partition_table_u *ptu = (typeof(ptu))ptp->pt;
+	unsigned int i;
+	Sector sect;
+	void *part;
+
+	for (i = 0; i < ptp->logical_sector_size / SZ_512; i++) {
+		/*
+		 * Partition table takes at maximum 4096 bytes, but
+		 * read_part_sector() guarantees only that SECTOR_SIZE will
+		 * be read at minimum.
+		 */
+		part = read_part_sector(ptp->state, ptp->sector + i, &sect);
+		if (!part) {
+			TEGRA_PT_ERR(ptp, "failed to read sector %llu\n",
+				     ptp->sector + i);
+			return 0;
+		}
+
+		memcpy(ptu->pt_parts[i], part, SZ_512);
+		put_dev_sector(sect);
+	}
+
+	return 1;
+}
+
+static int tegra_partition_scan(struct tegra_partition_table_parser *ptp)
+{
+	sector_t start_sector, num_sectors;
+	int ret = 0;
+
+	num_sectors  = TEGRA_PT_SECTOR(ptp, tegra_pt_logical_sectors_num);
+	start_sector = TEGRA_PT_SECTOR(ptp, tegra_pt_logical_sector_address);
+
+	if (start_sector < ptp->boot_offset) {
+		TEGRA_PT_ERR(ptp,
+			     "scanning eMMC boot partitions unimplemented\n");
+		return 0;
+	}
+
+	ptp->sector = start_sector - ptp->boot_offset;
+
+	/*
+	 * Partition table is duplicated for num_sectors.
+	 * If first table is corrupted, we will try next.
+	 */
+	while (num_sectors--) {
+		ret = tegra_read_partition_table(ptp);
+		if (!ret)
+			goto next_sector;
+
+		ret = tegra_partition_table_insec_hdr_valid(ptp);
+		if (!ret)
+			goto next_sector;
+
+		ret = tegra_partition_table_unencrypted(ptp);
+		if (!ret)
+			goto next_sector;
+
+		ret = tegra_partition_table_sec_hdr_valid(ptp);
+		if (!ret)
+			goto next_sector;
+
+		ret = tegra_partition_table_parsed(ptp);
+		if (ret)
+			break;
+next_sector:
+		ptp->sector += TEGRA_PT_SECTOR_SIZE(ptp);
+	}
+
+	return ret;
+}
+
+static const u32 tegra20_sdhci_bases[TEGRA_PT_SDHCI_DEVICE_INSTANCES] = {
+	0xc8000000, 0xc8000200, 0xc8000400, 0xc8000600,
+};
+
+static const u32 tegra30_sdhci_bases[TEGRA_PT_SDHCI_DEVICE_INSTANCES] = {
+	0x78000000, 0x78000200, 0x78000400, 0x78000600,
+};
+
+static const struct of_device_id tegra_sdhci_match[] = {
+	{ .compatible = "nvidia,tegra20-sdhci", .data = tegra20_sdhci_bases, },
+	{ .compatible = "nvidia,tegra30-sdhci", .data = tegra30_sdhci_bases, },
+	{}
+};
+
+static int
+tegra_partition_table_emmc_boot_offset(struct tegra_partition_table_parser *ptp)
+{
+	struct mmc_card *card = mmc_bdev_to_card(ptp->state->bdev);
+	const struct of_device_id *matched;
+	const u32 *sdhci_bases;
+	u32 sdhci_base;
+	unsigned int i;
+	int err;
+
+	/* filter out unexpected/untested boot sources */
+	if (!card || card->ext_csd.rev < 3 ||
+	    !mmc_card_is_blockaddr(card) ||
+	     mmc_card_is_removable(card->host) ||
+	     bdev_logical_block_size(ptp->state->bdev) != SZ_512) {
+		TEGRA_PT_ERR(ptp, "unexpected boot source\n");
+		return -1;
+	}
+
+	/* skip everything unrelated to Tegra eMMC */
+	matched = of_match_node(tegra_sdhci_match, card->host->parent->of_node);
+	if (!matched)
+		return -1;
+
+	sdhci_bases = matched->data;
+
+	/* figure out SDHCI instance ID by the base address */
+	err = of_property_read_u32_index(card->host->parent->of_node,
+					 "reg", 0, &sdhci_base);
+	if (err)
+		return -1;
+
+	for (i = 0; i < TEGRA_PT_SDHCI_DEVICE_INSTANCES; i++) {
+		if (sdhci_base == sdhci_bases[i])
+			break;
+	}
+
+	if (i == TEGRA_PT_SDHCI_DEVICE_INSTANCES)
+		return -1;
+
+	ptp->dev_id       = TEGRA_PT_SDHCI_DEVICE_ID;
+	ptp->dev_instance = i;
+
+	/*
+	 * eMMC storage has two special boot partitions in addition to the
+	 * main one.  NVIDIA's bootloader linearizes eMMC boot0->boot1->main
+	 * accesses, this means that the partition table addresses are shifted
