Re: new checkpatch flexible array test ? (was Re: [PATCH v4 12/14] gunyah: rsc_mgr: Add RPC for console services)
From: Gustavo A. R. Silva
Date: Wed Nov 02 2022 - 03:58:17 EST
On 10/3/22 00:38, Joe Perches wrote:
On Mon, 2022-10-03 at 07:29 +0200, Greg Kroah-Hartman wrote:
On Sun, Oct 02, 2022 at 06:46:30PM -0700, Joe Perches wrote:
On Fri, 2022-09-30 at 14:22 +0200, Greg Kroah-Hartman wrote:
On Wed, Sep 28, 2022 at 12:56:31PM -0700, Elliot Berman wrote:
Gunyah resource manager defines a simple API for virtual machine log
sharing with the console service.
[]
diff --git a/include/linux/gunyah_rsc_mgr.h b/include/linux/gunyah_rsc_mgr.h
[]
+struct gh_rm_notif_vm_console_chars {
+ u16 vmid;
+ u16 num_bytes;
+ u8 bytes[0];
Please do not use [0] for new structures, otherwise we will just have to
fix them up again as we are trying to get rid of all of these from the
kernel. Just use "bytes[];" instead.
Maybe a checkpatch addition like:
---
[]
diff --git a/scripts/checkpatch.pl b/scripts/checkpatch.pl
[]
@@ -3948,6 +3948,17 @@ sub process {
}
}
+# check for zero length array declarations in likely structs
+ if ($line =~ /^\+\t($Declare\s*$Ident)\s*\[\s*0\s*\]\s*;\s*$/ &&
+ defined $lines[$linenr] &&
+ $lines[$linenr] =~ /^[\+ ]\}\s*(?:__\w+\s*(?:$balanced_parens)?)\s*;\s*$/) {
This sounds great. We need the same for one-element arrays. :)
Both zero-length and one-element arrays are deprecated.
This should actually be:
$lines[$linenr] =~ /^[\+ ]\}(?:\s*__\w+\s*(?:$balanced_parens)?)*\s*;\s*$/) {
I agree. Thanks.
as it was missing a * for uses like
int foo[0];
} __packed __aligned(4);
and uses without any attribute at all
+ if (WARN("FLEXIBLE_ARRAY_ZERO",
+ "Prefer flexible length array declarations with [] over [0]\n" . $herecurr) &&
+ $fix) {
+ $fixed[$fixlinenr] =~ s/\[\s*0\s*\]/[]/;
+ }
+ }
+
# check for multiple consecutive blank lines
if ($prevline =~ /^[\+ ]\s*$/ &&
$line =~ /^\+\s*$/ &&
This is a question for Gustavo, who did all the work here. Gustavo,
does the above checkpatch change look good to you?
Yep; the idea is great. :)
Another alternative to stop those fake flex-arrays from entering the codebase
is to run Coccinelle scripts during linux-next builds, as suggested by Elena
Reshetova at LSSEU a couple of months ago.
However, if these can be stopped with checkpatch it'd be really helpful, as well.
Thanks
--
Gustavo