Message-ID: <8d2b30b76bb94255bc03006a69459b26@samsung.com>
Date: Thu, 11 Apr 2024 23:32:44 +0000
From: Dan Helmick <dan.helmick@...sung.com>
To: Luis Chamberlain <mcgrof@...nel.org>, John Garry
	<john.g.garry@...cle.com>
CC: "axboe@...nel.dk" <axboe@...nel.dk>, "kbusch@...nel.org"
	<kbusch@...nel.org>, "hch@....de" <hch@....de>, "sagi@...mberg.me"
	<sagi@...mberg.me>, "jejb@...ux.ibm.com" <jejb@...ux.ibm.com>,
	"martin.petersen@...cle.com" <martin.petersen@...cle.com>,
	"djwong@...nel.org" <djwong@...nel.org>, "viro@...iv.linux.org.uk"
	<viro@...iv.linux.org.uk>, "brauner@...nel.org" <brauner@...nel.org>,
	"dchinner@...hat.com" <dchinner@...hat.com>, "jack@...e.cz" <jack@...e.cz>,
	"linux-block@...r.kernel.org" <linux-block@...r.kernel.org>,
	"linux-kernel@...r.kernel.org" <linux-kernel@...r.kernel.org>,
	"linux-nvme@...ts.infradead.org" <linux-nvme@...ts.infradead.org>,
	"linux-fsdevel@...r.kernel.org" <linux-fsdevel@...r.kernel.org>,
	"tytso@....edu" <tytso@....edu>, "jbongio@...gle.com" <jbongio@...gle.com>,
	"linux-scsi@...r.kernel.org" <linux-scsi@...r.kernel.org>,
	"ojaswin@...ux.ibm.com" <ojaswin@...ux.ibm.com>, "linux-aio@...ck.org"
	<linux-aio@...ck.org>, "linux-btrfs@...r.kernel.org"
	<linux-btrfs@...r.kernel.org>, "io-uring@...r.kernel.org"
	<io-uring@...r.kernel.org>, "nilay@...ux.ibm.com" <nilay@...ux.ibm.com>,
	"ritesh.list@...il.com" <ritesh.list@...il.com>, "willy@...radead.org"
	<willy@...radead.org>, Alan Adamson <alan.adamson@...cle.com>
Subject: RE: [PATCH v6 10/10] nvme: Atomic write support

On Thu, April 11, 2024 10:23 AM, Luis Chaimberlain wrote:
> On Thu, Apr 11, 2024 at 09:59:57AM +0100, John Garry wrote:
> > On 11/04/2024 01:29, Luis Chamberlain wrote:
> > > On Tue, Mar 26, 2024 at 01:38:13PM +0000, John Garry wrote:
> > > > From: Alan Adamson <alan.adamson@...cle.com>
> > > >
> > > > Add support to set block layer request_queue atomic write limits.
> > > > The limits will be derived from either the namespace or controller
> > > > atomic parameters.
> > > >
> > > > NVMe atomic-related parameters are grouped into "normal" and "power-
> fail"
> > > > (or PF) class of parameter. For atomic write support, only PF
> > > > parameters are of interest. The "normal" parameters are concerned
> > > > with racing reads and writes (which also applies to PF). See NVM
> > > > Command Set Specification Revision 1.0d section 2.1.4 for reference.
> > > >
> > > > Whether to use per namespace or controller atomic parameters is
> > > > decided by NSFEAT bit 1 - see Figure 97: Identify – Identify
> > > > Namespace Data Structure, NVM Command Set.
> > > >
> > > > NVMe namespaces may define an atomic boundary, whereby no atomic
> > > > guarantees are provided for a write which straddles this per-lba
> > > > space boundary. The block layer merging policy is such that no
> > > > merges may occur in which the resultant request would straddle such a
> boundary.
> > > >
> > > > Unlike SCSI, NVMe specifies no granularity or alignment rules,
> > > > apart from atomic boundary rule.
> > >
> > > Larger IU drives a larger alignment *preference*, and it can be
> > > multiples of the LBA format, it's called Namespace Preferred Write
> > > Granularity (NPWG) and the NVMe driver already parses it. So say you
> > > have a 4k LBA format but a 16k NPWG. I suspect this means we'd want
> > > atomics writes to align to 16k but I can let Dan confirm.

