linux-ext4.vger.kernel.org archive mirror
 help / color / mirror / Atom feed
From: Alistair Popple <apopple@nvidia.com>
To: <akpm@linux-foundation.org>, <Felix.Kuehling@amd.com>,
	<linux-mm@kvack.org>, <rcampbell@nvidia.com>,
	<linux-ext4@vger.kernel.org>, <linux-xfs@vger.kernel.org>,
	"Sierra Guiza, Alejandro (Alex)" <alex.sierra@amd.com>
Cc: <amd-gfx@lists.freedesktop.org>,
	<dri-devel@lists.freedesktop.org>, <hch@lst.de>, <jgg@nvidia.com>,
	<jglisse@redhat.com>
Subject: Re: [PATCH v2 09/12] lib: test_hmm add module param for zone device type
Date: Tue, 21 Sep 2021 15:14:25 +1000	[thread overview]
Message-ID: <2161903.HsYN06obEU@nvdebian> (raw)
In-Reply-To: <729f833c-e880-96ce-5f49-2d72a93faa21@amd.com>

On Tuesday, 21 September 2021 6:05:30 AM AEST Sierra Guiza, Alejandro (Alex) wrote:
> 
> On 9/20/2021 3:53 AM, Alistair Popple wrote:
> > On Tuesday, 14 September 2021 2:16:01 AM AEST Alex Sierra wrote:
> >> In order to configure device public in test_hmm, two module parameters
> >> should be passed, which correspond to the SP start address of each
> >> device (2) spm_addr_dev0 & spm_addr_dev1. If no parameters are passed,
> >> private device type is configured.
> > It's a pity that testing this seems to require some amount of special setup to
> > test. Is there a way this could be made to work on a more standard setup
> > similar to how DEVICE_PRIVATE is tested?
> Hi Alistair
> We tried to do it as simpler as possible. Unfortunately, there are two main
> requirements to register dev memory as DEVICE_PUBLIC type. This memory must
> NOT be accessed by any memory allocator (SLAB, SLOB, SLUB) plus, it has 
> to be
> CPU coherently accessed.  We also want to avoid aliasing the same PFNs for
> different page types (regular system memory and DEVICE_PUBLIC). So we don't
> want the reserved memory to be part of the kernel's memory map before we 
> call
> memremap_pages. A transparent way of doing it, without any special HW, was
> setting a portion of system memory as SPM (Special purpose memory). And use
> this as our “device fake” memory.

Ok, I think it's great that we can test this without special HW but the boot
time configuration is still a bit annoying. Would it be possible to allocate
memory fitting the above requirements by hot unplugging it with something like
offline_and_remove_memory()?

I also don't see why the DEVICE_PRIVATE and DEVICE_PUBLIC testing should be
mutually exclusive - why can't we test both without reloading the module?

