systemsoftware
/
Linux-Architecture-Reference
35
1
Fork 0
Code Issues Pull Requests Projects Releases Wiki Activity

Continued Extension of H_Send_Logical_LAN for large send packets

Browse Source 
Signed-off-by: Jeff Scheel <scheel@us.ibm.com>
pull/1/head papr_2.9+
Jeff Scheel 5 years ago
parent 2f01076395
commit 87ca3cb207
1 changed files with 144 additions and 35 deletions
Show Stats Download Patch File Download Diff File

151
Virtualization/ch_virtual_io.xml
Unescape Escape View File

@ -9480,8 +9480,10 @@ hcall ( const uint64 H_SEND_LOGICAL_LAN, /* Send a message on the logical LAN */
uint64 buff-5, /* with a control byte of “invalid” indicates the end of */ uint64 buff-5, /* with a control byte of “invalid” indicates the end of */
uint64 buff-6, /* the message (only the number of buffers needed are */ uint64 buff-6, /* the message (only the number of buffers needed are */
/* used). */ /* used). */
uint64 continue-token /* value of 0 on first call, value returned in R4 on */ uint64 continue-token, /* value of 0 on first call, value returned in R4 on */
/* H_Busy */ /* H_Busy */
uint64 mss /* The Maximum Segment Size to use when segmenting */
/* the message */
);]]></programlisting> );]]></programlisting>


<para>The H_Dropped return code indicates to the sender that one or more <para>The H_Dropped return code indicates to the sender that one or more
@ -9527,6 +9529,11 @@ hcall ( const uint64 H_SEND_LOGICAL_LAN, /* Send a message on the logical LAN */
again but use the value returned in R4 from the previous call as the again but use the value returned in R4 from the previous call as the
value of continue-token.</para> value of continue-token.</para>
</listitem> </listitem>

<listitem>
<para>mss: The Maximum Segment Size to use if segmenting of the message
is necessary.</para>
</listitem>
</itemizedlist> </itemizedlist>
</simplesect> </simplesect>


@ -9548,10 +9555,22 @@ hcall ( const uint64 H_SEND_LOGICAL_LAN, /* Send a message on the logical LAN */
proceeding.</para> proceeding.</para>
</listitem> </listitem>


<listitem>
<para>If the caller has enabled large send/receive, verify the MSS is
0 or at least 16 – else H_Parameter.</para>
</listitem>

<listitem> <listitem>
<para>Verifies the VLAN number -- else H_Parameter.</para> <para>Verifies the VLAN number -- else H_Parameter.</para>
</listitem> </listitem>


<listitem>
<para>If the caller has enabled large send/receive and the MSS is not 0,
verify the message contains an Ethernet packet using IPv4 or IPv6 and
TCP – else H_Parameter. The complete TCP header must be within the
first 4096 bytes of the frame.</para>
</listitem>

<listitem> <listitem>
<para>Proceeds down the 6 buffer descriptors until the first one that <para>Proceeds down the 6 buffer descriptors until the first one that
has a length of 0</para> has a length of 0</para>
@ -9642,6 +9661,15 @@ hcall ( const uint64 H_SEND_LOGICAL_LAN, /* Send a message on the logical LAN */


</listitem> </listitem>


<listitem>
<para>If the MSS parameter is being used, the receiver IOA is not enabled for large receive, and the TCP payload in the frame is larger than the value of the mss parameter, the TCP payload of the message is divided into separate messages with copies of the MAC, VLAN, IP, and TCP headers in each.</para>
</listitem>

<listitem>
<para>For each message segment (packets without an MSS will be sent
as a single segment):</para>

<itemizedlist>
<listitem> <listitem>
<para>Searches the receiver&#8217;s receive queue for a suitable buffer <para>Searches the receiver&#8217;s receive queue for a suitable buffer
and atomically dequeues it:</para> and atomically dequeues it:</para>
@ -9655,6 +9683,37 @@ hcall ( const uint64 H_SEND_LOGICAL_LAN, /* Send a message on the logical LAN */
</itemizedlist> </itemizedlist>
</listitem> </listitem>


<listitem>
<para>If the receiver has enabled large-receive support, write
the MSS value in the second 8 bytes of the selected receive
buffer.</para>
</listitem>

<listitem>
<para>If segmentation is required, firmware will update portions
of the IP and TCP headers as necessary.

