DPDK patches and discussions
 help / color / mirror / Atom feed
From: "Prathyusha, Guduri" <Guduri.Prathyusha@cavium.com>
To: "Tu, LijuanX A" <lijuanx.a.tu@intel.com>
Cc: "dev@dpdk.org" <dev@dpdk.org>,
	"Chen, WeichunX" <weichunx.chen@intel.com>,
	 "Liu, Yu Y" <yu.y.liu@intel.com>,
	"Xu, Qian Q" <qian.q.xu@intel.com>,
	"Liu, Yong" <yong.liu@intel.com>,
	"Lu, PeipeiX" <peipeix.lu@intel.com>
Subject: Re: [dpdk-dev] dpdk-devbind can't be used misc with kernel tools
Date: Thu, 27 Apr 2017 07:30:40 +0000	[thread overview]
Message-ID: <BY2PR0701MB21012330871CB2981591296C8F100@BY2PR0701MB2101.namprd07.prod.outlook.com> (raw)
In-Reply-To: <38D041F150D4184C8114E499040E62343DBC19F2@SHSMSX103.ccr.corp.intel.com>

Hi LijuanX A,


Thanks for verifying it. I have already submitted a formal patch for the same

http://dpdk.org/dev/patchwork/patch/23921/


Regards,

Prathyusha









________________________________
From: Tu, LijuanX A <lijuanx.a.tu@intel.com>
Sent: Thursday, April 27, 2017 12:46:58 PM
To: Prathyusha, Guduri
Cc: dev@dpdk.org; Chen, WeichunX; Liu, Yu Y; Xu, Qian Q; Liu, Yong; Lu, PeipeiX
Subject: RE: dpdk-devbind can't be used misc with kernel tools

Hi Prathyusha,

The patch you created apply failed . I use the patch as attachment.
The tested result is OK ,please create another patch than can apply to DPDK.org

Thanks,
Lijuan

-----Original Message-----
From: Prathyusha, Guduri [mailto:Guduri.Prathyusha@cavium.com]
Sent: Wednesday, April 26, 2017 5:26 PM
To: Tu, LijuanX A
Cc: dev@dpdk.org; Chen, WeichunX; Liu, Yu Y; Xu, Qian Q; Liu, Yong; Lu, PeipeiX
Subject: Re: dpdk-devbind can't be used misc with kernel tools


Hi LijuanX A

Response inline...

Kindly apply the patch below (inline) and let me know the result. I will submit a formal patch once you confirm.

Thanks,
Prathyusha




From: Tu, LijuanX A <lijuanx.a.tu@intel.com>
Sent: Wednesday, April 26, 2017 11:49 AM
To: Prathyusha, Guduri
Cc: dev@dpdk.org; Chen, WeichunX; Liu, Yu Y; Xu, Qian Q; Liu, Yong; Lu, PeipeiX
Subject: dpdk-devbind can't be used misc with kernel tools


Hi Guduri,

I am a tester from intel dpdk team. I get a issues on usertools/dpdk-devbind.py

With the usertools/dpdk-devbind.py , I can't bind driver as expect.

I use the "dpdk-devbind.py" bind pci to igb_uio, then I using kernel tools  bind pci  to ixgbe, I can bind pci to igb_uio successfully ,but it  bind back to ixgbe failed..

/sys/bus/pci/devices/xxxxxx/driver_override is used to bind and unbind devices to drivers in kernels  >= 3.15 driver_override must be written to null for a device to bind to a driver. In the current script, null is written before unbinding a device.
Hence when you bind a device using script and unbind using kernel tools (/sys/bus/pci/drivers/xx/new_id, >>/sys/bus/pci/devices/xx/driver/unbind, /sys/bus/pci/drivers/xx/bind), driver_override is still set to the previous driver.
To avoid this, writing null to driver_override after binding a device to a driver. This will let the kernel tools unbind after binding using the dpdk script.

Bind pci to igb_uio and then  bind to ixgbe ,both use "dpdk-devbind.py", it works well.

Could you . have a look at this as soon as possible ,it block the daily regression test.
Thank you very much.