+	 * by the size of boot partitions.  In accordance with the eMMC
+	 * specification, the boot partition size is calculated as follows:
+	 *
+	 *	boot partition size = 128K byte x BOOT_SIZE_MULT
+	 *
+	 * This function returns number of sectors occupied by the both boot
+	 * partitions.
+	 */
+	return card->ext_csd.raw_boot_mult * SZ_128K /
+	       SZ_512 * MMC_NUM_BOOT_PARTITION;
+}
+
+/*
+ * Logical sector size may vary per device model and apparently there is no
+ * way to get information about the size from kernel. The info is hardcoded
+ * into bootloader and it doesn't tell us, so we'll just try all possible
+ * well-known sizes until succeed.
+ *
+ * For example Samsung Galaxy Tab 10.1 uses 2K sectors. While Acer A500,
+ * Nexus 7 and Ouya are using 4K sectors.
+ */
+static const unsigned int tegra_pt_logical_sector_sizes[] = {
+	SZ_4K, SZ_2K,
+};
+
+/*
+ * The 'tegraboot=<source>' command line option is provided to kernel by
+ * NVIDIA's proprietary bootloader.
+ */
+static bool tegra_boot_sdmmc;
+static int __init tegra_boot_fn(char *str)
+{
+	tegra_boot_sdmmc = !strcmp(str, "sdmmc");
+	return 1;
+}
+__setup("tegraboot=", tegra_boot_fn);
+
+int tegra_partition(struct parsed_partitions *state)
+{
+	struct tegra_partition_table_parser ptp = {};
+	unsigned int i;
+	int ret;
+
+	if (!soc_is_tegra() || !tegra_boot_sdmmc)
+		return 0;
+
+	ptp.state = state;
+
+	ptp.boot_offset = tegra_partition_table_emmc_boot_offset(&ptp);
+	if (ptp.boot_offset < 0)
+		return 0;
+
+	ptp.pt = kmalloc(SZ_4K, GFP_KERNEL);
+	if (!ptp.pt)
+		return 0;
+
+	for (i = 0; i < ARRAY_SIZE(tegra_pt_logical_sector_sizes); i++) {
+		ptp.logical_sector_size = tegra_pt_logical_sector_sizes[i];
+
+		ret = tegra_partition_scan(&ptp);
+		if (ret == 1) {
+			strlcat(state->pp_buf, "\n", PAGE_SIZE);
+			break;
+		}
+	}
+
+	kfree(ptp.pt);
+
+	return ret;
+}
diff --git a/include/soc/tegra/bootdata.h b/include/soc/tegra/bootdata.h
new file mode 100644
index 000000000000..7be207cb2519
--- /dev/null
+++ b/include/soc/tegra/bootdata.h
@@ -0,0 +1,46 @@
+/* SPDX-License-Identifier: GPL-2.0-only */
+
+#ifndef __SOC_TEGRA_BOOTDATA_H__
+#define __SOC_TEGRA_BOOTDATA_H__
+
+#include <linux/compiler.h>
+#include <linux/types.h>
+
+#define TEGRA_BOOTDATA_VERSION_T20	NVBOOT_BOOTDATA_VERSION(0x2, 0x1)
+#define TEGRA_BOOTDATA_VERSION_T30	NVBOOT_BOOTDATA_VERSION(0x3, 0x1)
+
+#define NVBOOT_BOOTDATA_VERSION(a, b)	((((a) & 0xffff) << 16) | \
+					  ((b) & 0xffff))
+#define NVBOOT_CMAC_AES_HASH_LENGTH	4
+
+struct tegra20_boot_info_table {
+	u32 unused_data1[14];
+	u32 bct_size;
+	u32 bct_ptr;
+} __packed;
+
+struct tegra20_boot_config_table {
+	u32 crypto_hash[NVBOOT_CMAC_AES_HASH_LENGTH];
+	u32 random_aes_blk[NVBOOT_CMAC_AES_HASH_LENGTH];
+	u32 boot_data_version;
+	u32 unused_data1[712];
+	u32 unused_consumer_data1;
+	u16 partition_table_logical_sector_address;
+	u16 partition_table_num_logical_sectors;
+	u32 unused_consumer_data[294];
+	u32 unused_data[3];
+} __packed;
+
+struct tegra30_boot_config_table {
+	u32 crypto_hash[NVBOOT_CMAC_AES_HASH_LENGTH];
+	u32 random_aes_blk[NVBOOT_CMAC_AES_HASH_LENGTH];
+	u32 boot_data_version;
+	u32 unused_data1[1016];
+	u32 unused_consumer_data1;
+	u16 partition_table_logical_sector_address;
+	u16 partition_table_num_logical_sectors;
+	u32 unused_consumer_data[502];
+	u32 unused_data[3];
+} __packed;
+
+#endif /* __SOC_TEGRA_BOOTDATA_H__ */
diff --git a/include/soc/tegra/common.h b/include/soc/tegra/common.h
index 98027a76ce3d..744280ecab5f 100644
--- a/include/soc/tegra/common.h
+++ b/include/soc/tegra/common.h
@@ -6,6 +6,15 @@
 #ifndef __SOC_TEGRA_COMMON_H__
 #define __SOC_TEGRA_COMMON_H__
 
