public inbox for devel@edk2.groups.io
 help / color / mirror / Atom feed
From: "Wu, Hao A" <hao.a.wu@intel.com>
To: "devel@edk2.groups.io" <devel@edk2.groups.io>,
	"jbrasen@nvidia.com" <jbrasen@nvidia.com>
Cc: "Ni, Ray" <ray.ni@intel.com>,
	"Kinney, Michael D" <michael.d.kinney@intel.com>,
	"Gao, Liming" <gaoliming@byosoft.com.cn>,
	"Zeng, Star" <star.zeng@intel.com>
Subject: Re: [edk2-devel] [PATCH] MdeModulePkg/ScsiDisk: Change TPL to NOTIFY
Date: Thu, 23 Jun 2022 02:46:47 +0000	[thread overview]
Message-ID: <DM6PR11MB4025995904C7C20CC66E4966CAB59@DM6PR11MB4025.namprd11.prod.outlook.com> (raw)
In-Reply-To: <DS7PR12MB5789F49B3FE7EF65CCCEE00ACBAF9@DS7PR12MB5789.namprd12.prod.outlook.com>

Hello,

Several concerns:
1. Raising the TPL to NOTIFY level for blocking/synchronous BlockIO service will break the using case when the co-existence of synchronous and non-blocking/asynchronous IO tasks
As far as I know, in DiskIoDxe, blocking I/O request will wait for all the synchronous I/O requests to complete first before sending down the synchronous request.
Please consider the scenario where a previous asynchronous IO task, it will take a long time to complete due to big data transfer size.
When a subsequent synchronous IO task is submitted before the asynchronous task completes, it will wait for the previous asynchronous task to complete.
However, if the TPL level for this synchronous task is NOTIFY, it will get stuck since the completion callback for the asynchronous task is also at NOTIFY TPL level, which will never be called.
More details can be referred to commit:
SHA-1: a717086c5f973821b9b49646a4ec725f6b898bdb
* MdeModulePkg ScsiDiskDxe: Raise the Tpl of async IO callback to TPL_NOTIFY

2. My personal take is that BlockIO is not a proper base service for variable service 
To my knowledge, the variable service will still exist during runtime, but the BlockIO services (as far as I know) do not.
My understanding is that the main purpose of the BlockIO services are to provide BIOS boot codes with the ability to locate partitions and file systems on media storage devices and search boot options on them.
I would recommend to implement a dedicated platform driver to produce the FVB services for variable driver consumption.

If you think using the BlockIO service as the base for variable service is the right direction, I have added people in the CC list that might help.
Could you please reach out to them for feedbacks on the TPL restriction on BlockIO services and Variable service?

Best Regards,
Hao Wu

> -----Original Message-----
> From: devel@edk2.groups.io <devel@edk2.groups.io> On Behalf Of Jeff
> Brasen via groups.io
> Sent: Friday, June 17, 2022 11:37 PM
> To: Wu, Hao A <hao.a.wu@intel.com>; devel@edk2.groups.io
> Cc: Ni, Ray <ray.ni@intel.com>
> Subject: Re: [edk2-devel] [PATCH] MdeModulePkg/ScsiDisk: Change TPL to
> NOTIFY
> 
> Resending this as I replying via edk2.groups.io doesn't seem to copy
> maintainers.
> 
> Resuming this patch to see if there are any additional thoughts on this.
> 
> In response to the query about DXE/BDS services we have some internal
> connection logic that runs in DXE to connect the devices that are needed for
> arch services that have to be connected prior the end of dxe.
> 
> Thanks,
> Jeff
> 
> > -----Original Message-----
> > From: Jeff Brasen
> > Sent: Tuesday, December 14, 2021 9:48 PM
> > To: Wu, Hao A <hao.a.wu@intel.com>; devel@edk2.groups.io
> > Cc: Ni, Ray <ray.ni@intel.com>
> > Subject: RE: [PATCH] MdeModulePkg/ScsiDisk: Change TPL to NOTIFY
> >
> >
> >
> > > -----Original Message-----
> > > From: Wu, Hao A <hao.a.wu@intel.com>
> > > Sent: Tuesday, December 14, 2021 8:00 PM
> > > To: Jeff Brasen <jbrasen@nvidia.com>; devel@edk2.groups.io
> > > Cc: Ni, Ray <ray.ni@intel.com>
> > > Subject: RE: [PATCH] MdeModulePkg/ScsiDisk: Change TPL to NOTIFY
> > >
> > > External email: Use caution opening links or attachments
> > >
> > >
> > > > -----Original Message-----
> > > > From: Jeff Brasen <jbrasen@nvidia.com>
> > > > Sent: Wednesday, December 15, 2021 1:59 AM
> > > > To: devel@edk2.groups.io
> > > > Cc: Wu, Hao A <hao.a.wu@intel.com>; Ni, Ray <ray.ni@intel.com>;
> > > > Jeff Brasen <jbrasen@nvidia.com>
> > > > Subject: [PATCH] MdeModulePkg/ScsiDisk: Change TPL to NOTIFY
> > > >
> > > > Increase TPL to TPL_NOTIFY to allow for use if caller is > TPL_CALLBACK.
> > > > This allows services like variable services that run at TPL_NOTIFY
> > > > to be hosted on ScsiDisks (i.e. UFS)
> > > >
> > > > Aligns with the eMMC driver that also uses a higher TPL.
> > > > This change was made in
> 3b1d8241d0dac25c5e678c364fa2754ac1731060
> > >
> > >
> > > Sorry, my take is that this change is not equivalent to the one made
> > > in the SD/MMC stack.
> > >
> > > For the SD/MMC change you mentioned (commit
> > > 3b1d8241d0dac25c5e678c364fa2754ac1731060), the TPL is raised to
> > > TPL_NOTIFY only when:
> > >   a) Operation to the linked lists that manage the asynchronous IO tasks
> > >   b) Callback functions that process the asynchronous IO tasks The
> > > TPL remains TPL_CALLBACK during the BlockIO services and the
> > > majority of the
> > > BlockIO2 services (operations to asynchronous tasks linked list are
> > > the exceptions).
> > >
> > > But the proposed change in ScsiDisk modifies the TPL level of the
> > > entire
> > > BlockIO/BlockIO2 (and other protocols) services to TPL_NOTIFY.
> > > For me, this is not aligned with the "TPL Restrictions" documented
> > > in the UEFI specification.
> > >
> > > Best Regards,
> > > Hao Wu
> > >
> > >
> >
> > I had sent out a query on this before and didn't see any response. The
> > core of the issue I am trying to solve it support variable services on a UFS
> device.
> > When the UFS blockIO is invoked from variable services it is not
> > allowed (which does align from the UEFI spec perspective but does not
> > allow me to implement variables services on UFS)
> >
> >  The other way that worked was lowering the lock TPL level in the PCD
> > driver and Variable down to callback. The PCD one seems like it should
> > be done as variable services is supposed to only be called from <=
> TPL_CALLBACK.
> > However, I was worried about that having a larger system impact on
> > that change.
> >
> > > >
> > > > Signed-off-by: Jeff Brasen <jbrasen@nvidia.com>
> > > > ---
> > > >  MdeModulePkg/Bus/Scsi/ScsiDiskDxe/ScsiDisk.c | 22
> > > > ++++++++++----------
> > > >  1 file changed, 11 insertions(+), 11 deletions(-)
> > > >
> > > > diff --git a/MdeModulePkg/Bus/Scsi/ScsiDiskDxe/ScsiDisk.c
> > > > b/MdeModulePkg/Bus/Scsi/ScsiDiskDxe/ScsiDisk.c
> > > > index 98e84b4ea8..b6e5848e77 100644
> > > > --- a/MdeModulePkg/Bus/Scsi/ScsiDiskDxe/ScsiDisk.c
> > > > +++ b/MdeModulePkg/Bus/Scsi/ScsiDiskDxe/ScsiDisk.c
> > > > @@ -514,7 +514,7 @@ ScsiDiskReset (
> > > >    SCSI_DISK_DEV  *ScsiDiskDevice;
> > > >    EFI_STATUS     Status;
> > > >
> > > > -  OldTpl = gBS->RaiseTPL (TPL_CALLBACK);
> > > > +  OldTpl = gBS->RaiseTPL (TPL_NOTIFY);
> > > >
> > > >    ScsiDiskDevice = SCSI_DISK_DEV_FROM_BLKIO (This);
> > > >
> > > > @@ -581,7 +581,7 @@ ScsiDiskReadBlocks (
