Differences

This shows you the differences between two versions of the page.

--- soc:2010:cooldavid:journal:week10 [2010/07/29 07:33]
cooldavid
+++ soc:2010:cooldavid:journal:week10 [2010/07/30 04:14] (current)
cooldavid
@@ Line 2: / Line 2: @@
 === Shortlog of iPXE merge commits ===
+[[http://git.etherboot.org/?p=people/cooldavid/gpxe.git;a=shortlog;h=refs/heads/tcpfix|Git-branch]]
   * Guo-Fu Tseng (2):
     * [tcp] Randomize TCP bind port
@@ Line 24: / Line 24: @@
 If we want to support Selective ACK in the future, we don't want to discard
 any packet.
+[[https://git.ipxe.org/ipxe.git/commitdiff/1d3b6619e5e35eecc29efcef6eb1dd3564a2eb45|Commit diff on iPXE]]
+[[https://git.ipxe.org/ipxe.git/commitdiff/9dc51afa2cc2b0d8310b540ae193edb37f148b8b|Commit diff on iPXE]]
 <file>
 commit 1d3b6619e5e35eecc29efcef6eb1dd3564a2eb45
@@ Line 56: / Line 61: @@
  src/include/ipxe/malloc.h |   17 +++++++
 files changed, 85 insertions(+), 46 deletions(-)
+</file>
+== Generic xfer interface ==
+"deliver_raw() no longer exists as a data-xfer method"
+We might need it someday, I did think of some approach to do
+Selective ACK support with possibly using deliver_raw() for
+performance concern.
+[[https://git.ipxe.org/ipxe.git/commitdiff/4327d5d39f101f1df0ace6c03f3b3ada5f6a6213|Commit diff on iPXE]]
+<file>
+commit 4327d5d39f101f1df0ace6c03f3b3ada5f6a6213
+Author: Michael Brown <mcb30@ipxe.org>
+Date:   Wed Jun 16 01:31:29 2010 +0100
+    [interface] Convert all data-xfer interfaces to generic interfaces
+    Remove data-xfer as an interface type, and replace data-xfer
+    interfaces with generic interfaces supporting the data-xfer methods.
+    Filter interfaces (as used by the TLS layer) are handled using the
+    generic pass-through interface capability.  A side-effect of this is
+    that deliver_raw() no longer exists as a data-xfer method.  (In
+    practice this doesn't lose any efficiency, since there are no
+    instances within the current codebase where xfer_deliver_raw() is used
+    to pass data to an interface supporting the deliver_raw() method.)
+    Signed-off-by: Michael Brown <mcb30@ipxe.org>
 </file>
 ==== Re-do my previous TCP patches against merged branch ====
-aaaaa
+[[http://git.etherboot.org/?p=people/cooldavid/gpxe.git;a=shortlog;h=refs/heads/tcpfix2|Git-branch]]
-==== Post the patches on the gpxe-devel and wait for feedback ====
+<file>
-bbbbb
+commit be160f2a9581ca791f038cff0ca476f19d8fbc0e
+Author: Guo-Fu Tseng <cooldavid@cooldavid.org>
+Date:   Thu Jul 15 11:02:52 2010 +0800
+    [tcp core][RFC] Wait for TCP to safely close
+    After some discussion with Piotr and Michael, I think it might
+    be an acceptable solution for the issue that gPXE leaves remote
+    peer TCP in undesired state.
+    The reason I don't mark activity_start at somewhere like tcp_open
+    or ESTABLISHED is that sanboot do not close the TCP connection
+    before hand of the control to OS.
+ src/core/activity.c         |   64 +++++++++++++++++++++++++++++++++++++++++++
+ src/core/image.c            |    8 +++++
+ src/core/main.c             |    2 +
+ src/include/gpxe/activity.h |   28 +++++++++++++++++++
+ src/include/gpxe/tcp.h      |   13 +++++++++
+ src/net/tcp.c               |   15 ++++++++++
+ src/usr/autoboot.c          |    6 ++++
+files changed, 136 insertions(+), 0 deletions(-)
+</file>
+<file>
+commit 4c985888acf02acc9bf2d94925279e7a14cbc779
+Author: Guo-Fu Tseng <cooldavid@cooldavid.org>
+Date:   Wed May 12 09:26:08 2010 +0800
+    [jme] Adding hardware RX checksum offload support
+    Read corresponding flags for RX checksum status, and update the
+    csum_stat field of the struct io_buffer.