+#include <linux/types.h>
+
+#ifdef CONFIG_ARCH_TEGRA
 bool soc_is_tegra(void);
+#else
+static inline bool soc_is_tegra(void)
+{
+	return false;
+}
+#endif
 
 #endif /* __SOC_TEGRA_COMMON_H__ */
diff --git a/include/soc/tegra/partition.h b/include/soc/tegra/partition.h
new file mode 100644
index 000000000000..e8fcce1a725d
--- /dev/null
+++ b/include/soc/tegra/partition.h
@@ -0,0 +1,84 @@
+/* SPDX-License-Identifier: GPL-2.0-only */
+
+#ifndef __SOC_TEGRA_PARTITION_H__
+#define __SOC_TEGRA_PARTITION_H__
+
+#include <linux/compiler.h>
+#include <linux/types.h>
+
+#define TEGRA_PT_MAGIC				0xffffffff8f9e8d8bULL
+#define TEGRA_PT_VERSION			0x100
+#define TEGRA_PT_AES_HASH_SIZE			4
+#define TEGRA_PT_NAME_SIZE			4
+
+#define TEGRA_PT_SDHCI_DEVICE_ID		18
+#define TEGRA_PT_SDHCI_DEVICE_INSTANCES		4
+
+#define TEGRA_PT_PART_TYPE_BCT			1
+#define TEGRA_PT_PART_TYPE_EBT			2
+#define TEGRA_PT_PART_TYPE_PT			3
+#define TEGRA_PT_PART_TYPE_GENERIC		6
+#define TEGRA_PT_PART_TYPE_GP1			9
+#define TEGRA_PT_PART_TYPE_GPT			10
+
+struct tegra_partition_mount_info {
+	u32 device_id;
+	u32 device_instance;
+	u32 device_attr;
+	u8  mount_path[TEGRA_PT_NAME_SIZE];
+	u32 file_system_type;
+	u32 file_system_attr;
+} __packed;
+
+struct tegra_partition_info {
+	u32 partition_attr;
+	u32 __pad1;
+	u64 logical_sector_address;
+	u64 logical_sectors_num;
+	u64 physical_sector_address;
+	u64 physical_sectors_num;
+	u32 partition_type;
+	u32 __pad2;
+} __packed;
+
+struct tegra_partition {
+	u32 partition_id;
+	u8  partition_name[TEGRA_PT_NAME_SIZE];
+	struct tegra_partition_mount_info mount_info;
+	struct tegra_partition_info part_info;
+} __packed;
+
+struct tegra_partition_header_insecure {
+	u64 magic;
+	u32 version;
+	u32 length;
+	u32 signature[TEGRA_PT_AES_HASH_SIZE];
+} __packed;
+
+struct tegra_partition_header_secure {
+	u32 random_data[TEGRA_PT_AES_HASH_SIZE];
+	u64 magic;
+	u32 version;
+	u32 length;
+	u32 num_partitions;
+	u32 __pad;
+} __packed;
+
+struct tegra_partition_table {
+	struct tegra_partition_header_insecure insecure;
+	struct tegra_partition_header_secure secure;
+	struct tegra_partition partitions[];
+} __packed;
+
+#ifdef CONFIG_TEGRA_PARTITION
+void tegra_partition_table_setup(unsigned int logical_sector_address,
+				 unsigned int logical_sectors_num);
+#else
+static inline void
+tegra_partition_table_setup(unsigned int logical_sector_address,
+			    unsigned int logical_sectors_num)
+{
+}
+#endif
+
+#endif /* __SOC_TEGRA_PARTITION_H__ */
-- 
2.26.0

[PATCH v5 4/6] partitions/efi: Support GPT entry lookup at a non-standard location

[Dmitry Osipenko]

Most of consumer-grade NVIDIA Tegra devices use a proprietary bootloader
that can't be easily replaced because it's locked down using Secure Boot
cryptography signing and the crypto keys aren't given to a device owner.
These devices usually have eMMC storage that is partitioned using a custom
NVIDIA Tegra partition table format.  Of course bootloader and other
"special things" are stored on the eMMC storage, and thus, the partition
format can't be changed.

The Tegra partition format has been reverse-engineered, but NVIDIA uses
a virtually merged "boot" and "main" eMMC HW partitions in theirs
bootloader. This is not supported by Linux kernel and can't be easily
implemented. Hence partition table entry isn't accessible by kernel if
it's located at the "boot" eMMC partition. Luckily this is a rare case in
practice and even if it's the case, likely that the proprietary bootloader
will supply kernel with a non-standard gpt_sector= cmdline option.  This
gpt_sector= argument points at a GPT entry which resides at a non-standard
location on the eMMC storage.

This patch allows to support the non-standard cmdline option for NVIDIA
Tegra devices without bothering any other platforms.  The force_gpt_sector
parser-state struct member should be set before invoking efi_partition()
and it should be unset after the invocation completion.  This new parser
member instructs GPT parser to look up GPT entry at the given sector if
state->force_gpt_sector != 0.

This patch is based on the original work done by Colin Cross for the
downstream Android kernel.

Signed-off-by: Dmitry Osipenko <digetx@gmail.com>
---
 block/partitions/check.h | 1 +
 block/partitions/efi.c   | 8 ++++++++
 2 files changed, 9 insertions(+)

diff --git a/block/partitions/check.h b/block/partitions/check.h
index ffa01cc4b0b0..55acf6340e5b 100644
--- a/block/partitions/check.h
+++ b/block/partitions/check.h
@@ -22,6 +22,7 @@ struct parsed_partitions {
 	int limit;
 	bool access_beyond_eod;
 	char *pp_buf;
+	sector_t force_gpt_sector;
 };
 
 typedef struct {
diff --git a/block/partitions/efi.c b/block/partitions/efi.c
index b64bfdd4326c..3af4660bc11f 100644
--- a/block/partitions/efi.c
+++ b/block/partitions/efi.c
@@ -621,6 +621,14 @@ static int find_valid_gpt(struct parsed_partitions *state, gpt_header **gpt,
         if (!good_agpt && force_gpt)
                 good_agpt = is_gpt_valid(state, lastlba, &agpt, &aptes);
 
+	/* The force_gpt_sector is used by NVIDIA Tegra partition parser in
+	 * order to convey a non-standard location of the GPT entry for lookup.
+	 * By default force_gpt_sector is set to 0 and has no effect.
+	 */
+	if (!good_agpt && force_gpt && state->force_gpt_sector)
+		good_agpt = is_gpt_valid(state, state->force_gpt_sector,
+					 &agpt, &aptes);
+
         /* The obviously unsuccessful case */
         if (!good_pgpt && !good_agpt)
                 goto fail;
-- 
2.26.0

[PATCH v5 5/6] partitions/tegra: Support enforced GPT scanning

[Dmitry Osipenko]

Downstream NVIDIA bootloader provides gpt_sector=<sector> kernel command
line option to the kernel. This option should instruct the GPT partition
parser to look at the specified sector for a valid GPT header if the GPT
is not found at the beginning or the end of a block device. Support of
this feature is needed by Tegra-based devices that have TegraPT and GPT
placed in inaccessible by kernel locations.  The GPT entry duplicates
TegraPT partitions.

