netdev - Re: [PATCH] bnx2x: fix occasional statistics off-by-4GB error

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: <CAHo-Ooz2W_jcbbOFLvihf+N-6fVSNoj0jGKTT8AXoOSC3_3DRg@mail.gmail.com>
Date:	Mon, 18 Mar 2013 13:36:21 -0700
From:	Maciej Żenczykowski <zenczykowski@...il.com>
To:	David Miller <davem@...emloft.net>
Cc:	eilong@...adcom.com, eric.dumazet@...il.com, dmitry@...adcom.com,
	netdev@...r.kernel.org, yuvalmin@...adcom.com
Subject: Re: [PATCH] bnx2x: fix occasional statistics off-by-4GB error

Not quite the right call sequence above.
  UPDATE_FSTAT_QSTAT(total_bytes_received); --> SUB_64 --> DIFF_64
is probably more relevant.

Regardless:

/* difference = minuend - subtrahend */
#define DIFF_64(d_hi, m_hi, s_hi, d_lo, m_lo, s_lo) \
<------>do { \
<------><------>if (m_lo < s_lo) { \
<------><------><------>/* underflow */ \
<------><------><------>d_hi = m_hi - s_hi; \
<------><------><------>if (d_hi > 0) { \
<------><------><------><------>/* we can 'loan' 1 */ \
<------><------><------><------>d_hi--; \
<------><------><------><------>d_lo = m_lo + (UINT_MAX - s_lo) + 1; \
<------><------><------>} else { \
<------><------><------><------>/* m_hi <= s_hi */ \
<------><------><------><------>d_hi = 0; \
<------><------><------><------>d_lo = 0; \
<------><------><------>} \
<------><------>} else { \
...
I believe this fails.  All parameters are most likely going to be u32,
since that's used for stats pretty much everywhere.
As such after d_hi = m_hi - s_hi;  d_hi will be >= 0 since it's u32.

As such if "m_hi == s_hi" && "m_lo < s_lo" we will return (0,0)
however if "m_hi < s_hi" && "m_lo < s_lo" we will not return (0,0)

I'm not sure which behaviour is desired, but either way this is obviously wrong.

0 - 1 returns 0
0 - (4GB+1) returns -4GB-1
--
To unsubscribe from this list: send the line "unsubscribe netdev" in
the body of a message to majordomo@...r.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html