From mboxrd@z Thu Jan 1 00:00:00 1970 Received: from mga02.intel.com (mga02.intel.com [134.134.136.20]) by mx.groups.io with SMTP id smtpd.web08.9908.1607001525015021349 for ; Thu, 03 Dec 2020 05:18:45 -0800 Authentication-Results: mx.groups.io; dkim=missing; spf=none, err=permanent DNS error (domain: linux.intel.com, ip: 134.134.136.20, mailfrom: maciej.rabeda@linux.intel.com) IronPort-SDR: 7jxaDZ5psB1VGfy2OlJhvfnJ+z6zXRCbSGZlg7fRjHi7jB9iJb8mbVCTvRMb5xyUr0Im2k6LHy 2U8Ta7ueAoYg== X-IronPort-AV: E=McAfee;i="6000,8403,9823"; a="160244775" X-IronPort-AV: E=Sophos;i="5.78,389,1599548400"; d="scan'208,217";a="160244775" X-Amp-Result: SKIPPED(no attachment in message) X-Amp-File-Uploaded: False Received: from orsmga005.jf.intel.com ([10.7.209.41]) by orsmga101.jf.intel.com with ESMTP/TLS/ECDHE-RSA-AES256-GCM-SHA384; 03 Dec 2020 05:18:43 -0800 IronPort-SDR: rUD9Pu5WY4p7gc/FxOoAGe/d7rOGdOySjRCnmOTfhU1ev5GEoU8mwdT9XACl0JFbTH7+pwXsCo ygZ46quJQgsQ== X-IronPort-AV: E=Sophos;i="5.78,389,1599548400"; d="scan'208,217";a="550479675" Received: from mrabeda-mobl.ger.corp.intel.com (HELO [10.252.54.77]) ([10.252.54.77]) by orsmga005-auth.jf.intel.com with ESMTP/TLS/ECDHE-RSA-AES256-GCM-SHA384; 03 Dec 2020 05:18:42 -0800 Subject: Re: [edk2-devel] reg: PCD To Control MNP Buffer Recycle Support To: devel@edk2.groups.io, sivaramann@amiindia.co.in References: <78af7771-8edf-ec2b-9db4-92ad9f506380@linux.intel.com> From: "Maciej Rabeda" Message-ID: Date: Thu, 3 Dec 2020 14:18:34 +0100 User-Agent: Mozilla/5.0 (Windows NT 10.0; Win64; x64; rv:78.0) Gecko/20100101 Thunderbird/78.5.1 MIME-Version: 1.0 In-Reply-To: Content-Type: multipart/alternative; boundary="------------FEE3343649FA94E1734FC341" Content-Language: pl --------------FEE3343649FA94E1734FC341 Content-Type: text/plain; charset=windows-1252; format=flowed Content-Transfer-Encoding: 7bit You are referring to the wrong document (PXE 2.1 spec for 16-bit UNDIs/legacy BIOS). Appendix E of UEFI specification defines UNDI functions, behaviors and supported opcodes in UEFI. Thanks, Maciej On 26-Nov-20 17:20, Sivaraman Nainar wrote: > > Hello Maciej: > > *As per the PXE Specification: * > > The protocol driver calls UNDI repeatedly with the FuncFlag equal to > PXENV_UNDI_ISR_IN_GET_NEXT to get all the buffers in a frame and also > all the received > > frames in the queue. On this call, UNDI must remember the previous > buffer given to the protoco,l remove it from the receive queue and > recycle it. In case of a multi-buffered frame, if the previous > > buffer is not the last buffer in the frame it must return the next > buffer in the frame in the parameter block. Otherwise it must return > the first buffer in the next frame. > > Assume the UNDI Driver does not properly Store that TxAddress in the > GetStatus then the possibility of MNP Layer having issues when it > treats the recycling based on the GetStatus. > > 2. If a NIC has a pre-allocated region for Tx buffers and cannot > dynamically swap them out, UNDI->Transmit() should copy the Tx buffer > into appropriate region. UNDI driver should also store that Tx buffer > somewhere within the driver, where it will be accessible by > UNDI->GetStatus() function. > > When this point 2 fails the Recycle support done in the SW Layer shall > also fail. > > Pleases correct my understanding if any . > > -Siva > > *From:* Rabeda, Maciej [mailto:maciej.rabeda@linux.intel.com] > *Sent:* Thursday, November 26, 2020 9:10 PM > *To:* devel@edk2.groups.io; Sivaraman Nainar > *Subject:* Re: [edk2-devel] reg: PCD To Control MNP Buffer Recycle Support > > Hi Nainar, > > What do you mean by "controllers do not support recycling of Tx buffers"? > MNP provides a transmit buffer to SNP (and with EDK2's SNP also to > UNDI) and at this point it is considered owned by SNP/UNDI. > After UNDI/SNP successfully transmits the packet, it is expected to > give it back via UNDI->GetStatus(). > > Adding network controller (NIC) and UNDI into the picture. > > 1. UNDI->Transmit() gets a single Tx buffer to put on the wire. > 2. If a NIC has a pre-allocated region for Tx buffers and cannot > dynamically swap them out, UNDI->Transmit() should copy the Tx buffer > into appropriate region. UNDI driver should also store that Tx buffer > somewhere within the driver, where it will be accessible by > UNDI->GetStatus() function. > 3. After successful transmit of Tx buffer, on next UNDI->GetStatus() > with PXE_OPFLAGS_GET_TRANSMITTED_BUFFERS OpFlag, driver should put > that Tx buffer address in PXE_CDB.DBAddr provided to UNDI->GetStatus() > as a parameter. > > Summing the above - Tx buffer recycling is purely a SW feature. I do > not understand your request nor find it valid. > > Thanks, > Maciej > > On 24-Nov-20 15:30, Sivaraman Nainar wrote: > > Hello all: > > MNPDxe supported with the Recycle buffer support from the below > git commit, > > https://github.com/tianocore/edk2/commit/0507449955c5c629cec196b62986afbb91203ed9#diff-fb5b97ad38efea22f5ddd745f6e43ebdb509dc4a5aef81997ba53af5f918a47b > > > But many network controllers does not support the Recycling Tx > Buffers which will cause the PXE Download failure or HTTP Boot > failure. > > Can this feature can be controlled by a Dynamic PCD so that the > same code can work for controller which does not support Tx Buffer > Recycling. > > Thanks > > Siva > > --------------FEE3343649FA94E1734FC341 Content-Type: text/html; charset=windows-1252 Content-Transfer-Encoding: 8bit You are referring to the wrong document (PXE 2.1 spec for 16-bit UNDIs/legacy BIOS).
Appendix E of UEFI specification defines UNDI functions, behaviors and supported opcodes in UEFI.