Secondly, some Tegra-based devices have bootloader that enforces the GPT
scanning of the backup/alternative GPT entry by providing "gpt" cmdline
option to the kernel, but doesn't provide the "gpt_sector" option.
In this case GPT entry resides at a special offset from the end of eMMC
storage.  It is a common situation for older bootloader versions.

The offset is calculated as a total number of eMMC sectors minus number of
eMMC boot sectors minus 1.  This equation is explicitly defined and used
by the downstream Tegra kernels for locating GPT entry.

Signed-off-by: Dmitry Osipenko <digetx@gmail.com>
---
 block/partitions/check.h |  1 +
 block/partitions/core.c  |  1 +
 block/partitions/efi.c   |  6 +++++
 block/partitions/tegra.c | 57 ++++++++++++++++++++++++++++++++++++++++
 4 files changed, 65 insertions(+)

diff --git a/block/partitions/check.h b/block/partitions/check.h
index 55acf6340e5b..1ce445d1c7f0 100644
--- a/block/partitions/check.h
+++ b/block/partitions/check.h
@@ -68,5 +68,6 @@ int osf_partition(struct parsed_partitions *state);
 int sgi_partition(struct parsed_partitions *state);
 int sun_partition(struct parsed_partitions *state);
 int sysv68_partition(struct parsed_partitions *state);
+int tegra_partition_forced_gpt(struct parsed_partitions *state);
 int tegra_partition(struct parsed_partitions *state);
 int ultrix_partition(struct parsed_partitions *state);
diff --git a/block/partitions/core.c b/block/partitions/core.c
index 0b4720372f07..1931647d9742 100644
--- a/block/partitions/core.c
+++ b/block/partitions/core.c
@@ -83,6 +83,7 @@ static int (*check_part[])(struct parsed_partitions *) = {
 	sysv68_partition,
 #endif
 #ifdef CONFIG_TEGRA_PARTITION
+	tegra_partition_forced_gpt,
 	tegra_partition,
 #endif
 	NULL
diff --git a/block/partitions/efi.c b/block/partitions/efi.c
index 3af4660bc11f..4eb4496fbb1b 100644
--- a/block/partitions/efi.c
+++ b/block/partitions/efi.c
@@ -98,6 +98,12 @@ static int force_gpt;
 static int __init
 force_gpt_fn(char *str)
 {
+	/* This check allows to parse "gpt gpt_sector=" properly since
+	 * "gpt" overlaps with "gpt_sector", see tegra_gpt_sector_fn().
+	 */
+	if (force_gpt)
+		return 0;
+
 	force_gpt = 1;
 	return 1;
 }
diff --git a/block/partitions/tegra.c b/block/partitions/tegra.c
index d3a00ade145a..831dedb9a11c 100644
--- a/block/partitions/tegra.c
+++ b/block/partitions/tegra.c
@@ -565,3 +565,60 @@ int tegra_partition(struct parsed_partitions *state)
 
 	return ret;
 }
+
+/*
+ * This allows a kernel command line option 'gpt_sector=<sector>' to
+ * enable GPT header lookup at a non-standard location. This option
+ * is provided to kernel by NVIDIA's proprietary bootloader.
+ */
+static sector_t tegra_gpt_sector;
+static int __init tegra_gpt_sector_fn(char *str)
+{
+	WARN_ON(kstrtoull(str, 10, &tegra_gpt_sector) < 0);
+	return 1;
+}
+__setup("gpt_sector=", tegra_gpt_sector_fn);
+
+int tegra_partition_forced_gpt(struct parsed_partitions *state)
+{
+	int ret = 0;
+
+#ifdef CONFIG_EFI_PARTITION
+	struct tegra_partition_table_parser ptp = {};
+
+	if (!soc_is_tegra() || !tegra_boot_sdmmc)
+		return 0;
+
+	ptp.state = state;
+
+	ptp.boot_offset = tegra_partition_table_emmc_boot_offset(&ptp);
+	if (ptp.boot_offset < 0)
+		return 0;
+
+	/*
+	 * Some Tegra devices do not use gpt_sector=<sector> kernel command
+	 * line option. In this case these devices usually have a GPT entry
+	 * at the end of the block device and the GPT entry address is
+	 * calculated this way for eMMC:
+	 *
+	 * gpt_sector = ext_csd.sectors_num - ext_csd.boot_sectors_num - 1
+	 *
+	 * This algorithm is defined and used by NVIDIA in the downstream
+	 * kernel of those devices.
+	 *
+	 * Please note that bootloader supplies the "gpt" cmdline option
+	 * which enforces the GPT scanning, meaning that the scanning will
+	 * be a NO-OP on devices that do not use GPT.
+	 */
+	if (tegra_gpt_sector) {
+		state->force_gpt_sector = tegra_gpt_sector;
+	} else {
+		state->force_gpt_sector  = get_capacity(state->bdev->bd_disk);
+		state->force_gpt_sector -= ptp.boot_offset + 1;
+	}
+
+	ret = efi_partition(state);
+	state->force_gpt_sector = 0;
+#endif
+	return ret;
+}
-- 
2.26.0

[PATCH v5 6/6] soc/tegra: Expose Boot Configuration Table via sysfs

[Dmitry Osipenko]

It's quite useful to have unencrypted BCT exposed to userspace for
debugging purposes, so let's expose it via sysfs.  The BCT data will be
present in '/sys/tegra/boot_config_table' binary file if BCT is available.

