linux-kernel - Re: [PATCH v5 5/5] x86/sgx: Enable automatic SVN updates for SGX enclaves

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]

Message-ID: <aCrlFfl2lAzzHOA8@gmail.com>
Date: Mon, 19 May 2025 10:00:21 +0200
From: Ingo Molnar <mingo@...nel.org>
To: Elena Reshetova <elena.reshetova@...el.com>
Cc: dave.hansen@...el.com, jarkko@...nel.org, seanjc@...gle.com,
	kai.huang@...el.com, linux-sgx@...r.kernel.org,
	linux-kernel@...r.kernel.org, x86@...nel.org,
	asit.k.mallick@...el.com, vincent.r.scarlata@...el.com,
	chongc@...gle.com, erdemaktas@...gle.com, vannapurve@...gle.com,
	dionnaglaze@...gle.com, bondarn@...gle.com, scott.raynor@...el.com
Subject: Re: [PATCH v5 5/5] x86/sgx: Enable automatic SVN updates for SGX
 enclaves


* Elena Reshetova <elena.reshetova@...el.com> wrote:

> @@ -19,10 +19,15 @@ static int sgx_open(struct inode *inode, struct file *file)
>  	struct sgx_encl *encl;
>  	int ret;
>  
> +	ret = sgx_inc_usage_count();
> +	if (ret)
> +		return -EBUSY;

So if sgx_inc_usage_count() returns nonzero, it's in use already and we 
return -EBUSY, right?

But:

>  int sgx_inc_usage_count(void)
>  {
> +	int ret;
> +
> +	/*
> +	 * Increments from non-zero indicate EPC other
> +	 * active EPC users and EUPDATESVN is not attempted.
> +	 */
> +	if (atomic64_inc_not_zero(&sgx_usage_count))
> +		return 0;

If sgx_usage_count is 1 here (ie. it's busy), this will return *zero*, 
and sgx_open() will not run into the -EBUSY condition and will continue 
assuming it has claimed the usage count, while it hasn't ...

Furthermore, in the sgx_open() error paths we can then run into 
sgx_dec_usage_count(), which will merrily underflow the counter into 
negative:

 +void sgx_dec_usage_count(void)
 +{
 +       atomic64_dec(&sgx_usage_count);
 +}

How is this all supposed to work?

Thanks,

	Ingo