From: Enrico Scholz <enrico.scholz@sigma-chemnitz.de>
To: barebox@lists.infradead.org
Cc: Enrico Scholz <enrico.scholz@sigma-chemnitz.de>
Subject: [PATCH 8/8] tftp: add some documentation about windowsize support
Date: Sun, 28 Aug 2022 16:02:31 +0200 [thread overview]
Message-ID: <20220828140231.930643-9-enrico.scholz@sigma-chemnitz.de> (raw)
In-Reply-To: <20220828140231.930643-1-enrico.scholz@sigma-chemnitz.de>
Signed-off-by: Enrico Scholz <enrico.scholz@sigma-chemnitz.de>
---
Documentation/filesystems/tftp.rst | 38 ++++++++++++++++++++++++++++++
1 file changed, 38 insertions(+)
diff --git a/Documentation/filesystems/tftp.rst b/Documentation/filesystems/tftp.rst
index a292765e2511..8929213d3c4a 100644
--- a/Documentation/filesystems/tftp.rst
+++ b/Documentation/filesystems/tftp.rst
@@ -20,3 +20,41 @@ Example:
In addition to the TFTP filesystem implementation, barebox does also have a
:ref:`tftp command <command_tftp>`.
+
+RFC 7440 "windowsize" support
+=============================
+
+barebox supports the tftp windowsize option for downloading files. It
+is not implemented for uploads.
+
+Generally, this option greatly improves the download speed (factors
+4-30 are not uncommon). But choosing a too large windowsize can have
+the opposite effect. Performance depends on:
+
+ - the network infrastructure: when the tftp server sends files with
+ 1Gb/s but there are components in the network (switches or the
+ target nic) which support only 100 Mb/s, packets will be dropped.
+
+ The lower the internal buffers of the bottleneck components, the
+ lower the optimal window size.
+
+ In practice (iMX8MP on a Netgear GS108Ev3 with a port configured to
+ 100 Mb/s) it had to be reduced to
+
+ .. code-block:: console
+ global tftp.windowsize=26
+
+ for example.
+
+ - the target network driver: datagrams from server will arive faster
+ than they can be processed and must be buffered internally. For
+ example, the `fec-imx` driver reserves place for
+
+ .. code-block:: c
+ #define FEC_RBD_NUM 64
+
+ packets before they are dropped
+
+ - partially the workload: copying downloaded files to ram will be
+ faster than burning them into flash. Latter can consume internal
+ buffers quicker so that windowsize might be reduced
--
2.37.2
next prev parent reply other threads:[~2022-08-28 14:05 UTC|newest]
Thread overview: 11+ messages / expand[flat|nested] mbox.gz Atom feed top
2022-08-28 14:02 [PATCH 0/8] tftp fixups Enrico Scholz
2022-08-28 14:02 ` [PATCH 1/8] tftp: make debug_assert() critical when selftest is enabled Enrico Scholz
2022-08-28 14:02 ` [PATCH 2/8] tftp: remove sanity check of first block Enrico Scholz
2022-08-28 14:02 ` [PATCH 3/8] tftp: split out allocation and cache initialization Enrico Scholz
2022-08-28 14:02 ` [PATCH 4/8] tftp: accept OACK + DATA datagrams only in certain states Enrico Scholz
2022-08-28 14:02 ` [PATCH 5/8] tftp: support non rfc 2347 servers Enrico Scholz
2022-08-28 14:02 ` [PATCH 6/8] tftp: do not set 'tsize' in wrq requests Enrico Scholz
2022-08-28 14:02 ` [PATCH 7/8] tftp: fix WRQ support Enrico Scholz
2022-08-28 14:02 ` Enrico Scholz [this message]
2022-08-29 10:52 ` [PATCH 0/8] tftp fixups Ahmad Fatoum
2022-08-30 7:30 ` Sascha Hauer
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
Avoid top-posting and favor interleaved quoting:
https://en.wikipedia.org/wiki/Posting_style#Interleaved_style
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to=20220828140231.930643-9-enrico.scholz@sigma-chemnitz.de \
--to=enrico.scholz@sigma-chemnitz.de \
--cc=barebox@lists.infradead.org \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox