linux-kernel - Re: [RFC][PATCH] almost certain bug in drivers/crypto/chelsio/chcr_algo.c:create_authenc

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] [day] [month] [year] [list]

Date:   Sat, 15 Feb 2020 06:29:30 +0000
From:   Al Viro <viro@...iv.linux.org.uk>
To:     Atul Gupta <atul.gupta@...lsio.com>
Cc:     linux-crypto@...r.kernel.org, linux-kernel@...r.kernel.org
Subject: Re: [RFC][PATCH] almost certain bug in
 drivers/crypto/chelsio/chcr_algo.c:create_authenc_wr()

	Another fairly bug in there (this time in
drivers/crypto/chelsio/chtls/chtls_io.c):

/* Read TLS header to find content type and data length */
static int tls_header_read(struct tls_hdr *thdr, struct iov_iter *from)  
{
        if (copy_from_iter(thdr, sizeof(*thdr), from) != sizeof(*thdr))
                return -EFAULT;
        return (__force int)cpu_to_be16(thdr->length);
}

For one thing, that kind of force-casts is pretty much always wrong.
This one clearly says "should've used be16_to_cpu()".  And that includes
annotating tls_hdr ->length as __be16; no idea about the other field
in there (->version, that is).

For another, the only caller is
                        recordsz = tls_header_read(&hdr, &msg->msg_iter);
                        size -= TLS_HEADER_LENGTH;
                        copied += TLS_HEADER_LENGTH;
                        csk->tlshws.txleft = recordsz;
                        csk->tlshws.type = hdr.type;
                        if (skb)
                                ULP_SKB_CB(skb)->ulp.tls.type = hdr.type;
which doesn't look like something that'll work if you get -EFAULT out of
that function.  If anything, that smells like we want
                        recordsz = tls_header_read(&hdr, &msg->msg_iter);
			if (recordsz < 0)
				goto do_fault;
			...

What's more, we do *not* want a header that has faulted halfway through
to be consumed.  IOW, we want copy_from_iter_full():

static int tls_header_read(struct tls_hdr *thdr, struct iov_iter *from)  
{
        if (!copy_from_iter_full(thdr, sizeof(*thdr), from))
                return -EFAULT;
        return be16_to_cpu(thdr->length);
}

in addition to that missing check in the caller...