> > > >    EFI_TPL             OldTpl;
> > > >
> > > >    MediaChange    = FALSE;
> > > > -  OldTpl         = gBS->RaiseTPL (TPL_CALLBACK);
> > > > +  OldTpl         = gBS->RaiseTPL (TPL_NOTIFY);
> > > >    ScsiDiskDevice = SCSI_DISK_DEV_FROM_BLKIO (This);
> > > >    Media          = ScsiDiskDevice->BlkIo.Media;
> > > >
> > > > @@ -733,7 +733,7 @@ ScsiDiskWriteBlocks (
> > > >    EFI_TPL             OldTpl;
> > > >
> > > >    MediaChange    = FALSE;
> > > > -  OldTpl         = gBS->RaiseTPL (TPL_CALLBACK);
> > > > +  OldTpl         = gBS->RaiseTPL (TPL_NOTIFY);
> > > >    ScsiDiskDevice = SCSI_DISK_DEV_FROM_BLKIO (This);
> > > >    Media          = ScsiDiskDevice->BlkIo.Media;
> > > >
> > > > @@ -898,7 +898,7 @@ ScsiDiskResetEx (
> > > >    SCSI_DISK_DEV  *ScsiDiskDevice;
> > > >    EFI_STATUS     Status;
> > > >
> > > > -  OldTpl = gBS->RaiseTPL (TPL_CALLBACK);
> > > > +  OldTpl = gBS->RaiseTPL (TPL_NOTIFY);
> > > >
> > > >    ScsiDiskDevice = SCSI_DISK_DEV_FROM_BLKIO2 (This);
> > > >
> > > > @@ -975,7 +975,7 @@ ScsiDiskReadBlocksEx (
> > > >    EFI_TPL             OldTpl;
> > > >
> > > >    MediaChange    = FALSE;
> > > > -  OldTpl         = gBS->RaiseTPL (TPL_CALLBACK);
> > > > +  OldTpl         = gBS->RaiseTPL (TPL_NOTIFY);
> > > >    ScsiDiskDevice = SCSI_DISK_DEV_FROM_BLKIO2 (This);
> > > >    Media          = ScsiDiskDevice->BlkIo.Media;
> > > >
> > > > @@ -1154,7 +1154,7 @@ ScsiDiskWriteBlocksEx (
> > > >    EFI_TPL             OldTpl;
> > > >
> > > >    MediaChange    = FALSE;
> > > > -  OldTpl         = gBS->RaiseTPL (TPL_CALLBACK);
> > > > +  OldTpl         = gBS->RaiseTPL (TPL_NOTIFY);
> > > >    ScsiDiskDevice = SCSI_DISK_DEV_FROM_BLKIO2 (This);
> > > >    Media          = ScsiDiskDevice->BlkIo.Media;
> > > >
> > > > @@ -1323,7 +1323,7 @@ ScsiDiskFlushBlocksEx (
> > > >    EFI_TPL             OldTpl;
> > > >
> > > >    MediaChange    = FALSE;
> > > > -  OldTpl         = gBS->RaiseTPL (TPL_CALLBACK);
> > > > +  OldTpl         = gBS->RaiseTPL (TPL_NOTIFY);
> > > >    ScsiDiskDevice = SCSI_DISK_DEV_FROM_BLKIO2 (This);
> > > >    Media          = ScsiDiskDevice->BlkIo.Media;
> > > >
> > > > @@ -1717,7 +1717,7 @@ ScsiDiskEraseBlocks (
> > > >    EFI_TPL             OldTpl;
> > > >
> > > >    MediaChange    = FALSE;
> > > > -  OldTpl         = gBS->RaiseTPL (TPL_CALLBACK);
> > > > +  OldTpl         = gBS->RaiseTPL (TPL_NOTIFY);
> > > >    ScsiDiskDevice = SCSI_DISK_DEV_FROM_ERASEBLK (This);
> > > >
> > > >    if (!IS_DEVICE_FIXED (ScsiDiskDevice)) { @@ -1907,7 +1907,7 @@
> > > > ScsiDiskReceiveData (
> > > >    AlignedBuffer          = NULL;
> > > >    MediaChange            = FALSE;
> > > >    AlignedBufferAllocated = FALSE;
> > > > -  OldTpl                 = gBS->RaiseTPL (TPL_CALLBACK);
> > > > +  OldTpl                 = gBS->RaiseTPL (TPL_NOTIFY);
> > > >    ScsiDiskDevice         = SCSI_DISK_DEV_FROM_STORSEC (This);
> > > >    Media                  = ScsiDiskDevice->BlkIo.Media;
> > > >
> > > > @@ -2122,7 +2122,7 @@ ScsiDiskSendData (
> > > >    AlignedBuffer          = NULL;
> > > >    MediaChange            = FALSE;
> > > >    AlignedBufferAllocated = FALSE;
> > > > -  OldTpl                 = gBS->RaiseTPL (TPL_CALLBACK);
> > > > +  OldTpl                 = gBS->RaiseTPL (TPL_NOTIFY);
> > > >    ScsiDiskDevice         = SCSI_DISK_DEV_FROM_STORSEC (This);
> > > >    Media                  = ScsiDiskDevice->BlkIo.Media;
> > > >
> > > > @@ -2294,7 +2294,7 @@ ScsiDiskDetectMedia (
> > > >
> > > >    Status = gBS->CreateEvent (
> > > >                    EVT_TIMER,
> > > > -                  TPL_CALLBACK,
> > > > +                  TPL_NOTIFY,
> > > >                    NULL,
> > > >                    NULL,
> > > >                    &TimeoutEvt
> > > > --
> > > > 2.17.1
> 
> 
> 
> 
> 


  reply	other threads:[~2022-06-23  2:46 UTC|newest]

Thread overview: 9+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2021-12-14 17:59 [PATCH] MdeModulePkg/ScsiDisk: Change TPL to NOTIFY Jeff Brasen
2021-12-15  2:59 ` Wu, Hao A
2021-12-15  4:48   ` Jeff Brasen
2021-12-15  5:43     ` Wu, Hao A
2021-12-17  4:07       ` Zeng, Star
2022-03-10 16:38         ` [edk2-devel] " Jeff Brasen
2022-06-17 15:37     ` Jeff Brasen
2022-06-23  2:46       ` Wu, Hao A [this message]
2022-06-23  2:53         ` [edk2-devel] " Wu, Hao A

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-list 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=DM6PR11MB4025995904C7C20CC66E4966CAB59@DM6PR11MB4025.namprd11.prod.outlook.com \
    --to=devel@edk2.groups.io \
    /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