netdev - bnx2x statistics occasionally wrong by near multiples of 4GB.

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-next>] [day] [month] [year] [list]

Message-ID: <CANP3RGdWMHv-7TZ4fvASrmz7NXkO7B2g2RCZwV8ecExUsQTQ2w@mail.gmail.com>
Date:	Fri, 15 Mar 2013 14:03:05 -0700
From:	Maciej Żenczykowski <maze@...gle.com>
To:	Mintz Yuval <yuvalmin@...adcom.com>
Cc:	Linux NetDev <netdev@...r.kernel.org>
Subject: bnx2x statistics occasionally wrong by near multiples of 4GB.

bnx2x_stats.h:

#define UPDATE_QSTAT(s, t) \
     do { \
              qstats->t##_hi = qstats_old->t##_hi + le32_to_cpu(s.hi); \
              qstats->t##_lo = qstats_old->t##_lo + le32_to_cpu(s.lo); \
      } while (0)

Looks wrong because it doesn't deal with overflow of _lo.

I'm not 100% sure if I tracked it down correctly, but I think this is
the root cause of us occasionally seeing stats (/proc/net/dev tx_bytes
in this case) jump back by roughly N*4GB and then go forward again.

In particular we saw successive reads of /proc/net/dev eth0 Transmit bytes show:

0x8F6 45356E60
0x8F4 4C8D72B9
0x8F6 5420D096

(which since this is a sum of 3 values (ucase/mcast/bcast) from all
queues, means we actually wrongly accounted for overflow twice...)

- Maciej
--
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