[U-Boot] [PATCH v3 11/31] blk: Call part_init() in the post_probe() method

[Bin Meng]

Hi Hannes,

On Thu, Dec 13, 2018 at 3:32 PM Hannes Schmelzer <hannes at schmelzer.or.at> wrote:
>
>
>
> On 12/13/18 8:26 AM, Bin Meng wrote:
> > Hi Hannes,
> Hi Bin,
> > On Thu, Dec 13, 2018 at 3:19 PM Hannes Schmelzer <hannes at schmelzer.or.at> wrote:
> >>
> >>
> >> On 10/15/18 11:21 AM, Bin Meng wrote:
> >>> part_init() is currently called in every DM BLK driver, either
> >>> in its bind() or probe() method. However we can use the BLK
> >>> uclass driver's post_probe() method to do it automatically.
> >>>
> >>> Update all DM BLK drivers to adopt this change.
> >>>
> >>> Signed-off-by: Bin Meng <bmeng.cn at gmail.com>
> >>> Reviewed-by: Simon Glass <sjg at chromium.org>
> >>> ---
> >>>
> >>> Changes in v3: None
> >>> Changes in v2: None
> >>>
> >>>    cmd/sata.c                        |  9 ---------
> >>>    common/usb_storage.c              |  4 +---
> >>>    drivers/block/blk-uclass.c        | 12 ++++++++++++
> >>>    drivers/block/ide.c               |  2 --
> >>>    drivers/block/sandbox.c           |  2 +-
> >>>    drivers/mmc/mmc.c                 |  3 ---
> >>>    drivers/nvme/nvme.c               |  1 -
> >>>    drivers/scsi/scsi.c               |  1 -
> >>>    lib/efi_driver/efi_block_device.c |  2 --
> >>>    9 files changed, 14 insertions(+), 22 deletions(-)
> >> Hi Bin,
> >>
> >> sorry for this very late report on this.
> >> I Just merged the 'v2019.01-rc1' tag into my branch and noticed that
> >> filesystem on my eMMC targets are not working anymore.
> >> Bisect showed up, that this commit breaks the stuff.
> >>
> >> I reviewed the change a bit and i would say that all other block devices
> >> should work as known, except MMC, NVME and EFI (i do not have one, so i
> >> cannot test).
> >>> diff --git a/drivers/mmc/mmc.c b/drivers/mmc/mmc.c
> >>> index 585951c..d6b9cdc 100644
> >>> --- a/drivers/mmc/mmc.c
> >>> +++ b/drivers/mmc/mmc.c
> >>> @@ -2444,9 +2444,6 @@ static int mmc_startup(struct mmc *mmc)
> >>>        bdesc->product[0] = 0;
> >>>        bdesc->revision[0] = 0;
> >>>    #endif
> >>> -#if !defined(CONFIG_SPL_BUILD) || defined(CONFIG_SPL_LIBDISK_SUPPORT)
> >>> -     part_init(bdesc);
> >>> -#endif
> >>>
> >>>        return 0;
> >>>    }
> >> i changed this to:
> >>
> >> #if (!defined(CONFIG_SPL_BUILD) || defined(CONFIG_SPL_LIBDISK_SUPPORT)) && \
> >>        !defined(CONFIG_BLK)
> >>       part_init(bdesc);
> >> #endif
> > I don't understand. The patch above removed 3 lines. Did you say that
> > you added back these 3 lines with additional "!defined(CONFIG_BLK)"?
> Yes,
> in case if CONFIG_BLK is disabled (meaning blk-uclass.c isn't compiled
> in and does therefore nothing),
> the part_init(...) has to be done here.

OK, now I understand. But you should convert your board to use DM MMC
instead. It's the right timing now.

> >
> >>> diff --git a/drivers/nvme/nvme.c b/drivers/nvme/nvme.c
> >>> index eb6fded..1ee0a0a 100644
> >>> --- a/drivers/nvme/nvme.c
> >>> +++ b/drivers/nvme/nvme.c
> >>> @@ -664,7 +664,6 @@ static int nvme_blk_probe(struct udevice *udev)
> >>>        sprintf(desc->vendor, "0x%.4x", pplat->vendor);
> >>>        memcpy(desc->product, ndev->serial, sizeof(ndev->serial));
> >>>        memcpy(desc->revision, ndev->firmware_rev, sizeof(ndev->firmware_rev));
> >>> -     part_init(desc);
> >>>
> >>>        return 0;
> >>>    }
> >>>
> >>> diff --git a/lib/efi_driver/efi_block_device.c b/lib/efi_driver/efi_block_device.c
> >>> index 7b71b4d..3f147cf 100644
> >>> --- a/lib/efi_driver/efi_block_device.c
> >>> +++ b/lib/efi_driver/efi_block_device.c
> >>> @@ -173,8 +173,6 @@ static int efi_bl_bind(efi_handle_t handle, void *interface)
> >>>                return ret;
> >>>        EFI_PRINT("%s: block device '%s' created\n", __func__, bdev->name);
> >>>
> >>> -     ret = blk_prepare_device(bdev);
> >>> -
> >>>        /* Create handles for the partions of the block device */
> >>>        disks = efi_bl_bind_partitions(handle, bdev);
> >>>        EFI_PRINT("Found %d partitions\n", disks);
> >> here i cannot see some alternative if CONFIG_BLK is enabled.
> >>
> >> I don't understand your comments here too...
> With alternative i mean, that:
> If CONFIG_BLK is disabled within the code here some part_init has to be
> done.

No, this one I still don't understand. The EFI block device is a DM
driver. It cannot work when CONFIG_BLK is disabled.

> Otherwise Partitions and following filesystems cannot work.
> >
> >> I know, everything shall be converted to DM since the deadline is quite
> >> close.
> >> But i also think that upon this day everything should work as expected.
> >>

Regards,
Bin