Re: [RFC PATCH 1/1] x86/sgx: Explicitly give up the CPU in EDMM's ioctl() to avoid softlockup

From: "Jarkko Sakkinen" <jarkko@kernel.org>
To: "Jarkko Sakkinen" <jarkko@kernel.org>,
	"Bojun Zhu" <zhubojun.zbj@antgroup.com>
Cc: "Huang, Kai" <kai.huang@intel.com>,
	"linux-sgx@vger.kernel.org" <linux-sgx@vger.kernel.org>,
	"linux-kernel@vger.kernel.org" <linux-kernel@vger.kernel.org>,
	"dave.hansen@linux.intel.com" <dave.hansen@linux.intel.com>,
	"刘双(轩屹)" <ls123674@antgroup.com>,
	"Chatre, Reinette" <reinette.chatre@intel.com>
Subject: Re: [RFC PATCH 1/1] x86/sgx: Explicitly give up the CPU in EDMM's ioctl() to avoid softlockup
Date: Wed, 24 Apr 2024 20:47:46 +0300	[thread overview]
Message-ID: <D0SJIC3GSSKA.3S3RTNVWC55TB@kernel.org> (raw)
In-Reply-To: <D0SJFJIPUE1M.36M8JVVXO89YQ@kernel.org>

On Wed Apr 24, 2024 at 8:44 PM EEST, Jarkko Sakkinen wrote:
> On Wed Apr 24, 2024 at 2:50 PM EEST, Bojun Zhu wrote:
> > I still have some questions:
> >
> > It seems that the variable "ret" is set to 0 if there is **some** EPC pages have been 
> > added when interrupted by signal(Supposed that sgx_encl_add_page() 
> > always returns successfully).
>
> Ah, ok.
>
> Returning zero is right thing to do because it also returns count of
> pages successfully added. I.e. the function does not guarantee that
> all pages are processsed but it does guarantee that the system is in
> predictable state.
>
> It could be that e.g. sgx_alloc_epc_page() calls fails.
>
> So, it is a bit like how read system call works.

Good that you asked that had rethink for a while.

What I would suggest that you just put out 2nd verson out and then we
see where it is going.

For sgx_ioc_encl_add_pages() it is important to maintain exact semantics
as we need to maintain backwards compatibility.

BR, Jarkko