<itemizedlist>
<listitem>
<para>If the packet is using IPv4, firmware will always
regenerate the checksum for each segment.</para>
</listitem>

<listitem>
<para>For each segment, firmware will increment the IP
ident field by 1. It is the sender's responsibility
to ensure that no other packets are sent with the
colliding ident values.</para>
</listitem>

<listitem>
<para>If the receiver has not enabled checksum offload,
firmware will update the checksum in the TCP header.</para>
</listitem>
</itemizedlist>
</para>
</listitem>

<listitem> <listitem>
<para>Copy the send data in to the selected receive buffer, build a <para>Copy the send data in to the selected receive buffer, build a
receive queue entry, and generate an interrupt to the receiver if the receive queue entry, and generate an interrupt to the receiver if the
@ -9665,8 +9724,13 @@ hcall ( const uint64 H_SEND_LOGICAL_LAN, /* Send a message on the logical LAN */


<listitem> <listitem>
<para>If any frames were dropped return H_Dropped else return <para>If any frames were dropped return H_Dropped else return
H_Success.</para> H_Success. This may happen if the receiver is low on buffers of the
correct size.</para>
</listitem> </listitem>

</itemizedlist>
</listitem>

</itemizedlist> </itemizedlist>


<para><emphasis role="bold">Firmware Implementation Note:</emphasis> If during the <para><emphasis role="bold">Firmware Implementation Note:</emphasis> If during the
@ -10005,7 +10069,9 @@ hcall ( const uint64 H_CHANGE_LOGICAL_LAN_MAC, /* Change the MAC address */
<para>This bit is implemented when PHYP supports sending TCP <para>This bit is implemented when PHYP supports sending TCP
packets with a non-zero TCP checksum field when bit 6 of the packets with a non-zero TCP checksum field when bit 6 of the
buffer descriptor (the "No Checksum" bit) is set. This bit buffer descriptor (the "No Checksum" bit) is set. This bit
indicates R1–17.3.6.2.2–3 is not required.</para> indicates
<xref linkend="req_illan_attributes_checksum_offload_3" xrefstyle="select: labelnumber nopage" />
is not required.</para>
</entry> </entry>
</row> </row>
<row> <row>
@ -10013,10 +10079,30 @@ hcall ( const uint64 H_CHANGE_LOGICAL_LAN_MAC, /* Change the MAC address */
<para>47</para> <para>47</para>
</entry> </entry>
<entry> <entry>
<para>Reserved</para> <para>Large send/receive enabled</para>
</entry> </entry>
<entry> <entry>
<para>&#160;</para> <para>This bit is implemented for a VIOA whenever the Large Send option is
implemented for TCP. This bit is initially set to 0 by the firmware and
the ILLAN DD may attempt to set it to 1 by use of the H_ILLAN_ATTRIBUTES
hcall() if the DD supports the option to receive TCP packets sent using
H_SEND_LOGICAL_LAN with a non-zero MSS. Firmware will not allow changing
the state of this bit if it does not support Large Send as indicated by
bit 48 for the VIOA (H_Constrained would be returned in this case from
the H_ILLAN_ATTRIBUTES hcall() when this bit is a 1 in the set-mask).
The state of this bit cannot be changed between the time that the ILLAN is
registered by an H_REGISTER_LOGICAL_LAN and it is de-registered by
an H_FREE_LOGICAL_LAN. See
<xref linkend="sec_illan_large_send_indication" />
for more information.<?linebreak?>
1: The partition software has indicated that it supports receiving
large-send packets sent using H_SEND_LOGICAL_LAN with a non-zero MSS.
If no buffers are available to fit the large packet, it will be
dropped.<?linebreak?>
0: The partition software has not indicated that it supports the
Large Receive option for packets sent using H_SEND_LOGICAL_LAN,
or it has but the firmware does not support the option, or supports
the option but not for this protocol or for this VIOA.</para>
</entry> </entry>
</row> </row>
<row> <row>
@ -10027,13 +10113,16 @@ hcall ( const uint64 H_CHANGE_LOGICAL_LAN_MAC, /* Change the MAC address */
<para>Large Send Indication Supported</para> <para>Large Send Indication Supported</para>
</entry> </entry>
<entry> <entry>
<para>The bit is implemented when the large send indication bit in <para>The bit is implemented when firmware supports the large send
the I/O descriptor passed to H_SEND_LOGICAL_LAN is indication bit in the I/O descriptor passed to H_SEND_LOGICAL_LAN.
supported by firmware.</para> This bit allows the partition software to enable large receive as
<para>0: Software must not request large send indication, described in
by setting Bit 5 of the buffer descriptor.</para> <xref linkend="sec_illan_large_send_indication" />.<?linebreak?>
<para>1: Software may request large send indication, by 0: Software must not request large send indication,
setting Bit 5 of the buffer descriptor.</para> by setting Bit 5 of the buffer descriptor.<?linebreak?>
1: Software may request large send indication, by
setting Bit 5 of the buffer descriptorand may supply an MSS value
greater than 0.</para>
</entry> </entry>
</row> </row>
<row> <row>
@ -10166,14 +10255,14 @@ hcall ( const uint64 H_CHANGE_LOGICAL_LAN_MAC, /* Change the MAC address */
H_REGISTER_LOGICAL_LAN and it is deregistered by an H_REGISTER_LOGICAL_LAN and it is deregistered by an
H_FREE_LOGICAL_LAN. See H_FREE_LOGICAL_LAN. See
<xref linkend="dbdoclet.50569350_53238" /> for more <xref linkend="dbdoclet.50569350_53238" /> for more
information.</para> information.<?linebreak?>
<para>1: The partition software has indicated that it supports 1: The partition software has indicated that it supports
the ILLAN Checksum Offload Support option for TCP and IPv6 the ILLAN Checksum Offload Support option for TCP and IPv6
protocol and for the above stated extension headers by using protocol and for the above stated extension headers by using
the H_ILLAN_ATTRIBUTES hcall() with this bit set to a 1 in the the H_ILLAN_ATTRIBUTES hcall() with this bit set to a 1 in the
set-mask, and the firmware has verified that it supports this set-mask, and the firmware has verified that it supports this
protocol for the option for the VIOA.</para> protocol for the option for the VIOA.<?linebreak?>
<para>0: The partition software has not indicated that it 0: The partition software has not indicated that it
supports the ILLAN Checksum Offload Support option for TCP and supports the ILLAN Checksum Offload Support option for TCP and
IPv6 protocol and for the above stated extension headers by IPv6 protocol and for the above stated extension headers by
using the H_ILLAN_ATTRIBUTES hcall() with this bit set to a 1 using the H_ILLAN_ATTRIBUTES hcall() with this bit set to a 1
@ -11428,7 +11517,7 @@ hcall ( const uint64 H_ILLAN_ATTRIBUTES,/* Returns in R4 the resulting ILLAN */
</listitem> </listitem>
</varlistentry> </varlistentry>


