All of lore.kernel.org
 help / color / mirror / Atom feed
From: Heinrich Schuchardt <heinrich.schuchardt@canonical.com>
To: AKASHI Takahiro <takahiro.akashi@linaro.org>
Cc: Marek Vasut <marex@denx.de>,
	agraf@csgraf.de, sjg@chromium.org, ilias.apalodimas@linaro.org,
	u-boot@lists.denx.de
Subject: Re: [RFC 3/3] cmd: efidebug: handle booting from removable media
Date: Wed, 10 Nov 2021 10:17:50 +0100	[thread overview]
Message-ID: <ed7724bf-fb88-01ad-70a2-872e6f32b215@canonical.com> (raw)
In-Reply-To: <66353088-caad-7953-6a12-eded28576c3d@canonical.com>

On 11/10/21 09:11, Heinrich Schuchardt wrote:
> 
> 
> On 11/10/21 07:24, AKASHI Takahiro wrote:
>> On Tue, Nov 09, 2021 at 10:50:37AM +0100, Heinrich Schuchardt wrote:
>>> On 11/9/21 02:32, AKASHI Takahiro wrote:
>>>> Support for booting from removable media is now added to UEFI boot
>>>> manager. Here we should modify efidebug command in order to define
>>>> a proper "BootXXXX" variable.
>>>>
>>>> With this patch applied, you will be able to specify the boot order,
>>>> usb and scsi:
>>>
>>> I guess for a removable device this should work even if the device is 
>>> not
>>> present. But currently:
>>>
>>> => efidebug boot add -b 1000 USB_present usb 0:1 EFI/BOOT/BOOTARM.EFI
>>> => efidebug boot add -b 1000 USB_not_present usb 1:1 
>>> EFI/BOOT/BOOTARM.EFI
>>> Cannot create device path for "usb 1:1"
>>
>> # In the second, I don't expect you to specify the file path,
>> # "EFI/BOOT/BOOTARM.EFI", for removable media support.
>>
>> Yes, I have been aware of this issue but it is not this-patch specific
>> but an existing issue as you clearly mentioned above.
>>
>>> A media device path like:
>>>
>>> /VenHw(e61d73b9-a384-4acc-aeab-82e828f3628b)/UsbClass(0x0,0x0,0x9,0x0,0x1)/UsbClass(0x781,0x5571,0x0,0x0,0x0)/HD(1,MBR,0x0c449046,0x800,0x800) 
>>>
>>>
>>> is not very helpful because the next device that you insert may have a
>>> different location of the ESP partition.
>>>
>>> I think you should store
>>>
>>> /VenHw(e61d73b9-a384-4acc-aeab-82e828f3628b)/UsbClass(0x0,0x0,0x9,0x0,0x1)/UsbClass(0x781,0x5571,0x0,0x0,0x0) 
>>>
>>
>> Apparently it is promising but actually not because
>> "UsbClass(0x781,0x5571,0x0,0x0,0x0)" contains Vendor ID and
>> Product ID which can only be retrieved from a real device attached
>> to the board.
> 
> 0x781,0x5571 relates to
> Vendor: SanDisk Corp.
> Device: Cruzer Fit
> 
> This is how Linux sees my OrangePi PC:
> 
> $ lsusb
> Bus 009 Device 001: ID 1d6b:0001 Linux Foundation 1.1 root hub
> Bus 007 Device 001: ID 1d6b:0002 Linux Foundation 2.0 root hub
> Bus 008 Device 001: ID 1d6b:0001 Linux Foundation 1.1 root hub
> Bus 006 Device 002: ID 0781:5571 SanDisk Corp. Cruzer Fit
> Bus 006 Device 001: ID 1d6b:0002 Linux Foundation 2.0 root hub
> Bus 005 Device 001: ID 1d6b:0001 Linux Foundation 1.1 root hub
> Bus 004 Device 001: ID 1d6b:0002 Linux Foundation 2.0 root hub
> Bus 003 Device 001: ID 1d6b:0001 Linux Foundation 1.1 root hub
> Bus 002 Device 001: ID 1d6b:0002 Linux Foundation 2.0 root hub
> Bus 001 Device 001: ID 1d6b:0002 Linux Foundation 2.0 root hub
> 
> The bus numbering seems not to be stable:
> 
> $ lsusb
> Bus 009 Device 001: ID 1d6b:0001 Linux Foundation 1.1 root hub
> Bus 008 Device 001: ID 1d6b:0002 Linux Foundation 2.0 root hub
> Bus 007 Device 001: ID 1d6b:0001 Linux Foundation 1.1 root hub
> Bus 005 Device 002: ID 0781:5571 SanDisk Corp. Cruzer Fit
> Bus 005 Device 001: ID 1d6b:0002 Linux Foundation 2.0 root hub
> Bus 006 Device 001: ID 1d6b:0001 Linux Foundation 1.1 root hub
> Bus 004 Device 001: ID 1d6b:0002 Linux Foundation 2.0 root hub
> Bus 003 Device 001: ID 1d6b:0001 Linux Foundation 1.1 root hub
> Bus 001 Device 001: ID 1d6b:0002 Linux Foundation 2.0 root hub
> Bus 002 Device 001: ID 1d6b:0002 Linux Foundation 2.0 root hub
> 
> 
> And this is in U-Boot:
> 
> usb           0  [ + ]   ehci_generic          |   |-- usb@1c1a000
> usb_hub       0  [ + ]   usb_hub               |   |   `-- usb_hub
> usb           1  [ + ]   ohci_generic          |   |-- usb@1c1a400
> usb_hub       1  [ + ]   usb_hub               |   |   `-- usb_hub
> usb           2  [ + ]   ehci_generic          |   |-- usb@1c1b000
> usb_hub       2  [ + ]   usb_hub               |   |   `-- usb_hub
> usb           3  [ + ]   ohci_generic          |   |-- usb@1c1b400
> usb_hub       3  [ + ]   usb_hub               |   |   `-- usb_hub
> usb           4  [ + ]   ehci_generic          |   |-- usb@1c1c000
> usb_hub       4  [ + ]   usb_hub               |   |   `-- usb_hub
> usb_mass_s    0  [ + ]   usb_mass_storage      |   |       `-- 
> usb_mass_storage
> blk           1  [   ]   usb_storage_blk       |   |           `-- 
> usb_mass_storage.lun0
> usb           5  [ + ]   ohci_generic          |   |-- usb@1c1c400
> usb_hub       5  [ + ]   usb_hub               |   |   `-- usb_hub
> usb           6  [ + ]   ehci_generic          |   |-- usb@1c1d000
> usb_hub       6  [ + ]   usb_hub               |   |   `-- usb_hub
> usb           7  [ + ]   ohci_generic          |   |-- usb@1c1d400
> usb_hub       7  [ + ]   usb_hub               |   |   `-- usb_hub
> 
> The location where a USB is plugged is identified by the port of the USB 
> hub it is connected to.
> 
> I think before we can properly support removable media we will have to 
> get the UEFI device path right. We need a node on all of the levels of 
> the DM tree of which more than one instance can exist.

Here is an example of device path created by an AMI BIOS for an ESP 
partition on a USB stick:

PciRoot(0x0)/Pci(0x1,0x2)/Pci(0x0,0x0)/Pci(0x8,0x0)/Pci(0x0,0x1)/USB(0x2,0x0)/USB(0x3,0x0)/HD(1,GPT,1AD3DE75-504A-47DF-A5CE-81D9EF0252A2,0x40,0x55D824)

The two number in PCI(foo, bar) are:
PCI device
PCI function

The two numbers in USB(foo, bar) are:
USB Parent Port Number
USB Interface Number

By using USB() instead of USBClass() the device path becomes generic. If 
you plug in a device from a different vendor into the same USB port, you 
will get the same device path for the device. Only the partition 
information may change.

Best regards

Heinrich

> 
>>
>> I'm not yet sure where "UsbClass(0x0,0x0,0x9,0x0,0x1)" comes from.
>>
>> "USB(X, Y)", where X is a parent_port_number and Y is a usb_interface,
>> would be a more appropriate candidate for this kind of device path,
>> but we don't utilise this form in the current implementation.
>>
>>> and find the ESP on it at runtime.
>>
>> I don't think the specification requires that the disk is an ESP
>> (at least explicitly).
>>
>>
>>>> => efidebug -b 1 SCSI scsi 0:1
>>>> => efidebug boot dump
>>>> Boot0001:
>>>> attributes: A-- (0x00000001)
>>>>     label: SCSI
>>>>     file_path: /VenHw(e61d73b9-a384-4acc-aeab-82e828f3628b)/Scsi(0,0)/
>>
>> Contrary to USB, we use Scsi(0,0) for scsi devices but it is NOT 
>> identical
>> to U-Boot's scsi0 (in boot_targets).
>>
>>>>     HD(1,GPT,0ed48d12-1b4c-4e08-b3ee-decf20428036,0x800,0xa000)
>>>>     data:
>>>>       00000000: 00 00
>>>> => efideubg -b 2 USB usb 0:1
>>>
>>> efidebug
>>
>> OK
>>
>> -Takahiro Akashi
>>
>>> Best regards
>>>
>>> Heinrich
>>>
>>>> => efidebug boot order 2 1
>>>> => bootefi bootmgr
>>>>
>>>> Signed-off-by: AKASHI Takahiro <takahiro.akashi@linaro.org>
>>>> ---
>>>>    cmd/efidebug.c | 46 ++++++++++++++++++++++++++++++++++++++++------
>>>>    1 file changed, 40 insertions(+), 6 deletions(-)
>>>>
>>>> diff --git a/cmd/efidebug.c b/cmd/efidebug.c
>>>> index a977ca9c72f5..aaf269cdf47d 100644
>>>> --- a/cmd/efidebug.c
>>>> +++ b/cmd/efidebug.c
>>>> @@ -933,6 +933,29 @@ out:
>>>>        return initrd_dp;
>>>>    }
>>>> +/**
>>>> + * count_arguments - count the number of arguments
>>>> + * @argc:    Total number of arguments
>>>> + * @argv:    Argument array
>>>> + * Return:    Number of arguments
>>>> + *
>>>> + * Count the number of arguments for a given option, "-?"
>>>> + * Specifically if the first argument is not "-?", return 0;
>>>> + */
>>>> +static int count_arguments(int argc, char *const argv[])
>>>> +{
>>>> +    int i;
>>>> +
>>>> +    if (argv[0][0] != '-')
>>>> +        return 0;
>>>> +
>>>> +    for (i = 1; i < argc; i++)
>>>> +        if (argv[i][0] == '-')
>>>> +            break;
>>>> +
>>>> +    return i;
>>>> +}
>>>> +
>>>>    /**
>>>>     * do_efi_boot_add() - set UEFI load option
>>>>     *
>>>> @@ -945,7 +968,7 @@ out:
>>>>     *
>>>>     * Implement efidebug "boot add" sub-command. Create or change 
>>>> UEFI load option.
>>>>     *
>>>> - * efidebug boot add -b <id> <label> <interface> <devnum>[:<part>] 
>>>> <file>
>>>> + * efidebug boot add -b <id> <label> <interface> <devnum>[:<part>] 
>>>> [<file>]
>>>>     *                   -i <file> <interface2> <devnum2>[:<part>] 
>>>> <initrd>
>>>>     *                   -s '<options>'
>>>>     */
>>>> @@ -979,7 +1002,10 @@ static int do_efi_boot_add(struct cmd_tbl 
>>>> *cmdtp, int flag,
>>>>        argv++; /* 'add' */
>>>>        for (; argc > 0; argc--, argv++) {
>>>>            if (!strcmp(argv[0], "-b")) {
>>>> -            if (argc <  5 || lo.label) {
>>>> +            int num_args;
>>>> +
>>>> +            num_args = count_arguments(argc, argv);
>>>> +            if (num_args <  5 || num_args > 6 || lo.label) {
>>>>                    r = CMD_RET_USAGE;
>>>>                    goto out;
>>>>                }
>>>> @@ -1000,7 +1026,8 @@ static int do_efi_boot_add(struct cmd_tbl 
>>>> *cmdtp, int flag,
>>>>                utf8_utf16_strncpy(&label, argv[2], label_len);
>>>>                /* file path */
>>>> -            ret = efi_dp_from_name(argv[3], argv[4], argv[5],
>>>> +            ret = efi_dp_from_name(argv[3], argv[4],
>>>> +                           num_args == 5 ? "" : argv[5],
>>>>                               &device_path, &file_path);
>>>>                if (ret != EFI_SUCCESS) {
>>>>                    printf("Cannot create device path for \"%s %s\"\n",
>>>> @@ -1008,10 +1035,16 @@ static int do_efi_boot_add(struct cmd_tbl 
>>>> *cmdtp, int flag,
>>>>                    r = CMD_RET_FAILURE;
>>>>                    goto out;
>>>>                }
>>>> +            /* if no file name is given, use device_path */
>>>> +            if (num_args == 5) {
>>>> +                efi_free_pool(file_path);
>>>> +                file_path = device_path;
>>>> +                device_path = NULL;
>>>> +            }
>>>>                fp_size += efi_dp_size(file_path) +
>>>>                    sizeof(struct efi_device_path);
>>>> -            argc -= 5;
>>>> -            argv += 5;
>>>> +            argc -= num_args;
>>>> +            argv += num_args;
>>>>            } else if (!strcmp(argv[0], "-i")) {
>>>>                if (argc < 3 || initrd_dp) {
>>>>                    r = CMD_RET_USAGE;
>>>> @@ -1078,7 +1111,8 @@ out:
>>>>        free(data);
>>>>        efi_free_pool(final_fp);
>>>>        efi_free_pool(initrd_dp);
>>>> -    efi_free_pool(device_path);
>>>> +    if (device_path)
>>>> +        efi_free_pool(device_path);
>>>>        efi_free_pool(file_path);
>>>>        free(lo.label);
>>>>