Here is the patch to be applied

diff --git a/usertools/dpdk-devbind.py b/usertools/dpdk-devbind.py
index bb4d536e0..1dc1065b1 100755
--- a/usertools/dpdk-devbind.py
+++ b/usertools/dpdk-devbind.py
@@ -386,25 +386,6 @@ def unbind_one(dev_id, force):
               "Skipping unbind" % (dev_id))
         return

-    # For kernels > 3.15 driver_override is used to bind a device to a driver.
-    # Before unbinding it, overwrite driver_override with empty string so that
-    # the device can be bound to any other driver
-    filename = "/sys/bus/pci/devices/%s/driver_override" % dev_id
-    if os.path.exists(filename):
-        try:
-            f = open(filename, "w")
-        except:
-            print("Error: unbind failed for %s - Cannot open %s"
-                  % (dev_id, filename))
-            sys.exit(1)
-        try:
-            f.write("\00")
-            f.close()
-        except:
-            print("Error: unbind failed for %s - Cannot open %s"
-                  % (dev_id, filename))
-            sys.exit(1)
-
     # write to /sys to unbind
     filename = "/sys/bus/pci/drivers/%s/unbind" % dev["Driver_str"]
     try:
@@ -507,6 +488,25 @@ def bind_one(dev_id, driver, force):
             bind_one(dev_id, saved_driver, force)
         return

+    # For kernels > 3.15 driver_override is used to bind a device to a driver.
+    # Before unbinding it, overwrite driver_override with empty string so that
+    # the device can be bound to any other driver
+    filename = "/sys/bus/pci/devices/%s/driver_override" % dev_id
+    if os.path.exists(filename):
+        try:
+            f = open(filename, "w")
+        except:
+            print("Error: unbind failed for %s - Cannot open %s"
+                  % (dev_id, filename))
+            sys.exit(1)
+        try:
+            f.write("\00")
+            f.close()
+        except:
+            print("Error: unbind failed for %s - Cannot open %s"
+                  % (dev_id, filename))
+            sys.exit(1)
+

 def unbind_all(dev_list, force=False):
     """Unbind method, takes a list of device locations"""


There are my test env and steps:

dpdk commit eba33e87ad37626604be7186e746751f99691084
Components: usertools/dpdk-devbind.py
kernel: 4.8.6-300.fc25.x86_64
driver: ixgbe
version: 5.0.4
firmware-version: 0x61bf0001
Expect: we can use dpdk-devbind.py to bind or unbind PCI-device, we also can use kernel tools to bind or unbind PCI-device, such as :
steps:
# ./dpdk-devbind.py --bind=igb_uio 0000:05:00.0
status:
0000:05:00.0 '82599ES 10-Gigabit SFI/SFP+ Network Connection 10fb' drv=igb_uio unused=

# echo "8086 10fd" >/sys/bus/pci/drivers/ixgbe/new_id
# echo "0000:05:00.0" >/sys/bus/pci/devices/0000\:05\:00.0/driver/unbind
# echo "0000:05:00.0" >/sys/bus/pci/drivers/ixgbe/ # echo "0000:05:00.0" >/sys/bus/pci/drivers/ixgbe/bind
-bash: echo: write error: No such device status
0000:05:00.0 '82599ES 10-Gigabit SFI/SFP+ Network Connection' unused=ixgbe,igb_uio
Result:
It can't bind to ixgbe, expect it can bind to ixgbe.
I think the related commit are ::
commit 2fc3502935700243d9a6d903166e6fd11e429843
Author: Guduri Prathyusha <gprathyusha@caviumnetworks.com>
Date:   Wed Mar 22 19:41:29 2017 +0530
usertools: use optimized driver override scheme to bind

commit c3ce205d5729867bd1c4c4429a80e01a528d5905
Author: Guduri Prathyusha <gprathyusha@caviumnetworks.com>
Date:   Wed Mar 22 19:41:28 2017 +0530
    usertools: optimize lspci invocation

      reply	other threads:[~2017-04-27  7:30 UTC|newest]

Thread overview: 5+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2017-04-26  6:19 Tu, LijuanX A
2017-04-26  9:26 ` Prathyusha, Guduri
2017-04-27  6:42   ` Tu, LijuanX A
2017-04-27  7:16   ` Tu, LijuanX A
2017-04-27  7:30     ` Prathyusha, Guduri [this message]

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=BY2PR0701MB21012330871CB2981591296C8F100@BY2PR0701MB2101.namprd07.prod.outlook.com \
    --to=guduri.prathyusha@cavium.com \
    --cc=dev@dpdk.org \
    --cc=lijuanx.a.tu@intel.com \
    --cc=peipeix.lu@intel.com \
    --cc=qian.q.xu@intel.com \
    --cc=weichunx.chen@intel.com \
    --cc=yong.liu@intel.com \
    --cc=yu.y.liu@intel.com \
    /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;
as well as URLs for NNTP newsgroup(s).