Suggested-by: Michał Mirosław <mirq-linux@rere.qmqm.pl>
Signed-off-by: Dmitry Osipenko <digetx@gmail.com>
---
 arch/arm/mach-tegra/tegra.c  |  4 +++
 drivers/soc/tegra/Makefile   |  1 +
 drivers/soc/tegra/bootdata.c | 51 ++++++++++++++++++++++++++++++++++++
 drivers/soc/tegra/common.c   | 17 ++++++++++++
 include/soc/tegra/bootdata.h |  2 ++
 include/soc/tegra/common.h   |  3 +++
 6 files changed, 78 insertions(+)
 create mode 100644 drivers/soc/tegra/bootdata.c

diff --git a/arch/arm/mach-tegra/tegra.c b/arch/arm/mach-tegra/tegra.c
index da6bcd85398b..5f40463f1b97 100644
--- a/arch/arm/mach-tegra/tegra.c
+++ b/arch/arm/mach-tegra/tegra.c
@@ -72,6 +72,7 @@ static void __init tegra_boot_config_table_init(void)
 	u32 iram_end   = TEGRA_IRAM_BASE + TEGRA_IRAM_SIZE;
 	u32 iram_start = TEGRA_IRAM_BASE;
 	u32 pt_addr, pt_size, bct_size;
+	void __iomem *bct_ptr;
 
 	t20_bit = IO_ADDRESS(TEGRA_IRAM_BASE);
 
@@ -90,6 +91,7 @@ static void __init tegra_boot_config_table_init(void)
 
 		pt_addr = t20_bct->partition_table_logical_sector_address;
 		pt_size = t20_bct->partition_table_num_logical_sectors;
+		bct_ptr = t20_bct;
 
 	} else if (of_machine_is_compatible("nvidia,tegra30")) {
 		bct_size = sizeof(*t30_bct);
@@ -106,12 +108,14 @@ static void __init tegra_boot_config_table_init(void)
 
 		pt_addr = t30_bct->partition_table_logical_sector_address;
 		pt_size = t30_bct->partition_table_num_logical_sectors;
+		bct_ptr = t30_bct;
 	} else {
 		return;
 	}
 
 	pr_info("%s: BCT found in IRAM\n", __func__);
 
+	tegra_bootdata_bct_setup(bct_ptr, bct_size);
 	tegra_partition_table_setup(pt_addr, pt_size);
 }
 
diff --git a/drivers/soc/tegra/Makefile b/drivers/soc/tegra/Makefile
index 9c809c1814bd..8be2bfb4d95d 100644
--- a/drivers/soc/tegra/Makefile
+++ b/drivers/soc/tegra/Makefile
@@ -1,6 +1,7 @@
 # SPDX-License-Identifier: GPL-2.0
 obj-y += fuse/
 
+obj-y += bootdata.o
 obj-y += common.o
 obj-$(CONFIG_SOC_TEGRA_FLOWCTRL) += flowctrl.o
 obj-$(CONFIG_SOC_TEGRA_PMC) += pmc.o
diff --git a/drivers/soc/tegra/bootdata.c b/drivers/soc/tegra/bootdata.c
new file mode 100644
index 000000000000..e18a27b74023
--- /dev/null
+++ b/drivers/soc/tegra/bootdata.c
@@ -0,0 +1,51 @@
+// SPDX-License-Identifier: GPL-2.0
+
+#include <linux/init.h>
+#include <linux/io.h>
+#include <linux/sizes.h>
+#include <linux/slab.h>
+#include <linux/sysfs.h>
+#include <linux/types.h>
+
+#include <soc/tegra/bootdata.h>
+#include <soc/tegra/common.h>
+
+/*
+ * spare_bct[] will be released once kernel is booted, hence not wasting
+ * kernel space if BCT is missing. The tegra_bct can't be allocated during
+ * of BCT setting up because it's too early for the slab allocator.
+ */
+static u8 spare_bct[SZ_8K] __initdata;
+static u8 *tegra_bct;
+
+static ssize_t boot_config_table_read(struct file *filp,
+				      struct kobject *kobj,
+				      struct bin_attribute *bin_attr,
+				      char *buf, loff_t off, size_t count)
+{
+	memcpy(buf, tegra_bct + off, count);
+	return count;
+}
+static BIN_ATTR_RO(boot_config_table, 0);
+
+static int __init tegra_bootdata_bct_sysfs_init(void)
+{
+	if (!bin_attr_boot_config_table.size)
+		return 0;
+
+	tegra_bct = kmalloc(GFP_KERNEL, bin_attr_boot_config_table.size);
+	if (!tegra_bct)
+		return -ENOMEM;
+
+	memcpy(tegra_bct, spare_bct, bin_attr_boot_config_table.size);
+
+	return sysfs_create_bin_file(tegra_soc_kobj,
+				     &bin_attr_boot_config_table);
+}
+late_initcall(tegra_bootdata_bct_sysfs_init)
+
+void __init tegra_bootdata_bct_setup(void __iomem *bct_ptr, size_t bct_size)
+{
+	memcpy_fromio(spare_bct, bct_ptr, bct_size);
+	bin_attr_boot_config_table.size = bct_size;
+}
diff --git a/drivers/soc/tegra/common.c b/drivers/soc/tegra/common.c
index 3dc54f59cafe..2b4b49eacb2e 100644
--- a/drivers/soc/tegra/common.c
+++ b/drivers/soc/tegra/common.c
@@ -3,10 +3,15 @@
  * Copyright (C) 2014 NVIDIA CORPORATION.  All rights reserved.
  */
 
+#include <linux/init.h>
+#include <linux/kernel.h>
 #include <linux/of.h>
+#include <linux/sysfs.h>
 
 #include <soc/tegra/common.h>
 
+struct kobject *tegra_soc_kobj;
+
 static const struct of_device_id tegra_machine_match[] = {
 	{ .compatible = "nvidia,tegra20", },
 	{ .compatible = "nvidia,tegra30", },
@@ -31,3 +36,15 @@ bool soc_is_tegra(void)
 
 	return match != NULL;
 }
+
+static int __init tegra_soc_sysfs_init(void)
+{
+	if (!soc_is_tegra())
+		return 0;
+
+	tegra_soc_kobj = kobject_create_and_add("tegra", NULL);
+	WARN_ON(!tegra_soc_kobj);
+
+	return 0;
+}
+arch_initcall(tegra_soc_sysfs_init)
diff --git a/include/soc/tegra/bootdata.h b/include/soc/tegra/bootdata.h
index 7be207cb2519..d5c7a251517d 100644
--- a/include/soc/tegra/bootdata.h
+++ b/include/soc/tegra/bootdata.h
@@ -43,4 +43,6 @@ struct tegra30_boot_config_table {
 	u32 unused_data[3];
 } __packed;
 
+void tegra_bootdata_bct_setup(void __iomem *bct_ptr, size_t bct_size);
+
 #endif /* __SOC_TEGRA_BOOTDATA_H__ */
diff --git a/include/soc/tegra/common.h b/include/soc/tegra/common.h
index 744280ecab5f..0bc11b45c98e 100644
--- a/include/soc/tegra/common.h
+++ b/include/soc/tegra/common.h
@@ -7,8 +7,11 @@
 #define __SOC_TEGRA_COMMON_H__
 
 #include <linux/types.h>
+#include <linux/sysfs.h>
 
 #ifdef CONFIG_ARCH_TEGRA
+extern struct kobject *tegra_soc_kobj;
+
 bool soc_is_tegra(void);
 #else
 static inline bool soc_is_tegra(void)
-- 
2.26.0

Re: [PATCH v5 4/6] partitions/efi: Support GPT entry lookup at a non-standard location

[Randy Dunlap]

On 5/16/20 8:36 AM, Dmitry Osipenko wrote:
> diff --git a/block/partitions/efi.c b/block/partitions/efi.c
> index b64bfdd4326c..3af4660bc11f 100644
> --- a/block/partitions/efi.c
> +++ b/block/partitions/efi.c
> @@ -621,6 +621,14 @@ static int find_valid_gpt(struct parsed_partitions *state, gpt_header **gpt,
>          if (!good_agpt && force_gpt)
>                  good_agpt = is_gpt_valid(state, lastlba, &agpt, &aptes);
>  
> +	/* The force_gpt_sector is used by NVIDIA Tegra partition parser in
> +	 * order to convey a non-standard location of the GPT entry for lookup.
> +	 * By default force_gpt_sector is set to 0 and has no effect.
> +	 */

Please fix the multi-line comment format as described in
Documentation/process/coding-style.rst.

> +	if (!good_agpt && force_gpt && state->force_gpt_sector)
> +		good_agpt = is_gpt_valid(state, state->force_gpt_sector,
> +					 &agpt, &aptes);
> +
>          /* The obviously unsuccessful case */
>          if (!good_pgpt && !good_agpt)
>                  goto fail;

thanks.
-- 
~Randy

Re: [PATCH v5 5/6] partitions/tegra: Support enforced GPT scanning

[Randy Dunlap]

On 5/16/20 8:36 AM, Dmitry Osipenko wrote:
> diff --git a/block/partitions/efi.c b/block/partitions/efi.c
> index 3af4660bc11f..4eb4496fbb1b 100644
> --- a/block/partitions/efi.c
> +++ b/block/partitions/efi.c
> @@ -98,6 +98,12 @@ static int force_gpt;
>  static int __init
>  force_gpt_fn(char *str)
>  {
> +	/* This check allows to parse "gpt gpt_sector=" properly since
> +	 * "gpt" overlaps with "gpt_sector", see tegra_gpt_sector_fn().
> +	 */

same here.

> +	if (force_gpt)
> +		return 0;
> +
>  	force_gpt = 1;
>  	return 1;
>  }


-- 
~Randy

Re: [PATCH v5 4/6] partitions/efi: Support GPT entry lookup at a non-standard location

[Dmitry Osipenko]

16.05.2020 18:51, Randy Dunlap пишет:
> On 5/16/20 8:36 AM, Dmitry Osipenko wrote:
>> diff --git a/block/partitions/efi.c b/block/partitions/efi.c
>> index b64bfdd4326c..3af4660bc11f 100644
>> --- a/block/partitions/efi.c
>> +++ b/block/partitions/efi.c
>> @@ -621,6 +621,14 @@ static int find_valid_gpt(struct parsed_partitions *state, gpt_header **gpt,
>>          if (!good_agpt && force_gpt)
>>                  good_agpt = is_gpt_valid(state, lastlba, &agpt, &aptes);
>>  
>> +	/* The force_gpt_sector is used by NVIDIA Tegra partition parser in
>> +	 * order to convey a non-standard location of the GPT entry for lookup.
>> +	 * By default force_gpt_sector is set to 0 and has no effect.
>> +	 */
> 
> Please fix the multi-line comment format as described in
> Documentation/process/coding-style.rst.
> 
>> +	if (!good_agpt && force_gpt && state->force_gpt_sector)
>> +		good_agpt = is_gpt_valid(state, state->force_gpt_sector,
>> +					 &agpt, &aptes);
>> +
>>          /* The obviously unsuccessful case */
>>          if (!good_pgpt && !good_agpt)
>>                  goto fail;
> 
> thanks.
> 

Hello Randy,

I know that it's not a proper kernel-style formatting, but that's the
style used by the whole efi.c source code and I wanted to maintain the
same style, for consistency. Of course I can change to a proper style if
it's more desirable than the consistency. Thank you for the comment!

