From: Heinrich Schuchardt <xypron.glpk@gmx.de>
To: Paul Barbieri <plb365@gmail.com>
Cc: AKASHI Takahiro <takahiro.akashi@linaro.org>, u-boot@lists.denx.de
Subject: Re: [PATCH v2] EFI: Fix ReadBlocks API reading incorrect sector for UCLASS_PARTITION devices
Date: Thu, 30 Jun 2022 10:38:44 +0200 [thread overview]
Message-ID: <90538c2d-2fa2-3c7d-7e4f-ec26164cf06e@gmx.de> (raw)
In-Reply-To: <20220629174413.2053-1-plb365@gmail.com>
On 6/29/22 19:44, Paul Barbieri wrote:
> The requsted partition disk sector incorrectly has the parition start
> sector added in twice for UCLASS_PARTITION devices. The efi_disk_rw_blocks()
> routine adds the diskobj->offset to the requested lba. When the device
> is a UCLASS_PARTITION, the dev_read() or dev_write() routine is called
> which adds part-gpt_part_info.start. This causes I/O to the wrong sector.
>
> Takahiro Akashi suggested removing the offset field from the efi_disk_obj
> structure since disk-uclass.c handles the partition start biasing. Device
> types other than UCLASS_PARTITION set the diskobj->offset field to zero
> which makes the field unnecessary. This change removes the offset field
> from the structure and removes all references from the code which is
> isolated to the lib/efi_loader/efi_disk.c module.
>
> This change also adds a test for the EFI ReadBlocks() API in the EFI
> selftest code. There is already a test for reading a FAT file. The new
> test uses ReadBlocks() to read the same "disk" block and compare it to
> the data read from the file system API.
>
> I verified that the test was using the correct code path by defining DEBUG
> in lib/efi_loader/efi_disk.c This showed that when the test was run,
> efi_disk_read_blocks() was executed and it called efi_disk_rw_blocks().
> Adding a little extra debug, I saw that efi_disk_rw_blocks() called dev_read().
> I also added a debug print in the read_blocks() routine of
> efi_selftest_block_device.c. One thing to note, if I moved the read_blocks test
> before the file system test, I would see the read_blocks() routine print
> in the middle of the efi_disk_rw_blocks() routine. If I put the test after
> the file system test, I don't see read_blocks() get called. I assume this
> is because dev_read() is able to get the block from the cache.
>
>
> Signed-Off-by: Paul Barbieri <plb365@gmail.com>
> Cc: Heinrich Schuchardt <xypron.glpk@gmx.de>
> Cc: AKASHI Takahiro <takahiro.akashi@linaro.org>
> ---
> lib/efi_loader/efi_disk.c | 8 +------
> lib/efi_selftest/efi_selftest_block_device.c | 23 ++++++++++++++++++++
> 2 files changed, 24 insertions(+), 7 deletions(-)
>
> diff --git a/lib/efi_loader/efi_disk.c b/lib/efi_loader/efi_disk.c
> index 1e82f52dc0..1d700b2a6b 100644
> --- a/lib/efi_loader/efi_disk.c
> +++ b/lib/efi_loader/efi_disk.c
> @@ -35,7 +35,6 @@ const efi_guid_t efi_system_partition_guid = PARTITION_SYSTEM_GUID;
> * @dp: device path to the block device
> * @part: partition
> * @volume: simple file system protocol of the partition
> - * @offset: offset into disk for simple partition
> * @dev: associated DM device
> */
> struct efi_disk_obj {
> @@ -47,7 +46,6 @@ struct efi_disk_obj {
> struct efi_device_path *dp;
> unsigned int part;
> struct efi_simple_file_system_protocol *volume;
> - lbaint_t offset;
> struct udevice *dev; /* TODO: move it to efi_object */
> };
>
> @@ -117,7 +115,6 @@ static efi_status_t efi_disk_rw_blocks(struct efi_block_io *this,
> diskobj = container_of(this, struct efi_disk_obj, ops);
> blksz = diskobj->media.block_size;
> blocks = buffer_size / blksz;
> - lba += diskobj->offset;
>
> EFI_PRINT("blocks=%x lba=%llx blksz=%x dir=%d\n",
> blocks, lba, blksz, direction);
> @@ -440,13 +437,11 @@ static efi_status_t efi_disk_add_dev(
>
> diskobj->dp = efi_dp_append_node(dp_parent, node);
> efi_free_pool(node);
> - diskobj->offset = part_info->start;
> diskobj->media.last_block = part_info->size - 1;
> if (part_info->bootable & PART_EFI_SYSTEM_PARTITION)
> guid = &efi_system_partition_guid;
> } else {
> diskobj->dp = efi_dp_from_part(desc, part);
> - diskobj->offset = 0;
> diskobj->media.last_block = desc->lba - 1;
> }
> diskobj->part = part;
> @@ -501,12 +496,11 @@ static efi_status_t efi_disk_add_dev(
> *disk = diskobj;
>
> EFI_PRINT("BlockIO: part %u, present %d, logical %d, removable %d"
> - ", offset " LBAF ", last_block %llu\n",
> + ", last_block %llu\n",
> diskobj->part,
> diskobj->media.media_present,
> diskobj->media.logical_partition,
> diskobj->media.removable_media,
> - diskobj->offset,
> diskobj->media.last_block);
>
> /* Store first EFI system partition */
> diff --git a/lib/efi_selftest/efi_selftest_block_device.c b/lib/efi_selftest/efi_selftest_block_device.c
> index 60fa655766..7b50c829ad 100644
> --- a/lib/efi_selftest/efi_selftest_block_device.c
> +++ b/lib/efi_selftest/efi_selftest_block_device.c
> @@ -11,6 +11,7 @@
> * ConnectController is used to setup partitions and to install the simple
> * file protocol.