Thanks,
Maciej

On 26-Nov-20 17:20, Sivaraman Nainar wrote:

Hello Maciej:

 

As per the PXE Specification:

The protocol driver calls UNDI repeatedly with the FuncFlag equal to PXENV_UNDI_ISR_IN_GET_NEXT to get all the buffers in a frame and also all the received

frames in the queue. On this call, UNDI must remember the previous buffer given to the protoco,l remove it from the receive queue and recycle it. In case of a multi-buffered frame, if the previous

buffer is not the last buffer in the frame it must return the next buffer in the frame in the parameter block. Otherwise it must return the first buffer in the next frame.

 

Assume the UNDI Driver does not properly Store that TxAddress in the GetStatus then the possibility of MNP Layer having issues when it treats the recycling based on the GetStatus.

2. If a NIC has a pre-allocated region for Tx buffers and cannot dynamically swap them out, UNDI->Transmit() should copy the Tx buffer into appropriate region. UNDI driver should also store that Tx buffer somewhere within the driver, where it will be accessible by UNDI->GetStatus() function.

When this point 2 fails the Recycle support done in the SW Layer shall also fail.

 

Pleases correct my understanding if any .

 

-Siva

From: Rabeda, Maciej [mailto:maciej.rabeda@linux.intel.com]
Sent: Thursday, November 26, 2020 9:10 PM
To: devel@edk2.groups.io; Sivaraman Nainar
Subject: Re: [edk2-devel] reg: PCD To Control MNP Buffer Recycle Support

 

Hi Nainar,

What do you mean by "controllers do not support recycling of Tx buffers"?
MNP provides a transmit buffer to SNP (and with EDK2's SNP also to UNDI) and at this point it is considered owned by SNP/UNDI.
After UNDI/SNP successfully transmits the packet, it is expected to give it back via UNDI->GetStatus().

Adding network controller (NIC) and UNDI into the picture.

1. UNDI->Transmit() gets a single Tx buffer to put on the wire.
2. If a NIC has a pre-allocated region for Tx buffers and cannot dynamically swap them out, UNDI->Transmit() should copy the Tx buffer into appropriate region. UNDI driver should also store that Tx buffer somewhere within the driver, where it will be accessible by UNDI->GetStatus() function.
3. After successful transmit of Tx buffer, on next UNDI->GetStatus() with PXE_OPFLAGS_GET_TRANSMITTED_BUFFERS OpFlag, driver should put that Tx buffer address in PXE_CDB.DBAddr provided to UNDI->GetStatus() as a parameter.

Summing the above - Tx buffer recycling is purely a SW feature. I do not understand your request nor find it valid.

Thanks,
Maciej

On 24-Nov-20 15:30, Sivaraman Nainar wrote:

Hello all:

 

MNPDxe supported with the Recycle buffer support from the below git commit,

 

https://github.com/tianocore/edk2/commit/0507449955c5c629cec196b62986afbb91203ed9#diff-fb5b97ad38efea22f5ddd745f6e43ebdb509dc4a5aef81997ba53af5f918a47b

 

But many network controllers does not support the Recycling Tx Buffers which will cause the PXE Download failure or HTTP Boot failure.

 

Can this feature can be controlled by a Dynamic PCD so that the same code can work for controller which does not support Tx Buffer Recycling.

 

Thanks

Siva

 


--------------FEE3343649FA94E1734FC341--