Received-SPF: pass (google.com: domain of linux-kernel-owner@vger.kernel.org designates 23.128.96.18 as permitted sender) client-ip=23.128.96.18;
Date:   Fri, 2 Oct 2020 11:01:14 -0500
From:   Josh Poimboeuf <jpoimboe@redhat.com>
To:     Vasily Gorbik <gor@linux.ibm.com>
Cc:     Peter Zijlstra <peterz@infradead.org>,
        Thomas Gleixner <tglx@linutronix.de>,
        Masami Hiramatsu <mhiramat@kernel.org>,
        David Laight <David.Laight@aculab.com>,
        Ingo Molnar <mingo@redhat.com>, Borislav Petkov <bp@alien8.de>,
        x86@kernel.org, "H. Peter Anvin" <hpa@zytor.com>,
        Miroslav Benes <mbenes@suse.cz>,
        Alexandre Chartre <alexandre.chartre@oracle.com>,
        Julien Thierry <jthierry@redhat.com>,
        linux-kernel@vger.kernel.org
Subject: Re: [RFC PATCH v3 4/4] objtool: fix x86 orc generation on big endian
 cross compiles
Message-ID: <20201002160114.7yb7z7aeijhchpwl@treble>
References: <cover.thread-6ec90b.your-ad-here.call-01601502173-ext-7769@work.hours>
 <patch-4.thread-6ec90b.git-6ec90b880ce6.your-ad-here.call-01601502173-ext-7769@work.hours>
MIME-Version: 1.0
Content-Type: text/plain; charset=utf-8
Content-Disposition: inline
In-Reply-To: <patch-4.thread-6ec90b.git-6ec90b880ce6.your-ad-here.call-01601502173-ext-7769@work.hours>
Precedence: bulk

On Thu, Oct 01, 2020 at 12:17:32AM +0200, Vasily Gorbik wrote:
> +++ b/tools/objtool/arch/x86/special.c
> @@ -9,7 +9,7 @@
>  
>  void arch_handle_alternative(unsigned short feature, struct special_alt *alt)
>  {
> -	switch (feature) {
> +	switch (le16_to_cpu(feature)) {

It might be cleaner for the endian conversion to be done when the
'feature' value is first read.

		feature = *(unsigned short *)(sec->data->d_buf + offset +
					      entry->feature);


>  	case X86_FEATURE_SMAP:
>  		/*
>  		 * If UACCESS validation is enabled; force that alternative;
> diff --git a/tools/objtool/check.c b/tools/objtool/check.c
> index 2df9f769412e..f20a4be2fb22 100644
> --- a/tools/objtool/check.c
> +++ b/tools/objtool/check.c
> @@ -1370,7 +1370,7 @@ static int read_unwind_hints(struct objtool_file *file)
>  		cfa = &insn->cfi.cfa;
>  
>  		if (hint->type == UNWIND_HINT_TYPE_RET_OFFSET) {
> -			insn->ret_offset = hint->sp_offset;
> +			insn->ret_offset = le16_to_cpu(hint->sp_offset);

Since this is common code, we might not always be able to assume the
value is little endian.  Could you make a more generic conversion macro
which -- when the target ELF file's endianness doesn't match the host
CPU's -- does a byte swap?  For example:

	insn->ret_offset = bswap_if_needed(hint->sp_offset);

The macro could detect the type size, and would also know the
host/target endianness, and could swap accordingly.  It could then be
called for all such multi-byte reads.

-- 
Josh