> * A known file is read from the file system and verified.
> + * Test that the read_blocks API correctly reads a block from the device.
> */
>
> #include <efi_selftest.h>
> @@ -312,6 +313,9 @@ static int execute(void)
> char buf[16] __aligned(ARCH_DMA_MINALIGN);
> u32 part1_size;
> u64 pos;
> + char block[2 * (1 << LB_BLOCK_SIZE)];
> + char *block_io_aligned;
> + u32 align;
>
> /* Connect controller to virtual disk */
> ret = boottime->connect_controller(disk_handle, NULL, NULL, 1);
> @@ -449,6 +453,25 @@ static int execute(void)
> return EFI_ST_FAILURE;
> }
>
> + /* Test read_blocks() can read same file data. */
> + boottime->set_mem(block, sizeof(block), 0);
> + align = block_io_protocol->media->io_align;
> + block_io_aligned = (char *)(((uintptr_t)block + align-1) & ~(align-1));
> + ret = block_io_protocol->read_blocks(block_io_protocol,
> + block_io_protocol->media->media_id,
> + (0x5000 >> LB_BLOCK_SIZE) - 1,
This code really needs a comment. E.g.
In the test data
* the partition starts at block 1
* the file hello.txt with content 'Hello world!' is located at 0x5000 of
the disk
Here we read block 0x27 (offset 0x4e00 of the partition) and expect the
string 'Hello world!' to be at the start of the block.
Best regards
Heinrich
> + block_io_protocol->media->block_size,
> + block_io_aligned);
> + if (ret != EFI_SUCCESS) {
> + efi_st_error("ReadBlocks failed\n");
> + return EFI_ST_FAILURE;
> + }
> +
> + if (memcmp(&block_io_aligned[1], buf, 11)) {
> + efi_st_error("Unexpected block content\n");
> + return EFI_ST_FAILURE;
> + }
> +
> #ifdef CONFIG_FAT_WRITE
> /* Write file */
> ret = root->open(root, &file, u"u-boot.txt", EFI_FILE_MODE_READ |
next prev parent reply other threads:[~2022-06-30 8:38 UTC|newest]
Thread overview: 8+ messages / expand[flat|nested] mbox.gz Atom feed top
2022-06-29 1:02 [PATCH] EFI: Fix ReadBlocks API reading incorrect sector for, UCLASS_PARTITION devices Paul Barbieri
2022-06-29 4:43 ` AKASHI Takahiro
2022-06-29 4:58 ` Heinrich Schuchardt
2022-06-29 17:44 ` [PATCH v2] EFI: Fix ReadBlocks API reading incorrect sector for " Paul Barbieri
2022-06-30 8:38 ` Heinrich Schuchardt [this message]
2022-06-30 11:02 ` [PATCH v3] " Paul Barbieri
2022-07-02 8:18 ` Heinrich Schuchardt
2022-07-02 11:20 ` [PATCH v4] " Paul Barbieri
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
Avoid top-posting and favor interleaved quoting:
https://en.wikipedia.org/wiki/Posting_style#Interleaved_style
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to=90538c2d-2fa2-3c7d-7e4f-ec26164cf06e@gmx.de \
--to=xypron.glpk@gmx.de \
--cc=plb365@gmail.com \
--cc=takahiro.akashi@linaro.org \
--cc=u-boot@lists.denx.de \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox