On Fri, 2021-01-15 at 10:07 -0800, James Bottomley wrote:

On Fri, 2021-01-15 at 09:26 -0800, James Bottomley wrote:

quoted

On Fri, 2021-01-15 at 14:54 +0100, Greg KH wrote:

quoted

On Thu, Jan 14, 2021 at 04:21:08PM -0800, James Bottomley wrote:

[...]

quoted

quoted

It looks like we already have a couple of bugs in the kernel
introduced by this confusion ...  return sysfs_emit() vs return
sysfs_emit_at() being the most tricky ...

Hm, Joe, you did the conversion to these functions (and wrote the
api), care to review this?

A cursory glance tells me that summary_show in 
drivers/infiniband/hw/usnic/usnic_ib_sysfs.c has a problem, I think
the last = should be +=

No, it's correct and overwriting what would otherwise be a trailing space.

The use in drivers/base/node.c:node_read_meminfo() is highly
questionable.  While currently not emitting wrong code, it depends on
len being 0 when passed in to sysfs_emit_at().  That argues it should
either be using sysfs_emit() or it should have a len += just in case
something gets prepended that makes len non zero.

<shrug>, it's currently correct and would be OK to change to += as
len is initialized though I think it's extremely doubtful it would
ever be changed.

sysfs is ABI right so prepending would break things.