<varlistentry> <varlistentry xml:id="req_illan_attributes_checksum_offload_3">
<term><emphasis role="bold">R1-<xref linkend="dbdoclet.50569350_37925" <term><emphasis role="bold">R1-<xref linkend="dbdoclet.50569350_37925"
xrefstyle="select: labelnumber nopage"/>-3.</emphasis></term> xrefstyle="select: labelnumber nopage"/>-3.</emphasis></term>


@ -11562,7 +11651,7 @@ hcall ( const uint64 H_ILLAN_ATTRIBUTES,/* Returns in R4 the resulting ILLAN */
</section> </section>


<section xml:id="sec_illan_large_send_indication"> <section xml:id="sec_illan_large_send_indication">
<title>ILLAN Large Send Indication option</title> <title>ILLAN Large Send/Receive option</title>


<para>This option allows the virtual device to send an <para>This option allows the virtual device to send an
indication to the receiver that the data being sent by indication to the receiver that the data being sent by
@ -11592,6 +11681,12 @@ hcall ( const uint64 H_ILLAN_ATTRIBUTES,/* Returns in R4 the resulting ILLAN */
<para>Implement the Large Send Indication bit of the ILLAN Attributes as defined in <para>Implement the Large Send Indication bit of the ILLAN Attributes as defined in
<xref linkend="dbdoclet.50569350_11661" />.</para> <xref linkend="dbdoclet.50569350_11661" />.</para>
</listitem> </listitem>

<listitem>
<para>Handle the large send/receive enabled bit being set in the
ILLAN Attributes set mask as defined in
<xref linkend="dbdoclet.50569350_11661" />.</para>
</listitem>
</orderedlist> </orderedlist>
</listitem> </listitem>
</varlistentry> </varlistentry>
@ -11612,13 +11707,27 @@ hcall ( const uint64 H_ILLAN_ATTRIBUTES,/* Returns in R4 the resulting ILLAN */
<listitem> <listitem>
<para><emphasis role="bold">For the ILLAN Large send indication option:</emphasis> <para><emphasis role="bold">For the ILLAN Large send indication option:</emphasis>
When the Large Send Indication bit of the first buffer descriptor is set to 1, When the Large Send Indication bit of the first buffer descriptor is set to 1,
then the firmware for the H_SEND_LOGICAL_LAN hcall() must set the Large Send Indication then the firmware for the H_SEND_LOGICAL_LAN hcall() must set the Large
bit in the receiver's receive queue entry to 1 when the packet is copied to the Send Indication bit in the receiver's receive queue entry to 1 when the
destination receive buffer.</para> packet is copied to the destination receive buffer and examine the MSS
parameter. If the MSS is greater than 0 and the both the sender and
receiver have enabled large send/receive using H_ILLAN_ATTRIBUTES,
firmware will copy the MSS into the receive buffer.</para>
</listitem> </listitem>
</varlistentry> </varlistentry>
</variablelist> </variablelist>
</section> </section>

<section xml:id="sec_illan_receive_buffer">
<title>Receive Buffer</title>

<para>Bit 5 in the Receive Queue entry indicates that the buffer contains a
large-send packet. The software device driver should inspect the 8-byte
value at byte offset 8 of the receive buffer for the MSS. If that value is 0,
the MSS is stored in the packet itself. In the case that the buffer contains
a large-send packet, the Ipv4 checksum can be considered good and does not
need to be verified.</para>
</section>
</section> </section>
</section> </section>
</section> </section>

Write Preview
Loading…
Cancel
Save