On 08/30/2017 06:37 AM, Heinrich Schuchardt wrote:

On 08/30/2017 03:54 AM, Bin Meng wrote:

...

Hi Heinrich,

On Wed, Aug 30, 2017 at 4:26 AM, Heinrich Schuchardt xypron.glpk@gmx.de wrote:

...

Hello Simon,

U-Boot HEAD qemu-86_defconfig cannot discover an IDE disk with one FAT partition in qemu-system-x86_64.

By bisection I found this patch.

b7c6baef2891ce8978cbfddb66e944943473ac21 x86: Convert MMC to driver model

With this patch I get

IDE: Bus 0: OK Bus 1: OK Device 0: Model: QEMU HARDDISK Firm: 2.5+ Ser#: QM00001 Type: Hard Disk Supports 48-bit addressing Capacity: 128.0 MB = 0.1 GB (262144 x 512) ** Can't read Driver Desriptor Block ** Device 1: not available Device 2: Model: QEMU Firm: 2.5+ Ser#: QEMU DVD-ROM Type: Removable CD ROM Capacity: not available Device 3: not available

=> ide info =>

Without the patch I get=> ide info Device 0: Model: QEMU HARDDISK Firm: 2.5+ Ser#: QM00001 Type: Hard Disk Supports 48-bit addressing Capacity: 128.0 MB = 0.1 GB (262144 x 512) Device 2: Model: QEMU Firm: 2.5+ Ser#: QEMU DVD-ROM Type: Removable CD ROM Capacity: not available

I think we observe two independent errors here:

• The hard disk Device 0 is not read.
• The ide command stops at the first device that is not available.

I guess only the first is caused by your patch.

Both logs look fine to me. The "Can't read Driver Desriptor Block" comes from part_mac.c. Did you verify the actual IDE read/write fails with current HEAD?

Regards, Bin

Hello Bin,

I have not checked block level read but used the shell commands for testing.

Before the patch I can read the directory of the drive:

=> ide info Device 0: Model: QEMU HARDDISK Firm: 2.5+ Ser#: QM00001 Type: Hard Disk Supports 48-bit addressing Capacity: 128.0 MB = 0.1 GB (262144 x 512) Device 2: Model: QEMU Firm: 2.5+ Ser#: QEMU DVD-ROM Type: Removable CD ROM Capacity: not available => fat2ls ide 0:1 Unknown command 'fat2ls' - try 'help' => fatls ide 0:1 164768 snp.efi 0 file1 0 file2

After the patch (including HEAD) I cannot read the directory and cannot load the file snp.efi either:

=> ide info => fatls ide 0:1 ** Bad device ide 0 ** => => fatls mmc 0:1 ** Bad device mmc 0 **

In both cases I have loaded the same image with:

export BUILD_ROM=y make distclean && make qemu-x86_defconfig && make -j6

qemu-system-x86_64 -m 1G -bios u-boot.rom -nographic \ -netdev \ user,id=eth0,tftp=tftp,net=192.168.76.0/24,dhcpstart=192.168.76.9 \ -device e1000,netdev=eth0 -machine pc-i440fx-2.8 -hda img

Best regards

Heinrich

Hello Bin, hello Simon,

I think the bug is in functions ide_init (drivers/block/ide.c).

Platform X86 implies CONFIG_BLK=y.

So we should initialize ide_dev_desc[i].bdev.

We don't, so blk_dread fails after finding no read operation with ENOSYS when called from part_test_dos.

The following comment in include/blk.h confirms that bdev has to be filled:

#if CONFIG_IS_ENABLED(BLK) /* * For now we have a few functions which take struct blk_desc as a * parameter. This field allows them to look up the associated * device. Once these functions are removed we can drop this field. */ struct udevice *bdev; #else

I assume that we have the same issue in in __sata_initialize (drivers/ata/sata.c) but have not tested.

Best regards

Heinrich

Hi Bin,

On 31 August 2017 at 10:53, Bin Meng bmeng.cn@gmail.com wrote:

Hi Heinrich,

On Thu, Aug 31, 2017 at 5:19 AM, Heinrich Schuchardt xypron.glpk@gmx.de wrote:

...

On 08/30/2017 06:37 AM, Heinrich Schuchardt wrote:

...

On 08/30/2017 03:54 AM, Bin Meng wrote:

...

Hi Heinrich,

On Wed, Aug 30, 2017 at 4:26 AM, Heinrich Schuchardt xypron.glpk@gmx.de wrote:

...

Hello Simon,

U-Boot HEAD qemu-86_defconfig cannot discover an IDE disk with one FAT partition in qemu-system-x86_64.

By bisection I found this patch.

b7c6baef2891ce8978cbfddb66e944943473ac21 x86: Convert MMC to driver model

With this patch I get

IDE: Bus 0: OK Bus 1: OK Device 0: Model: QEMU HARDDISK Firm: 2.5+ Ser#: QM00001 Type: Hard Disk Supports 48-bit addressing Capacity: 128.0 MB = 0.1 GB (262144 x 512) ** Can't read Driver Desriptor Block ** Device 1: not available Device 2: Model: QEMU Firm: 2.5+ Ser#: QEMU DVD-ROM Type: Removable CD ROM Capacity: not available Device 3: not available

=> ide info =>

Without the patch I get=> ide info Device 0: Model: QEMU HARDDISK Firm: 2.5+ Ser#: QM00001 Type: Hard Disk Supports 48-bit addressing Capacity: 128.0 MB = 0.1 GB (262144 x 512) Device 2: Model: QEMU Firm: 2.5+ Ser#: QEMU DVD-ROM Type: Removable CD ROM Capacity: not available

I think we observe two independent errors here:

• The hard disk Device 0 is not read.
• The ide command stops at the first device that is not available.

I guess only the first is caused by your patch.

Both logs look fine to me. The "Can't read Driver Desriptor Block" comes from part_mac.c. Did you verify the actual IDE read/write fails with current HEAD?

Regards, Bin

Hello Bin,

I have not checked block level read but used the shell commands for testing.

Before the patch I can read the directory of the drive:

=> ide info Device 0: Model: QEMU HARDDISK Firm: 2.5+ Ser#: QM00001 Type: Hard Disk Supports 48-bit addressing Capacity: 128.0 MB = 0.1 GB (262144 x 512) Device 2: Model: QEMU Firm: 2.5+ Ser#: QEMU DVD-ROM Type: Removable CD ROM Capacity: not available => fat2ls ide 0:1 Unknown command 'fat2ls' - try 'help' => fatls ide 0:1 164768 snp.efi 0 file1 0 file2

After the patch (including HEAD) I cannot read the directory and cannot load the file snp.efi either:

=> ide info => fatls ide 0:1 ** Bad device ide 0 ** => => fatls mmc 0:1 ** Bad device mmc 0 **

In both cases I have loaded the same image with:

export BUILD_ROM=y make distclean && make qemu-x86_defconfig && make -j6

qemu-system-x86_64 -m 1G -bios u-boot.rom -nographic \ -netdev \ user,id=eth0,tftp=tftp,net=192.168.76.0/24,dhcpstart=192.168.76.9 \ -device e1000,netdev=eth0 -machine pc-i440fx-2.8 -hda img

Best regards

Heinrich

Hello Bin, hello Simon,

I think the bug is in functions ide_init (drivers/block/ide.c).

Platform X86 implies CONFIG_BLK=y.

So we should initialize ide_dev_desc[i].bdev.

We don't, so blk_dread fails after finding no read operation with ENOSYS when called from part_test_dos.

The following comment in include/blk.h confirms that bdev has to be filled:

#if CONFIG_IS_ENABLED(BLK) /*

• For now we have a few functions which take struct blk_desc as a
• parameter. This field allows them to look up the associated
• device. Once these functions are removed we can drop this field.

*/ struct udevice *bdev; #else

I assume that we have the same issue in in __sata_initialize (drivers/ata/sata.c) but have not tested.

Thanks for the testing!

Simon, are you going to fix this?

I am not going to race you to it, if you are thinking of fixing it. I am back from travels in a few days but have a busy week ahead and the release is imminent :-(

Regards, Simon

On 08/31/2017 02:55 PM, Bin Meng wrote:

Hi Simon,

On Thu, Aug 31, 2017 at 8:52 PM, Simon Glass sjg@chromium.org wrote:

...

Hi Bin,

On 31 August 2017 at 10:53, Bin Meng bmeng.cn@gmail.com wrote:

...

Hi Heinrich,

On Thu, Aug 31, 2017 at 5:19 AM, Heinrich Schuchardt xypron.glpk@gmx.de wrote:

...

On 08/30/2017 06:37 AM, Heinrich Schuchardt wrote:

...

On 08/30/2017 03:54 AM, Bin Meng wrote:

...

Hi Heinrich,

On Wed, Aug 30, 2017 at 4:26 AM, Heinrich Schuchardt xypron.glpk@gmx.de wrote: > Hello Simon, > > U-Boot HEAD qemu-86_defconfig cannot discover an IDE disk with one FAT > partition in qemu-system-x86_64. > > By bisection I found this patch. > > b7c6baef2891ce8978cbfddb66e944943473ac21 > x86: Convert MMC to driver model > > With this patch I get > > IDE: Bus 0: OK Bus 1: OK > Device 0: Model: QEMU HARDDISK Firm: 2.5+ Ser#: QM00001 > Type: Hard Disk > Supports 48-bit addressing > Capacity: 128.0 MB = 0.1 GB (262144 x 512) > ** Can't read Driver Desriptor Block ** > Device 1: not available > Device 2: Model: QEMU Firm: 2.5+ Ser#: QEMU DVD-ROM > Type: Removable CD ROM > Capacity: not available > Device 3: not available > > => ide info > => > > Without the patch I get=> ide info > Device 0: Model: QEMU HARDDISK Firm: 2.5+ Ser#: QM00001 > Type: Hard Disk > Supports 48-bit addressing > Capacity: 128.0 MB = 0.1 GB (262144 x 512) > Device 2: Model: QEMU Firm: 2.5+ Ser#: QEMU DVD-ROM > Type: Removable CD ROM > Capacity: not available > > I think we observe two independent errors here: > > - The hard disk Device 0 is not read. > - The ide command stops at the first device that is not available. > > I guess only the first is caused by your patch.

Both logs look fine to me. The "Can't read Driver Desriptor Block" comes from part_mac.c. Did you verify the actual IDE read/write fails with current HEAD?

Regards, Bin

Hello Bin,

I have not checked block level read but used the shell commands for testing.

Before the patch I can read the directory of the drive:

=> ide info Device 0: Model: QEMU HARDDISK Firm: 2.5+ Ser#: QM00001 Type: Hard Disk Supports 48-bit addressing Capacity: 128.0 MB = 0.1 GB (262144 x 512) Device 2: Model: QEMU Firm: 2.5+ Ser#: QEMU DVD-ROM Type: Removable CD ROM Capacity: not available => fat2ls ide 0:1 Unknown command 'fat2ls' - try 'help' => fatls ide 0:1 164768 snp.efi 0 file1 0 file2

After the patch (including HEAD) I cannot read the directory and cannot load the file snp.efi either:

=> ide info => fatls ide 0:1 ** Bad device ide 0 ** => => fatls mmc 0:1 ** Bad device mmc 0 **

In both cases I have loaded the same image with:

export BUILD_ROM=y make distclean && make qemu-x86_defconfig && make -j6

qemu-system-x86_64 -m 1G -bios u-boot.rom -nographic \ -netdev \ user,id=eth0,tftp=tftp,net=192.168.76.0/24,dhcpstart=192.168.76.9 \ -device e1000,netdev=eth0 -machine pc-i440fx-2.8 -hda img

Best regards

Heinrich

Hello Bin, hello Simon,

I think the bug is in functions ide_init (drivers/block/ide.c).

Platform X86 implies CONFIG_BLK=y.

So we should initialize ide_dev_desc[i].bdev.

We don't, so blk_dread fails after finding no read operation with ENOSYS when called from part_test_dos.

The following comment in include/blk.h confirms that bdev has to be filled:

#if CONFIG_IS_ENABLED(BLK) /*

• For now we have a few functions which take struct blk_desc as a
• parameter. This field allows them to look up the associated
• device. Once these functions are removed we can drop this field.

*/ struct udevice *bdev; #else

I assume that we have the same issue in in __sata_initialize (drivers/ata/sata.c) but have not tested.

Thanks for the testing!

Simon, are you going to fix this?

I am not going to race you to it, if you are thinking of fixing it. I am back from travels in a few days but have a busy week ahead and the release is imminent :-(

Or maybe Heinrich, do you plan to work on a fix? If not, I will put it on my todo list.

Regards, Bin

Hello hello Bin,

I am aware that this bug is release critical. But unfortunately I am departing on vacation tomorrow.

I guess there should be CONFIG_SYS_IDE_MAXDEVICE times an U_BOOT_IDE_DEVICE(ide_disk[i]) which should refer to driver U_BOOT_DRIVER(ide_blk).

Just as prove of concept I append a patch to this mail. With the patch the IDE partitions are discovered in disk/part.c. But they are not usable to access the disk because the devices are not registered in a uclass.

Hopefully you will find the time to fix the problem.

Regards

Heinrich

On 09/04/2017 05:59 AM, Bin Meng wrote:

Hi Heinrich,

On Fri, Sep 1, 2017 at 1:24 AM, Heinrich Schuchardt xypron.glpk@gmx.de wrote:

...

On 08/31/2017 02:55 PM, Bin Meng wrote:

...

Hi Simon,

On Thu, Aug 31, 2017 at 8:52 PM, Simon Glass sjg@chromium.org wrote:

...

Hi Bin,

On 31 August 2017 at 10:53, Bin Meng bmeng.cn@gmail.com wrote:

...

Hi Heinrich,

On Thu, Aug 31, 2017 at 5:19 AM, Heinrich Schuchardt xypron.glpk@gmx.de wrote:

...

On 08/30/2017 06:37 AM, Heinrich Schuchardt wrote: > > > On 08/30/2017 03:54 AM, Bin Meng wrote: >> Hi Heinrich, >> >> On Wed, Aug 30, 2017 at 4:26 AM, Heinrich Schuchardt xypron.glpk@gmx.de wrote: >>> Hello Simon, >>> >>> U-Boot HEAD qemu-86_defconfig cannot discover an IDE disk with one FAT >>> partition in qemu-system-x86_64. >>> >>> By bisection I found this patch. >>> >>> b7c6baef2891ce8978cbfddb66e944943473ac21 >>> x86: Convert MMC to driver model >>> >>> With this patch I get >>> >>> IDE: Bus 0: OK Bus 1: OK >>> Device 0: Model: QEMU HARDDISK Firm: 2.5+ Ser#: QM00001 >>> Type: Hard Disk >>> Supports 48-bit addressing >>> Capacity: 128.0 MB = 0.1 GB (262144 x 512) >>> ** Can't read Driver Desriptor Block ** >>> Device 1: not available >>> Device 2: Model: QEMU Firm: 2.5+ Ser#: QEMU DVD-ROM >>> Type: Removable CD ROM >>> Capacity: not available >>> Device 3: not available >>> >>> => ide info >>> => >>> >>> Without the patch I get=> ide info >>> Device 0: Model: QEMU HARDDISK Firm: 2.5+ Ser#: QM00001 >>> Type: Hard Disk >>> Supports 48-bit addressing >>> Capacity: 128.0 MB = 0.1 GB (262144 x 512) >>> Device 2: Model: QEMU Firm: 2.5+ Ser#: QEMU DVD-ROM >>> Type: Removable CD ROM >>> Capacity: not available >>> >>> I think we observe two independent errors here: >>> >>> - The hard disk Device 0 is not read. >>> - The ide command stops at the first device that is not available. >>> >>> I guess only the first is caused by your patch. >> >> Both logs look fine to me. The "Can't read Driver Desriptor Block" >> comes from part_mac.c. Did you verify the actual IDE read/write fails >> with current HEAD? >> >> Regards, >> Bin >> > > Hello Bin, > > I have not checked block level read but used the shell commands for testing. > > Before the patch I can read the directory of the drive: > > => ide info > Device 0: Model: QEMU HARDDISK Firm: 2.5+ Ser#: QM00001 > Type: Hard Disk > Supports 48-bit addressing > Capacity: 128.0 MB = 0.1 GB (262144 x 512) > Device 2: Model: QEMU Firm: 2.5+ Ser#: QEMU DVD-ROM > Type: Removable CD ROM > Capacity: not available > => fat2ls ide 0:1 > Unknown command 'fat2ls' - try 'help' > => fatls ide 0:1 > 164768 snp.efi > 0 file1 > 0 file2 > > > After the patch (including HEAD) I cannot read the directory and cannot > load the file snp.efi either: > > => ide info > => fatls ide 0:1 > ** Bad device ide 0 ** > => > => fatls mmc 0:1 > ** Bad device mmc 0 ** > > > In both cases I have loaded the same image with: > > export BUILD_ROM=y > make distclean && make qemu-x86_defconfig && make -j6 > > qemu-system-x86_64 -m 1G -bios u-boot.rom -nographic \ > -netdev \ > user,id=eth0,tftp=tftp,net=192.168.76.0/24,dhcpstart=192.168.76.9 \ > -device e1000,netdev=eth0 -machine pc-i440fx-2.8 -hda img > > Best regards > > Heinrich >

Hello Bin, hello Simon,

I think the bug is in functions ide_init (drivers/block/ide.c).

Platform X86 implies CONFIG_BLK=y.

So we should initialize ide_dev_desc[i].bdev.

We don't, so blk_dread fails after finding no read operation with ENOSYS when called from part_test_dos.

The following comment in include/blk.h confirms that bdev has to be filled:

#if CONFIG_IS_ENABLED(BLK) /*

• For now we have a few functions which take struct blk_desc as a
• parameter. This field allows them to look up the associated
• device. Once these functions are removed we can drop this field.

*/ struct udevice *bdev; #else

I assume that we have the same issue in in __sata_initialize (drivers/ata/sata.c) but have not tested.

Thanks for the testing!

Simon, are you going to fix this?

I am not going to race you to it, if you are thinking of fixing it. I am back from travels in a few days but have a busy week ahead and the release is imminent :-(

Or maybe Heinrich, do you plan to work on a fix? If not, I will put it on my todo list.

Regards, Bin

Hello hello Bin,

I am aware that this bug is release critical. But unfortunately I am departing on vacation tomorrow.

Since there is not much time before the release, I did minimal conversion of the IDE driver to the driver model, and now it works.

...

I guess there should be CONFIG_SYS_IDE_MAXDEVICE times an U_BOOT_IDE_DEVICE(ide_disk[i]) which should refer to driver U_BOOT_DRIVER(ide_blk).

Just as prove of concept I append a patch to this mail. With the patch the IDE partitions are discovered in disk/part.c. But they are not usable to access the disk because the devices are not registered in a uclass.

Hopefully you will find the time to fix the problem.

Please check the patches [1] that I just sent.

[1] http://patchwork.ozlabs.org/project/uboot/list/?series=1298

Regards, Bin

Hello Bin,

thank you for working on this.

The command 'ide info' now produces a reasonable output for the one drive I attached to qemu.

But 'fatls ide 0:1' still claims there is no device ide 0.

See output below.

This is how I start my image:

qemu-system-x86_64 -m 1G -bios denx/u-boot.rom -nographic \ -netdev \ user,id=eth0,tftp=tftp,net=192.168.76.0/24,dhcpstart=192.168.76.9 \ -device e1000,netdev=eth0 -machine pc-i440fx-2.8 -hda img

Best regards

Heinrich

U-Boot 2017.09-rc3-00130-g54c022cc0e (Sep 04 2017 - 07:30:09 +0200)

CPU: x86_64, vendor AMD, device 663h DRAM: 1 GiB Using default environment

Video: 640x480x16 Model: QEMU x86 (I440FX) Net: e1000: 52:54:00:12:34:56

Warning: e1000#0 using MAC address from ROM eth0: e1000#0 IDE: Bus 0: OK Bus 1: OK Device 0: Model: QEMU HARDDISK Firm: 2.5+ Ser#: QM00001 Type: Hard Disk Supports 48-bit addressing Capacity: 128.0 MB = 0.1 GB (262144 x 512) Device 1: not available Device 2: Model: QEMU Firm: 2.5+ Ser#: QEMU DVD-ROM Type: Removable CD ROM Capacity: not available Device 3: not available Hit any key to stop autoboot: 0 ** Bad device scsi 0 ** Error: Invalid Boot Flag (found 0x0000, expected 0xaa55) ## Kernel loading failed ... zboot - Boot bzImage

Usage: zboot [addr] [size] [initrd addr] [initrd size] addr - The optional starting address of the bzimage. If not set it defaults to the environment variable "fileaddr". size - The optional size of the bzimage. Defaults to zero. initrd addr - The address of the initrd image to use, if any. initrd size - The size of the initrd image to use, if any.

=> ide info Device 0: Model: QEMU HARDDISK Firm: 2.5+ Ser#: QM00001 Type: Hard Disk Capacity: 128.0 MB = 0.1 GB (262144 x 512) => fatls ide 0:1 ** Bad device ide 0 **

Hi Heinrich,

On Mon, Sep 4, 2017 at 3:30 PM, Bin Meng bmeng.cn@gmail.com wrote:

Hi Heinrich,

On Mon, Sep 4, 2017 at 1:41 PM, Heinrich Schuchardt xypron.glpk@gmx.de wrote:

...

On 09/04/2017 05:59 AM, Bin Meng wrote:

...

Hi Heinrich,

On Fri, Sep 1, 2017 at 1:24 AM, Heinrich Schuchardt xypron.glpk@gmx.de wrote:

...

On 08/31/2017 02:55 PM, Bin Meng wrote:

...

Hi Simon,

On Thu, Aug 31, 2017 at 8:52 PM, Simon Glass sjg@chromium.org wrote:

...