  reply	other threads:[~2021-11-10  9:18 UTC|newest]

Thread overview: 13+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2021-11-09  1:32 [RFC 0/3] efi_loader: bootmgr itself should support removable media AKASHI Takahiro
2021-11-09  1:32 ` [RFC 1/3] efi_loader: export efi_locate_device_handle() AKASHI Takahiro
2021-11-09  1:32 ` [RFC 2/3] efi_loader: bootmgr: add booting from removable media AKASHI Takahiro
2021-11-09  9:14   ` Heinrich Schuchardt
2021-11-10  5:38     ` AKASHI Takahiro
2021-11-09 13:53   ` Mark Kettenis
2021-11-10  5:52     ` AKASHI Takahiro
2021-11-09  1:32 ` [RFC 3/3] cmd: efidebug: handle " AKASHI Takahiro
2021-11-09  9:50   ` Heinrich Schuchardt
2021-11-10  6:24     ` AKASHI Takahiro
2021-11-10  8:11       ` Heinrich Schuchardt
2021-11-10  9:17         ` Heinrich Schuchardt [this message]
2021-11-12  1:51           ` AKASHI Takahiro

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=ed7724bf-fb88-01ad-70a2-872e6f32b215@canonical.com \
    --to=heinrich.schuchardt@canonical.com \
    --cc=agraf@csgraf.de \
    --cc=ilias.apalodimas@linaro.org \
    --cc=marex@denx.de \
    --cc=sjg@chromium.org \
    --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 an external index of several public inboxes,
see mirroring instructions on how to clone and mirror
all data and code used by this external index.