In one test, when modprobe zram, no zram device was found in the /dev. 
But don't see any errors printed in jouranls/dmesg. Later we found out 
that the reason was that device_add() did not check its return value when
calling devtmpfs_create_node().

The test steps:
1. Set the SElinux label of /dev to user_home_t 
2. modprobe zram num_devices=1000

v1 -> v2:
- New patch 1 to add error handling for devtmpfs_create_node().
- use dev_err() to replace pr_err_ratelimited in [2].
- only remove return value of devtmpfs_delete_node() in [3].

Longlong Xia (3):
  driver core: add error handling for devtmpfs_create_node()
  devtmpfs: add debug info to handle()
  devtmpfs: remove return value of devtmpfs_delete_node()

 drivers/base/base.h     |  4 ++--
 drivers/base/core.c     |  6 +++++-
 drivers/base/devtmpfs.c | 20 ++++++++++++++------
 3 files changed, 21 insertions(+), 9 deletions(-)

-- 
2.25.1

On Fri, Feb 10, 2023 at 09:54:41AM +0000, Longlong Xia wrote:
> In one test, when modprobe zram, no zram device was found in the /dev. 
> But don't see any errors printed in jouranls/dmesg. Later we found out 
> that the reason was that device_add() did not check its return value when
> calling devtmpfs_create_node().
> 
> The test steps:
> 1. Set the SElinux label of /dev to user_home_t 
> 2. modprobe zram num_devices=1000
> 
> v1 -> v2:
> - New patch 1 to add error handling for devtmpfs_create_node().
> - use dev_err() to replace pr_err_ratelimited in [2].
> - only remove return value of devtmpfs_delete_node() in [3].
> 
> Longlong Xia (3):
>   driver core: add error handling for devtmpfs_create_node()
>   devtmpfs: add debug info to handle()
>   devtmpfs: remove return value of devtmpfs_delete_node()
> 
>  drivers/base/base.h     |  4 ++--
>  drivers/base/core.c     |  6 +++++-
>  drivers/base/devtmpfs.c | 20 ++++++++++++++------
>  3 files changed, 21 insertions(+), 9 deletions(-)
> 
> -- 
> 2.25.1
> 

Thanks for the quick resend, nice work, all now queued up!

greg k-h