Re: [PATCH v5 4/6] partitions/efi: Support GPT entry lookup at a non-standard location

[Randy Dunlap]

On 5/16/20 9:50 AM, Dmitry Osipenko wrote:
> 16.05.2020 18:51, Randy Dunlap пишет:
>> On 5/16/20 8:36 AM, Dmitry Osipenko wrote:
>>> diff --git a/block/partitions/efi.c b/block/partitions/efi.c
>>> index b64bfdd4326c..3af4660bc11f 100644
>>> --- a/block/partitions/efi.c
>>> +++ b/block/partitions/efi.c
>>> @@ -621,6 +621,14 @@ static int find_valid_gpt(struct parsed_partitions *state, gpt_header **gpt,
>>>          if (!good_agpt && force_gpt)
>>>                  good_agpt = is_gpt_valid(state, lastlba, &agpt, &aptes);
>>>  
>>> +	/* The force_gpt_sector is used by NVIDIA Tegra partition parser in
>>> +	 * order to convey a non-standard location of the GPT entry for lookup.
>>> +	 * By default force_gpt_sector is set to 0 and has no effect.
>>> +	 */
>>
>> Please fix the multi-line comment format as described in
>> Documentation/process/coding-style.rst.
>>
>>> +	if (!good_agpt && force_gpt && state->force_gpt_sector)
>>> +		good_agpt = is_gpt_valid(state, state->force_gpt_sector,
>>> +					 &agpt, &aptes);
>>> +
>>>          /* The obviously unsuccessful case */
>>>          if (!good_pgpt && !good_agpt)
>>>                  goto fail;
>>
>> thanks.
>>
> 
> Hello Randy,
> 
> I know that it's not a proper kernel-style formatting, but that's the
> style used by the whole efi.c source code and I wanted to maintain the
> same style, for consistency. Of course I can change to a proper style if
> it's more desirable than the consistency. Thank you for the comment!
> 

too bad. Sorry to hear that.
It should have been "fixed" much earlier.
It's probably too late now.

-- 
~Randy

Re: [PATCH v5 4/6] partitions/efi: Support GPT entry lookup at a non-standard location

[Dmitry Osipenko]

16.05.2020 19:58, Randy Dunlap пишет:
> On 5/16/20 9:50 AM, Dmitry Osipenko wrote:
>> 16.05.2020 18:51, Randy Dunlap пишет:
>>> On 5/16/20 8:36 AM, Dmitry Osipenko wrote:
>>>> diff --git a/block/partitions/efi.c b/block/partitions/efi.c
>>>> index b64bfdd4326c..3af4660bc11f 100644
>>>> --- a/block/partitions/efi.c
>>>> +++ b/block/partitions/efi.c
>>>> @@ -621,6 +621,14 @@ static int find_valid_gpt(struct parsed_partitions *state, gpt_header **gpt,
>>>>          if (!good_agpt && force_gpt)
>>>>                  good_agpt = is_gpt_valid(state, lastlba, &agpt, &aptes);
>>>>  
>>>> +	/* The force_gpt_sector is used by NVIDIA Tegra partition parser in
>>>> +	 * order to convey a non-standard location of the GPT entry for lookup.
>>>> +	 * By default force_gpt_sector is set to 0 and has no effect.
>>>> +	 */
>>>
>>> Please fix the multi-line comment format as described in
>>> Documentation/process/coding-style.rst.
>>>
>>>> +	if (!good_agpt && force_gpt && state->force_gpt_sector)
>>>> +		good_agpt = is_gpt_valid(state, state->force_gpt_sector,
>>>> +					 &agpt, &aptes);
>>>> +
>>>>          /* The obviously unsuccessful case */
>>>>          if (!good_pgpt && !good_agpt)
>>>>                  goto fail;
>>>
>>> thanks.
>>>
>>
>> Hello Randy,
>>
>> I know that it's not a proper kernel-style formatting, but that's the
>> style used by the whole efi.c source code and I wanted to maintain the
>> same style, for consistency. Of course I can change to a proper style if
>> it's more desirable than the consistency. Thank you for the comment!
>>
> 
> too bad. Sorry to hear that.
> It should have been "fixed" much earlier.
> It's probably too late now.

Actually, I now see that there is a mix of different comment styles in
the efi.c code. So it should be fine to use the proper style, I'll
change it in v6.

I don't think it's too late, it's never late to make a correction :)
There are some other coding style problems in the efi.c that won't hurt
to fix, I may take a look at fixing them later on.

Re: [PATCH v5 4/6] partitions/efi: Support GPT entry lookup at a non-standard location

[Randy Dunlap]

On 5/16/20 5:11 PM, Dmitry Osipenko wrote:
> 16.05.2020 19:58, Randy Dunlap пишет:
>> On 5/16/20 9:50 AM, Dmitry Osipenko wrote:
>>> 16.05.2020 18:51, Randy Dunlap пишет:
>>>> On 5/16/20 8:36 AM, Dmitry Osipenko wrote:
>>>>> diff --git a/block/partitions/efi.c b/block/partitions/efi.c
>>>>> index b64bfdd4326c..3af4660bc11f 100644
>>>>> --- a/block/partitions/efi.c
>>>>> +++ b/block/partitions/efi.c
>>>>> @@ -621,6 +621,14 @@ static int find_valid_gpt(struct parsed_partitions *state, gpt_header **gpt,
>>>>>          if (!good_agpt && force_gpt)
>>>>>                  good_agpt = is_gpt_valid(state, lastlba, &agpt, &aptes);
>>>>>  
>>>>> +	/* The force_gpt_sector is used by NVIDIA Tegra partition parser in
>>>>> +	 * order to convey a non-standard location of the GPT entry for lookup.
>>>>> +	 * By default force_gpt_sector is set to 0 and has no effect.
>>>>> +	 */
>>>>
>>>> Please fix the multi-line comment format as described in
>>>> Documentation/process/coding-style.rst.
>>>>
>>>>> +	if (!good_agpt && force_gpt && state->force_gpt_sector)
>>>>> +		good_agpt = is_gpt_valid(state, state->force_gpt_sector,
>>>>> +					 &agpt, &aptes);
>>>>> +
>>>>>          /* The obviously unsuccessful case */
>>>>>          if (!good_pgpt && !good_agpt)
>>>>>                  goto fail;
>>>>
>>>> thanks.
>>>>
>>>
>>> Hello Randy,
>>>
>>> I know that it's not a proper kernel-style formatting, but that's the
>>> style used by the whole efi.c source code and I wanted to maintain the
>>> same style, for consistency. Of course I can change to a proper style if
>>> it's more desirable than the consistency. Thank you for the comment!
>>>
>>
>> too bad. Sorry to hear that.
>> It should have been "fixed" much earlier.
>> It's probably too late now.
> 
> Actually, I now see that there is a mix of different comment styles in
> the efi.c code. So it should be fine to use the proper style, I'll
> change it in v6.
> 
> I don't think it's too late, it's never late to make a correction :)
> There are some other coding style problems in the efi.c that won't hurt
> to fix, I may take a look at fixing them later on.
> 

OK, great. Thanks.

