aboutsummaryrefslogtreecommitdiff
path: root/doc
diff options
context:
space:
mode:
authorHaren Myneni <haren@linux.vnet.ibm.com>2017-07-17 19:51:48 -0700
committerStewart Smith <stewart@linux.vnet.ibm.com>2017-08-01 12:39:43 +1000
commitd689cb96ac58681bcff087e31162ff92fefd411d (patch)
tree4ec1361a41693ff26c3ffc1a3a6bd36f9879609c /doc
parent8a665d6e1d1cd5e4491cfa44f4b8e80e8ce543f6 (diff)
downloadskiboot-d689cb96ac58681bcff087e31162ff92fefd411d.zip
skiboot-d689cb96ac58681bcff087e31162ff92fefd411d.tar.gz
skiboot-d689cb96ac58681bcff087e31162ff92fefd411d.tar.bz2
NX: Describe device-tree changes for P9 NX compression engines
Describe device-tree entries needed for 842 and gzip compression engines on P9 NX. Signed-off-by: Haren Myneni <haren@us.ibm.com> Signed-off-by: Stewart Smith <stewart@linux.vnet.ibm.com>
Diffstat (limited to 'doc')
-rw-r--r--doc/device-tree/nx.rst37
1 files changed, 33 insertions, 4 deletions
diff --git a/doc/device-tree/nx.rst b/doc/device-tree/nx.rst
index 4c54d4b..1589be6 100644
--- a/doc/device-tree/nx.rst
+++ b/doc/device-tree/nx.rst
@@ -11,11 +11,11 @@ With unique xscom and nx addresses. Their compatible node contains
"ibm,power-nx".
-NX 842 Coprocessor
-------------------
+NX Compression Coprocessor
+--------------------------
-This is the memory compression coprocessor, which uses the IBM proprietary
-842 compression algorithm and format. Each nx node contains an 842 engine.
+This is the memory compression coprocessor. which uses the IBM proprietary
+842 compression algorithm and format. Each NX node contains an 842 engine.
::
ibm,842-coprocessor-type : CT value common to all 842 coprocessors
@@ -29,6 +29,35 @@ all 842 coprocessors in the system, the CT value will (should) be the same,
while each will have a different CI value. The driver can use CI 0 to allow
the hardware to automatically select which coprocessor instance to use.
+On P9, this compression coprocessor also supports standard GZIP/ZLIB
+compression algorithm and format. Virtual Accelerator Swirchboard (VAS) is used
+to access this coprocessor. VAS writes each request to receive FIFOs (RXFIFO)
+which are either high or normal priority and these FIFOs are bound to
+coprocessor types (842 and gzip).
+
+VAS distinguishes NX requests for the target engines based on logical
+partition ID (lpid), process ID (pid) and Thread ID (tid). So (lpid, pid, tid)
+combination has to be unique in the system. Each NX node contains high and
+normal FIFOs for each 842 and GZIP engines.
+::
+ /ibm,842-high-fifo : High priority 842 RxFIFO
+ /ibm,842-normal-fifo : Normal priority 842 RxFIFO
+ /ibm,gzip-high-fifo : High priority gzip RxFIFO
+ /ibm,gzip-normal-fifo : Normal priority gzip RxFIFO
+
+Each RxFIFO node contains: ::
+ compatible : ibm,p9-nx-842 or ibm,p9-nx-gzip
+ priority : High or Normal
+ rx-fifo-address : RxFIFO buffer address
+ rx-fifo-size : RxFIFO size
+ lpid : 0xfff (1's for 12 bits in UMAC notify match
+ register)
+ pid : Coprocessor type (either 842 or gzip)
+ tid : counter in each coprocessor type
+
+During initialization, the driver invokes VAS interface for each coprocessor
+type (842 and gzip) to configure the RxFIFO with rx_fifo_address, lpid, pid
+and tid for high and nornmal priority FIFOs.
NX RNG Coprocessor
------------------