+ src/drivers/net/jme.c |   24 ++++++++++++++++++++++++
+files changed, 24 insertions(+), 0 deletions(-)
+</file>
+<file>
+commit 134fe7eef6f1e1a4603d58524e0b4e73265ea9cb
+Author: Guo-Fu Tseng <cooldavid@cooldavid.org>
+Date:   Wed May 12 09:22:11 2010 +0800
+    [net] Adding RX packet checksum offload support
+    Add a field in struct io_buffer to pass checksum status from hardware.
+    If the hardware dose not support just ignore this field, it'll work as
+    usual.
+    I see that net/ipv6 does not check for checksum, so I temporary ignored
+    it.
+ src/core/iobuf.c         |    1 +
+ src/include/gpxe/iobuf.h |   13 +++++++++++++
+ src/net/ipv4.c           |    8 +++++---
+ src/net/tcp.c            |   18 ++++++++++--------
+ src/net/udp.c            |    2 +-
+files changed, 30 insertions(+), 12 deletions(-)
+</file>
+<file>
+commit 155da1968f5e89c38cd5f7d76c4c44c9784f83ff
+Author: Guo-Fu Tseng <cooldavid@cooldavid.org>
+Date:   Fri Jul 30 16:08:28 2010 +0800
+    [tcp] Cleanup tcp_close()
+    The complexity of tcp_close() makes the TCP closing control
+    hard to maintain.
+    This patch try to solve it by separating tcp_terminate() and
+    tcp_dataxfer_close() from tcp_close().
+    Use tcp_terminate() to immediately free all tcp resources, and
+    close data transfer interface by calling tcp_dataxfer_close().
+    tcp_close() now simply bing TCP connection to closing state.
+    Which makes it safe to call in tcp_rx_fin().
+    tcp_dataxfer_close() is separated due to early close timing of
+    tcp_xfer_close() which is called by upper-layer to indicate no
+    more data to send or receive.
+ src/net/tcp.c |  104 +++++++++++++++++++++++++++++---------------------------
+files changed, 54 insertions(+), 50 deletions(-)
+</file>
+<file>
+commit 7345fc858e6ee2f414ee8328b710b6a333a2be19
+Author: Guo-Fu Tseng <cooldavid@cooldavid.org>
+Date:   Thu Jul 29 21:43:05 2010 +0800
+    [tcp] Distinguish passive and active close with proper actions
+    The original design of TCP stack dose not distinguish passive and active
+    close, support is to make close action more compatible with RFC 793 and
+    having the ability to benefit from passive closing that can release all
+    resources immediately.
+ src/include/gpxe/tcp.h |   43 ++++++++++++++++++++++++++++++++++++-------
+ src/net/tcp.c          |   26 +++++++++++++++++++++-----
+files changed, 57 insertions(+), 12 deletions(-)
+</file>
+<file>
+commit bdee331b5ba32557bbd092987be0af3cc1c45479
+Author: Guo-Fu Tseng <cooldavid@cooldavid.org>
+Date:   Thu Jul 29 21:16:24 2010 +0800
+    [tcp] Deliver data only after updating TCP state
+    Michael suggested something like:
+    ===============================================================
+      struct list_head received = LIST_HEAD_INIT ( received );
+      ...
+      while ( ! list_empty ( &tcp->rx_queue ) ) {
+         ...
+         tcp_rx_data ( tcp, seq, iob_disown ( iobuf ), &received );
+         ...
+      }
+      list_for_each_entry_safe ( iobuf, tmp, &received, list ) {
+        // deliver iobuf via xfer_deliver_iob()
+      }
+      if ( tcp->state & TCP_STATE_RCVD ( TCP_FIN ) )
+        tcp_close ( tcp, 0 );
+    ===============================================================
+    But after some thought I think making the change like this commit
+    can have same behavior, but simplified by not using extra queue.
+    Which can also save some code size.
+    In this patch:
+. We call xfer_deliver_iob() after fully updated/handled received TCP
+         state. In which to have better timing that upper-layer protocol
+         might call tcp_xfer_close() to initiate a shutdown, or sending
+         extra data.
+. We move tcp_close() out from tcp_rx_fin() because tcp_close()
+         nullified the xfer interface. It'll cause error to call
+         xfer_deliver_iob() after tcp_rx_fin().
+         (So that I proposed an idea to clean up tcp_close(), I'll post it
+          in later commit.)
+ src/net/tcp.c |   34 +++++++++++++++++++---------------
+files changed, 19 insertions(+), 15 deletions(-)
+</file>
+==== Post the patches on the gpxe-devel and wait for feedback ====