diff options
author | Ali Saidi <Ali.Saidi@ARM.com> | 2012-09-07 14:20:53 -0500 |
---|---|---|
committer | Ali Saidi <Ali.Saidi@ARM.com> | 2012-09-07 14:20:53 -0500 |
commit | 5217d5a451322199b4165ee0293ed2681dae5da3 (patch) | |
tree | f9a03a52fcf38177872bf96483462a8b143456c3 /src/dev/i8254xGBe.cc | |
parent | 124aac72b41f2dd29778168f808a9d08282abcfe (diff) | |
download | gem5-5217d5a451322199b4165ee0293ed2681dae5da3.tar.xz |
Igbe: Newer kernels seem to allow TSO headers and packet data to be in one desc
Implement some code we used to panic on as it actually does happen with the
e1000 driver in Linux 3.3+. We used to assume that a TSO header would never
be part of a larger payload, however it appears as though it now can be.
Diffstat (limited to 'src/dev/i8254xGBe.cc')
-rw-r--r-- | src/dev/i8254xGBe.cc | 8 |
1 files changed, 3 insertions, 5 deletions
diff --git a/src/dev/i8254xGBe.cc b/src/dev/i8254xGBe.cc index 71b88377d..d2b7c0f75 100644 --- a/src/dev/i8254xGBe.cc +++ b/src/dev/i8254xGBe.cc @@ -1650,11 +1650,9 @@ IGbE::TxDescCache::headerComplete() unusedCache.pop_front(); usedCache.push_back(desc); } else { - // I don't think this case happens, I think the headrer is always - // it's own packet, if it wasn't it might be as simple as just - // incrementing descBytesUsed by the header length, but I'm not - // completely sure - panic("TSO header part of bigger packet, not implemented\n"); + DPRINTF(EthernetDesc, "TSO: header part of larger payload\n"); + tsoDescBytesUsed = tsoHeaderLen; + tsoLoadedHeader = true; } enableSm(); igbe->checkDrain(); |