Hi Bin,

On 31 August 2017 at 10:53, Bin Meng bmeng.cn@gmail.com wrote: > Hi Heinrich, > > On Thu, Aug 31, 2017 at 5:19 AM, Heinrich Schuchardt xypron.glpk@gmx.de wrote: >> On 08/30/2017 06:37 AM, Heinrich Schuchardt wrote: >>> >>> >>> On 08/30/2017 03:54 AM, Bin Meng wrote: >>>> Hi Heinrich, >>>> >>>> On Wed, Aug 30, 2017 at 4:26 AM, Heinrich Schuchardt xypron.glpk@gmx.de wrote: >>>>> Hello Simon, >>>>> >>>>> U-Boot HEAD qemu-86_defconfig cannot discover an IDE disk with one FAT >>>>> partition in qemu-system-x86_64. >>>>> >>>>> By bisection I found this patch. >>>>> >>>>> b7c6baef2891ce8978cbfddb66e944943473ac21 >>>>> x86: Convert MMC to driver model >>>>> >>>>> With this patch I get >>>>> >>>>> IDE: Bus 0: OK Bus 1: OK >>>>> Device 0: Model: QEMU HARDDISK Firm: 2.5+ Ser#: QM00001 >>>>> Type: Hard Disk >>>>> Supports 48-bit addressing >>>>> Capacity: 128.0 MB = 0.1 GB (262144 x 512) >>>>> ** Can't read Driver Desriptor Block ** >>>>> Device 1: not available >>>>> Device 2: Model: QEMU Firm: 2.5+ Ser#: QEMU DVD-ROM >>>>> Type: Removable CD ROM >>>>> Capacity: not available >>>>> Device 3: not available >>>>> >>>>> => ide info >>>>> => >>>>> >>>>> Without the patch I get=> ide info >>>>> Device 0: Model: QEMU HARDDISK Firm: 2.5+ Ser#: QM00001 >>>>> Type: Hard Disk >>>>> Supports 48-bit addressing >>>>> Capacity: 128.0 MB = 0.1 GB (262144 x 512) >>>>> Device 2: Model: QEMU Firm: 2.5+ Ser#: QEMU DVD-ROM >>>>> Type: Removable CD ROM >>>>> Capacity: not available >>>>> >>>>> I think we observe two independent errors here: >>>>> >>>>> - The hard disk Device 0 is not read. >>>>> - The ide command stops at the first device that is not available. >>>>> >>>>> I guess only the first is caused by your patch. >>>> >>>> Both logs look fine to me. The "Can't read Driver Desriptor Block" >>>> comes from part_mac.c. Did you verify the actual IDE read/write fails >>>> with current HEAD? >>>> >>>> Regards, >>>> Bin >>>> >>> >>> Hello Bin, >>> >>> I have not checked block level read but used the shell commands for testing. >>> >>> Before the patch I can read the directory of the drive: >>> >>> => ide info >>> Device 0: Model: QEMU HARDDISK Firm: 2.5+ Ser#: QM00001 >>> Type: Hard Disk >>> Supports 48-bit addressing >>> Capacity: 128.0 MB = 0.1 GB (262144 x 512) >>> Device 2: Model: QEMU Firm: 2.5+ Ser#: QEMU DVD-ROM >>> Type: Removable CD ROM >>> Capacity: not available >>> => fat2ls ide 0:1 >>> Unknown command 'fat2ls' - try 'help' >>> => fatls ide 0:1 >>> 164768 snp.efi >>> 0 file1 >>> 0 file2 >>> >>> >>> After the patch (including HEAD) I cannot read the directory and cannot >>> load the file snp.efi either: >>> >>> => ide info >>> => fatls ide 0:1 >>> ** Bad device ide 0 ** >>> => >>> => fatls mmc 0:1 >>> ** Bad device mmc 0 ** >>> >>> >>> In both cases I have loaded the same image with: >>> >>> export BUILD_ROM=y >>> make distclean && make qemu-x86_defconfig && make -j6 >>> >>> qemu-system-x86_64 -m 1G -bios u-boot.rom -nographic \ >>> -netdev \ >>> user,id=eth0,tftp=tftp,net=192.168.76.0/24,dhcpstart=192.168.76.9 \ >>> -device e1000,netdev=eth0 -machine pc-i440fx-2.8 -hda img >>> >>> Best regards >>> >>> Heinrich >>> >> >> Hello Bin, hello Simon, >> >> I think the bug is in functions ide_init (drivers/block/ide.c). >> >> Platform X86 implies CONFIG_BLK=y. >> >> So we should initialize ide_dev_desc[i].bdev. >> >> We don't, so blk_dread fails after finding no read operation with ENOSYS >> when called from part_test_dos. >> >> The following comment in include/blk.h confirms that bdev has to be filled: >> >> #if CONFIG_IS_ENABLED(BLK) >> /* >> * For now we have a few functions which take struct blk_desc as a >> * parameter. This field allows them to look up the associated >> * device. Once these functions are removed we can drop this field. >> */ >> struct udevice *bdev; >> #else >> >> I assume that we have the same issue in in __sata_initialize >> (drivers/ata/sata.c) but have not tested. >> > > Thanks for the testing! > > Simon, are you going to fix this?