-- 
~Randy

Re: [PATCH v5 6/6] soc/tegra: Expose Boot Configuration Table via sysfs

[Michał Mirosław]

On Sat, May 16, 2020 at 06:36:44PM +0300, Dmitry Osipenko wrote:
> It's quite useful to have unencrypted BCT exposed to userspace for
> debugging purposes, so let's expose it via sysfs.  The BCT data will be
> present in '/sys/tegra/boot_config_table' binary file if BCT is available.
[...]
> +/*
> + * spare_bct[] will be released once kernel is booted, hence not wasting
> + * kernel space if BCT is missing. The tegra_bct can't be allocated during
> + * of BCT setting up because it's too early for the slab allocator.
> + */
> +static u8 spare_bct[SZ_8K] __initdata;
> +static u8 *tegra_bct;
> +
> +static ssize_t boot_config_table_read(struct file *filp,
> +				      struct kobject *kobj,
> +				      struct bin_attribute *bin_attr,
> +				      char *buf, loff_t off, size_t count)
> +{
> +	memcpy(buf, tegra_bct + off, count);
> +	return count;
> +}
> +static BIN_ATTR_RO(boot_config_table, 0);
> +
> +static int __init tegra_bootdata_bct_sysfs_init(void)
> +{
> +	if (!bin_attr_boot_config_table.size)
> +		return 0;
> +
> +	tegra_bct = kmalloc(GFP_KERNEL, bin_attr_boot_config_table.size);
> +	if (!tegra_bct)
> +		return -ENOMEM;
> +
> +	memcpy(tegra_bct, spare_bct, bin_attr_boot_config_table.size);
> +
> +	return sysfs_create_bin_file(tegra_soc_kobj,
> +				     &bin_attr_boot_config_table);

Should we consider freeing the table if sysfs registration fails?

> +}
> +late_initcall(tegra_bootdata_bct_sysfs_init)
> +
> +void __init tegra_bootdata_bct_setup(void __iomem *bct_ptr, size_t bct_size)
> +{
> +	memcpy_fromio(spare_bct, bct_ptr, bct_size);
> +	bin_attr_boot_config_table.size = bct_size;
[...]

The size isn't checked anywhere. How the maximum is obtained? At least
a comment would be good if the 8k limit is guaranteed by other means.

Best Regards,
Michał Mirosław

Re: [PATCH v5 6/6] soc/tegra: Expose Boot Configuration Table via sysfs

[Dmitry Osipenko]

18.05.2020 10:45, Michał Mirosław пишет:
> On Sat, May 16, 2020 at 06:36:44PM +0300, Dmitry Osipenko wrote:
>> It's quite useful to have unencrypted BCT exposed to userspace for
>> debugging purposes, so let's expose it via sysfs.  The BCT data will be
>> present in '/sys/tegra/boot_config_table' binary file if BCT is available.
> [...]
>> +/*
>> + * spare_bct[] will be released once kernel is booted, hence not wasting
>> + * kernel space if BCT is missing. The tegra_bct can't be allocated during
>> + * of BCT setting up because it's too early for the slab allocator.
>> + */
>> +static u8 spare_bct[SZ_8K] __initdata;
>> +static u8 *tegra_bct;
>> +
>> +static ssize_t boot_config_table_read(struct file *filp,
>> +				      struct kobject *kobj,
>> +				      struct bin_attribute *bin_attr,
>> +				      char *buf, loff_t off, size_t count)
>> +{
>> +	memcpy(buf, tegra_bct + off, count);
>> +	return count;
>> +}
>> +static BIN_ATTR_RO(boot_config_table, 0);
>> +
>> +static int __init tegra_bootdata_bct_sysfs_init(void)
>> +{
>> +	if (!bin_attr_boot_config_table.size)
>> +		return 0;
>> +
>> +	tegra_bct = kmalloc(GFP_KERNEL, bin_attr_boot_config_table.size);
>> +	if (!tegra_bct)
>> +		return -ENOMEM;
>> +
>> +	memcpy(tegra_bct, spare_bct, bin_attr_boot_config_table.size);
>> +
>> +	return sysfs_create_bin_file(tegra_soc_kobj,
>> +				     &bin_attr_boot_config_table);
> 
> Should we consider freeing the table if sysfs registration fails?

This is a good suggestion, thank you :)

>> +}
>> +late_initcall(tegra_bootdata_bct_sysfs_init)
>> +
>> +void __init tegra_bootdata_bct_setup(void __iomem *bct_ptr, size_t bct_size)
>> +{
>> +	memcpy_fromio(spare_bct, bct_ptr, bct_size);
>> +	bin_attr_boot_config_table.size = bct_size;
> [...]
> 
> The size isn't checked anywhere. How the maximum is obtained? At least
> a comment would be good if the 8k limit is guaranteed by other means.

Yes, I can add a clarifying comment about the 8K. It also should be
possible to create a union of T20/T30 BCT structs and then use use
sizeof(bct_union). I'll consider this change for the next version, thank
you for the suggestion.