On Thu, 06/30 15:10, Dan Williams wrote:

On Wed, Jun 29, 2016 at 6:59 PM, Fam Zheng [off-list ref] wrote:

quoted

It is documented that KOBJ_ADD should be generated after the object's
attributes and children are ready.  We can achieve this with the new
disk_gen_uevents interface.

Signed-off-by: Fam Zheng <redacted>
---
 arch/powerpc/sysdev/axonram.c | 3 ++-
 1 file changed, 2 insertions(+), 1 deletion(-)

diff --git a/arch/powerpc/sysdev/axonram.c b/arch/powerpc/sysdev/axonram.c
index 4efd69b..27e7175 100644
--- a/arch/powerpc/sysdev/axonram.c
+++ b/arch/powerpc/sysdev/axonram.c

@@ -238,7 +238,7 @@ static int axon_ram_probe(struct platform_device *device)
        set_capacity(bank->disk, bank->size >> AXON_RAM_SECTOR_SHIFT);
        blk_queue_make_request(bank->disk->queue, axon_ram_make_request);
        blk_queue_logical_block_size(bank->disk->queue, AXON_RAM_SECTOR_SIZE);
-       add_disk(bank->disk, true);
+       add_disk(bank->disk, false);

        bank->irq_id = irq_of_parse_and_map(device->dev.of_node, 0);
        if (bank->irq_id == NO_IRQ) {

@@ -262,6 +262,7 @@ static int axon_ram_probe(struct platform_device *device)
                rc = -EFAULT;
                goto failed;
        }
+       disk_gen_uevents(bank->disk);

I assume you are doing this after:

   rc = device_create_file(&device->dev, &dev_attr_ecc);

...so that userspace gets notified of the new attribute, but this
attribute is on the parent device, not the disk itself.  Instead I
think this attribute should simply be registered before the call to
add_disk().  Then the KOBJ_ADD event for the disk comes after the
attribute is available.  It's still not a clean fit, because userspace
should not be expecting a child device uevent to signal new attributes
available on the parent.

Yes you are right, this patch is a mistake. Moving to before add_disk makes
sense to me.

Thanks for taking a look!

Fam