I am not going to race you to it, if you are thinking of fixing it. I am back from travels in a few days but have a busy week ahead and the release is imminent :-(

Or maybe Heinrich, do you plan to work on a fix? If not, I will put it on my todo list.

Regards, Bin

Hello hello Bin,

I am aware that this bug is release critical. But unfortunately I am departing on vacation tomorrow.

Since there is not much time before the release, I did minimal conversion of the IDE driver to the driver model, and now it works.

...

I guess there should be CONFIG_SYS_IDE_MAXDEVICE times an U_BOOT_IDE_DEVICE(ide_disk[i]) which should refer to driver U_BOOT_DRIVER(ide_blk).

Just as prove of concept I append a patch to this mail. With the patch the IDE partitions are discovered in disk/part.c. But they are not usable to access the disk because the devices are not registered in a uclass.

Hopefully you will find the time to fix the problem.

Please check the patches [1] that I just sent.

[1] http://patchwork.ozlabs.org/project/uboot/list/?series=1298

Regards, Bin

Hello Bin,

thank you for working on this.

The command 'ide info' now produces a reasonable output for the one drive I attached to qemu.

But 'fatls ide 0:1' still claims there is no device ide 0.

Can you please try the v2 patch series @ http://patchwork.ozlabs.org/project/uboot/list/?series=1338?

Regards, Bin

Hi,

On Wed, 30 Aug 2017 14:34:18 +0800 Simon Glass wrote:

Hi,

On 30 August 2017 at 09:54, Bin Meng bmeng.cn@gmail.com wrote:

...

Hi Heinrich,

On Wed, Aug 30, 2017 at 4:26 AM, Heinrich Schuchardt xypron.glpk@gmx.de wrote:

...

Hello Simon,

U-Boot HEAD qemu-86_defconfig cannot discover an IDE disk with one FAT partition in qemu-system-x86_64.

By bisection I found this patch.

b7c6baef2891ce8978cbfddb66e944943473ac21 x86: Convert MMC to driver model

With this patch I get

IDE: Bus 0: OK Bus 1: OK Device 0: Model: QEMU HARDDISK Firm: 2.5+ Ser#: QM00001 Type: Hard Disk Supports 48-bit addressing Capacity: 128.0 MB = 0.1 GB (262144 x 512) ** Can't read Driver Desriptor Block ** Device 1: not available Device 2: Model: QEMU Firm: 2.5+ Ser#: QEMU DVD-ROM Type: Removable CD ROM Capacity: not available Device 3: not available

=> ide info =>

Without the patch I get=> ide info Device 0: Model: QEMU HARDDISK Firm: 2.5+ Ser#: QM00001 Type: Hard Disk Supports 48-bit addressing Capacity: 128.0 MB = 0.1 GB (262144 x 512) Device 2: Model: QEMU Firm: 2.5+ Ser#: QEMU DVD-ROM Type: Removable CD ROM Capacity: not available

I think we observe two independent errors here:

• The hard disk Device 0 is not read.
• The ide command stops at the first device that is not available.

I guess only the first is caused by your patch.

Both logs look fine to me. The "Can't read Driver Desriptor Block" comes from part_mac.c. Did you verify the actual IDE read/write fails with current HEAD?

We should definitely change that message to debug(). It is very

... and fix the missing 'c' in 'Desriptor'

Lothar Waßmann