Re: [PATCH v3] HID: debug: fix the ring buffer implementation

[Vladis Dronov <vdronov@redhat.com>]

Hello, Sasha, all,

> The reason for that mess is that on <=4.14 kernels, there is a
> dependency on 6396bb22151 ("treewide: kzalloc() -> kcalloc()"), but
> since that patch is a treewide change, to bring that in we need a big
> pile of seemingly random patches from all over the tree, which is the
> output you see.

Indeed, this patch depends on 2 tree-wide patches, one being 6396bb22151
("treewide: kzalloc() -> kcalloc()") as you've mentioned and another one
is a9a08845e9ac ("vfs: do bulk POLL* -> EPOLL* replacement"). In the part
of HID driver they basically do:

#6396bb22151

-        buf = kzalloc(sizeof(char) * HID_DEBUG_BUFSIZE, GFP_ATOMIC);
+        buf = kzalloc(HID_DEBUG_BUFSIZE, GFP_ATOMIC);

-    if (!(list->hid_debug_buf = kzalloc(sizeof(char) * HID_DEBUG_BUFSIZE, GFP_KERNEL))) {
+    if (!(list->hid_debug_buf = kzalloc(HID_DEBUG_BUFSIZE, GFP_KERNEL))) {

#a9a08845e9ac
-        return POLLIN | POLLRDNORM;
+        return EPOLLIN | EPOLLRDNORM;

-        return POLLERR | POLLHUP;
+        return EPOLLERR | EPOLLHUP;

Adding this 2 changes to the 4.14.y make HID patch to apply cleanly (except
offsets). I suppose, I need to backport HID patch (adding these 2 changes
above) and post to stable@.

I will follow this process:
https://www.kernel.org/doc/html/latest/process/stable-kernel-rules.html

In case I do smth wrong, please, tell.

Best regards,
Vladis Dronov | Red Hat, Inc. | Product Security | Senior Software Engineer