Apologies for my delayed reply.  I confirm.  

FYI: I authored the first draft of the OPTPERF section and I tried also at a point to help clarify the atomics section.  After my 1st drafts on these sections, there was a fair amount of translations from English into Standards type language.  So, some drive specifics get removed to enable other medias and so-forth.  

NPWG is a preference.  It does not dictate the atomic behavior though.  So, don't go assuming you can rely on that behavior.  
 

> >
> > If we need to be aligned to NPWG, then the min atomic write unit would
> > also need to be NPWG. Any NPWG relation to atomic writes is not
> > defined in the spec, AFAICS.
> 
> NPWG is just a preference, not a requirement, so it is different than logical
> block size. As far as I can tell we have no block topology information to
> represent it. LBS will help users opt-in to align to the NPWG, and a respective
> NAWUPF will ensure you can also atomically write the respective sector size.
> 
> For atomics, NABSPF is what we want to use.
> 
> The above statement on the commit log just seems a bit misleading then.
> 
> > We simply use the LBA data size as the min atomic unit in this patch.
> 
> I thought NABSPF is used.

Yes, use NABSPF.  But most SSDs don't actually have boundaries.  This is more of a legacy SSD need.  

> 
> > > > Note on NABSPF:
> > > > There seems to be some vagueness in the spec as to whether NABSPF
> > > > applies for NSFEAT bit 1 being unset. Figure 97 does not
> > > > explicitly mention NABSPF and how it is affected by bit 1. However
> > > > Figure 4 does tell to check Figure
> > > > 97 for info about per-namespace parameters, which NABSPF is, so it
> > > > is implied. However currently nvme_update_disk_info() does check
> > > > namespace parameter NABO regardless of this bit.

NABO is a parameter that was carried forward, and it was already in the spec.  I didn't get an ability to impact that one with my changes.  

The story that was relayed to me says this parameter first existed in SATA and SCSI, and NVMe just pulled over an equivalent parameter even though the problem was resolved in the landscape NVMe SSDs ship into.  I was told that NABO was a parameter from Windows 95-ish days.  Something about the BIOS being written in 512B sectors with an ending that didn't align to 4KB.  But all the HDDs were trying to move over to 4KB for efficiencies of their ECCs.  So, there was this NABO parameter added to get the OS portion of the drive to be aligned nicely with the HDD's ECC.  

Anyways: add in the offset as queried from the drive even though it will most likely always be zero.

> > >
> > > Yeah that its quirky.
> > >
> > > Also today we set the physical block size to min(npwg, atomic) and
> > > that means for a today's average 4k IU drive if they get 16k atomic
> > > the physical block size would still be 4k. As the physical block
> > > size in practice can also lift the sector size filesystems used it
> > > would seem odd only a larger npwg could lift it.
> > It seems to me that if you want to provide atomic guarantees for this
> > large "physical block size", then it needs to be based on (N)AWUPF and NPWG.
> 
> For atomicity, I read it as needing to use NABSPF. Aligning to NPWG will just
> help performance.
> 
> The NPWG comes from an internal mapping table constructed and kept on
> DRAM on a drive in units of an IU size [0], and so not aligning to the IU just
> causes having to work with entries in the able rather than just one, and also
> incurs a read-modify-write. Contrary to the logical block size, a write below
> NPWG but respecting the logical block size is allowed, its just not optimal.
> 
> [0]
> https://urldefense.com/v3/__https://protect2.fireeye.com/v1/url?k=eae43295-
> b57ffcf1-eae5b9da-000babda0201-21ccc05e04b9be40&q=1&e=a20d17e2-
> 9e56-47e4-b5e0-
> 05494254a286&u=https*3A*2F*2Fkernelnewbies.org*2FKernelProjects*2Flarg
> e-block-size*23Indirection_Unit_size_increases__;JSUlJSUl!!EwVzqGoTKBqv-
> 0DWAJBm!Wkzd2Bo5MWgYPXDhfJYso2nocO0kAtKHvtYD1NT6p4QIkC846TclDRJ
> pPd683WDGeJSTbRBLKq5Fy-V9mBa8$
> 
>   Luis

Powered by blists - more mailing lists