Re: [PATCH v41 20/24] selftests/x86: Add a selftest for SGX

From: Borislav Petkov <bp@alien8.de>
To: Jarkko Sakkinen <jarkko@kernel.org>
Cc: x86@kernel.org, linux-sgx@vger.kernel.org,
	linux-kernel@vger.kernel.org, Shuah Khan <shuah@kernel.org>,
	linux-kselftest@vger.kernel.org,
	Jethro Beekman <jethro@fortanix.com>,
	akpm@linux-foundation.org, andriy.shevchenko@linux.intel.com,
	asapek@google.com, cedric.xing@intel.com,
	chenalexchen@google.com, conradparker@google.com,
	cyhanish@google.com, dave.hansen@intel.com,
	haitao.huang@intel.com, kai.huang@intel.com, kai.svahn@intel.com,
	kmoy@google.com, ludloff@google.com, luto@kernel.org,
	nhorman@redhat.com, npmccallum@redhat.com, puiterwijk@redhat.com,
	rientjes@google.com, sean.j.christopherson@intel.com,
	tglx@linutronix.de, yaozhangx@google.com, mikko.ylinen@intel.com
Subject: Re: [PATCH v41 20/24] selftests/x86: Add a selftest for SGX
Date: Tue, 17 Nov 2020 18:26:50 +0100	[thread overview]
Message-ID: <20201117172650.GI5719@zn.tnic> (raw)
In-Reply-To: <20201112220135.165028-21-jarkko@kernel.org>

On Fri, Nov 13, 2020 at 12:01:31AM +0200, Jarkko Sakkinen wrote:
> +bool encl_load(const char *path, struct encl *encl)
> +{
> +	Elf64_Phdr *phdr_tbl;
> +	off_t src_offset;
> +	Elf64_Ehdr *ehdr;
> +	int i, j;
> +	int ret;
> +
> +	memset(encl, 0, sizeof(*encl));
> +
> +	ret = open("/dev/sgx_enclave", O_RDWR);
> +	if (ret < 0) {
> +		fprintf(stderr, "Unable to open /dev/sgx_enclave\n");
> +		goto err;
> +	}
> +
> +	encl->fd = ret;
> +
> +	if (!encl_map_bin(path, encl))
> +		goto err;
> +
> +	ehdr = encl->bin;
> +	phdr_tbl = encl->bin + ehdr->e_phoff;
> +
> +	for (i = 0; i < ehdr->e_phnum; i++) {
> +		Elf64_Phdr *phdr = &phdr_tbl[i];
> +
> +		if (phdr->p_type == PT_LOAD)
> +			encl->nr_segments++;
> +	}
> +
> +	encl->segment_tbl = calloc(encl->nr_segments,
> +				   sizeof(struct encl_segment));
> +	if (!encl->segment_tbl)
> +		goto err;
> +
> +	for (i = 0, j = 0; i < ehdr->e_phnum; i++) {
> +		Elf64_Phdr *phdr = &phdr_tbl[i];
> +		unsigned int flags = phdr->p_flags;
> +		struct encl_segment *seg;
> +
> +		if (phdr->p_type != PT_LOAD)
> +			continue;
> +
> +		seg = &encl->segment_tbl[j];
> +
> +		if (!!(flags & ~(PF_R | PF_W | PF_X))) {
> +			fprintf(stderr,
> +				"%d has invalid segment flags 0x%02x.\n", i,
> +				phdr->p_flags);
> +			goto err;
> +		}
> +
> +		if (j == 0 && flags != (PF_R | PF_W)) {
> +			fprintf(stderr,
> +				"TCS has invalid segment flags 0x%02x.\n",
> +				phdr->p_flags);
> +			goto err;
> +		}
> +
> +		if (j == 0) {
> +			src_offset = (phdr->p_offset & PAGE_MASK) - src_offset;
> +
> +			seg->prot = PROT_READ | PROT_WRITE;
> +			seg->flags = SGX_PAGE_TYPE_TCS << 8;
> +		} else  {
> +			seg->prot = (phdr->p_flags & PF_R) ? PROT_READ : 0;
> +			seg->prot |= (phdr->p_flags & PF_W) ? PROT_WRITE : 0;
> +			seg->prot |= (phdr->p_flags & PF_X) ? PROT_EXEC : 0;
> +			seg->flags = (SGX_PAGE_TYPE_REG << 8) | seg->prot;
> +		}
> +
> +		seg->offset = (phdr->p_offset & PAGE_MASK) - src_offset;
> +		seg->size = (phdr->p_filesz + PAGE_SIZE - 1) & PAGE_MASK;
> +
> +		printf("0x%016lx 0x%016lx 0x%02x\n", seg->offset, seg->size,
> +		       seg->prot);
> +
> +		j++;
> +	}
> +
> +	assert(j == encl->nr_segments);
> +
> +	encl->src = encl->bin + src_offset;
> +	encl->src_size = encl->segment_tbl[j - 1].offset +
> +			 encl->segment_tbl[j - 1].size;
> +
> +	for (encl->encl_size = 4096; encl->encl_size < encl->src_size; )
> +		encl->encl_size <<= 1;

Something's fishy. That selftest fails with

mmap: Cannot allocate memory

I sprinkled some printfs at this size computation above and here's what
it says:

0x00007fdd3b4ca190 0x0000000000002000 0x03
0x00007fdd3b4cc190 0x0000000000001000 0x05
0x00007fdd3b4cd190 0x0000000000003000 0x03
encl_load: encl->nr_segments: 3
encl_load: seg2 offset: 0x7fdd3b4cd190, seg2 size: 12288
encl_load: encl_size: 140737488355328, src_size: 140588159402384
encl_map_area: encl_size: 140737488355328
mmap: Cannot allocate memory

src_size is computed by adding the offset and size of the last segment
which is at index 2.

The loop computes encl_size to 0x0000800000000000 then and mmap tries to
map double that in encl_map_area(). Looks like too much to me too.

What's up?

-- 
Regards/Gruss,
    Boris.

https://people.kernel.org/tglx/notes-about-netiquette