 - Alistair

> Regards,
> Alex Sierra
> 
> >
> >> Signed-off-by: Alex Sierra <alex.sierra@amd.com>
> >> ---
> >> v5:
> >> Remove devmem->pagemap.type = MEMORY_DEVICE_PRIVATE at
> >> dmirror_allocate_chunk that was forcing to configure pagemap.type
> >> to MEMORY_DEVICE_PRIVATE
> >>
> >> v6:
> >> Check for null pointers for resource and memremap references
> >> at dmirror_allocate_chunk
> >>
> >> v7:
> >> Due to patch dropped from these patch series "kernel: resource:
> >> lookup_resource as exported symbol", lookup_resource was not longer a
> >> callable function. This was used in public device configuration, to
> >> get start and end addresses, to create pgmap->range struct. This
> >> information is now taken directly from the spm_addr_devX parameters and
> >> the fixed size DEVMEM_CHUNK_SIZE.
> >> ---
> >>   lib/test_hmm.c      | 66 +++++++++++++++++++++++++++++++--------------
> >>   lib/test_hmm_uapi.h |  1 +
> >>   2 files changed, 47 insertions(+), 20 deletions(-)
> >>
> >> diff --git a/lib/test_hmm.c b/lib/test_hmm.c
> >> index 3cd91ca31dd7..ef27e355738a 100644
> >> --- a/lib/test_hmm.c
> >> +++ b/lib/test_hmm.c
> >> @@ -33,6 +33,16 @@
> >>   #define DEVMEM_CHUNK_SIZE		(256 * 1024 * 1024U)
> >>   #define DEVMEM_CHUNKS_RESERVE		16
> >>   
> >> +static unsigned long spm_addr_dev0;
> >> +module_param(spm_addr_dev0, long, 0644);
> >> +MODULE_PARM_DESC(spm_addr_dev0,
> >> +		"Specify start address for SPM (special purpose memory) used for device 0. By setting this Generic device type will be used. Make sure spm_addr_dev1 is set too");
> >> +
> >> +static unsigned long spm_addr_dev1;
> >> +module_param(spm_addr_dev1, long, 0644);
> >> +MODULE_PARM_DESC(spm_addr_dev1,
> >> +		"Specify start address for SPM (special purpose memory) used for device 1. By setting this Generic device type will be used. Make sure spm_addr_dev0 is set too");
> >> +
> >>   static const struct dev_pagemap_ops dmirror_devmem_ops;
> >>   static const struct mmu_interval_notifier_ops dmirror_min_ops;
> >>   static dev_t dmirror_dev;
> >> @@ -450,11 +460,11 @@ static int dmirror_write(struct dmirror *dmirror, struct hmm_dmirror_cmd *cmd)
> >>   	return ret;
> >>   }
> >>   
> >> -static bool dmirror_allocate_chunk(struct dmirror_device *mdevice,
> >> +static int dmirror_allocate_chunk(struct dmirror_device *mdevice,
> >>   				   struct page **ppage)
> >>   {
> >>   	struct dmirror_chunk *devmem;
> >> -	struct resource *res;
> >> +	struct resource *res = NULL;
> >>   	unsigned long pfn;
> >>   	unsigned long pfn_first;
> >>   	unsigned long pfn_last;
> >> @@ -462,17 +472,29 @@ static bool dmirror_allocate_chunk(struct dmirror_device *mdevice,
> >>   
> >>   	devmem = kzalloc(sizeof(*devmem), GFP_KERNEL);
> >>   	if (!devmem)
> >> -		return false;
> >> +		return -ENOMEM;
> >>   
> >> -	res = request_free_mem_region(&iomem_resource, DEVMEM_CHUNK_SIZE,
> >> -				      "hmm_dmirror");
> >> -	if (IS_ERR(res))
> >> -		goto err_devmem;
> >> +	if (!spm_addr_dev0 && !spm_addr_dev1) {
> >> +		res = request_free_mem_region(&iomem_resource, DEVMEM_CHUNK_SIZE,
> >> +					      "hmm_dmirror");
> >> +		if (IS_ERR_OR_NULL(res))
> >> +			goto err_devmem;
> >> +		devmem->pagemap.range.start = res->start;
> >> +		devmem->pagemap.range.end = res->end;
> >> +		devmem->pagemap.type = MEMORY_DEVICE_PRIVATE;
> >> +		mdevice->zone_device_type = HMM_DMIRROR_MEMORY_DEVICE_PRIVATE;
> >> +	} else if (spm_addr_dev0 && spm_addr_dev1) {
> >> +		devmem->pagemap.range.start = MINOR(mdevice->cdevice.dev) ?
> >> +							spm_addr_dev0 :
> >> +							spm_addr_dev1;
> >> +		devmem->pagemap.range.end = devmem->pagemap.range.start +
> >> +					    DEVMEM_CHUNK_SIZE - 1;
> >> +		devmem->pagemap.type = MEMORY_DEVICE_PUBLIC;
> >> +		mdevice->zone_device_type = HMM_DMIRROR_MEMORY_DEVICE_PUBLIC;
> >> +	} else {
> >> +		pr_err("Both spm_addr_dev parameters should be set\n");
> >> +	}
> >>   
> >> -	mdevice->zone_device_type = HMM_DMIRROR_MEMORY_DEVICE_PRIVATE;
> >> -	devmem->pagemap.type = MEMORY_DEVICE_PRIVATE;
> >> -	devmem->pagemap.range.start = res->start;
> >> -	devmem->pagemap.range.end = res->end;
> >>   	devmem->pagemap.nr_range = 1;
> >>   	devmem->pagemap.ops = &dmirror_devmem_ops;
> >>   	devmem->pagemap.owner = mdevice;
> >> @@ -493,10 +515,14 @@ static bool dmirror_allocate_chunk(struct dmirror_device *mdevice,
> >>   		mdevice->devmem_capacity = new_capacity;
> >>   		mdevice->devmem_chunks = new_chunks;
> >>   	}
> >> -
> >>   	ptr = memremap_pages(&devmem->pagemap, numa_node_id());
> >> -	if (IS_ERR(ptr))
> >> +	if (IS_ERR_OR_NULL(ptr)) {
> >> +		if (ptr)
> >> +			ret = PTR_ERR(ptr);
> >> +		else
> >> +			ret = -EFAULT;
> >>   		goto err_release;
> >> +	}
> >>   
> >>   	devmem->mdevice = mdevice;
> >>   	pfn_first = devmem->pagemap.range.start >> PAGE_SHIFT;
> >> @@ -529,7 +555,8 @@ static bool dmirror_allocate_chunk(struct dmirror_device *mdevice,
> >>   
> >>   err_release:
> >>   	mutex_unlock(&mdevice->devmem_lock);
> >> -	release_mem_region(devmem->pagemap.range.start, range_len(&devmem->pagemap.range));
> >> +	if (res)
> >> +		release_mem_region(devmem->pagemap.range.start, range_len(&devmem->pagemap.range));
> >>   err_devmem:
> >>   	kfree(devmem);
> >>   
> >> @@ -1097,10 +1124,8 @@ static int dmirror_device_init(struct dmirror_device *mdevice, int id)
> >>   	if (ret)
> >>   		return ret;
> >>   
> >> -	/* Build a list of free ZONE_DEVICE private struct pages */
> >> -	dmirror_allocate_chunk(mdevice, NULL);
> >> -
> >> -	return 0;
> >> +	/* Build a list of free ZONE_DEVICE struct pages */
> >> +	return dmirror_allocate_chunk(mdevice, NULL);
> >>   }
> >>   
> >>   static void dmirror_device_remove(struct dmirror_device *mdevice)
> >> @@ -1113,8 +1138,9 @@ static void dmirror_device_remove(struct dmirror_device *mdevice)
> >>   				mdevice->devmem_chunks[i];
> >>   
> >>   			memunmap_pages(&devmem->pagemap);
> >> -			release_mem_region(devmem->pagemap.range.start,
> >> -					   range_len(&devmem->pagemap.range));
> >> +			if (devmem->pagemap.type == MEMORY_DEVICE_PRIVATE)
> >> +				release_mem_region(devmem->pagemap.range.start,
> >> +						   range_len(&devmem->pagemap.range));
> >>   			kfree(devmem);
> >>   		}
> >>   		kfree(mdevice->devmem_chunks);
> >> diff --git a/lib/test_hmm_uapi.h b/lib/test_hmm_uapi.h
> >> index ee88701793d5..00259d994410 100644
> >> --- a/lib/test_hmm_uapi.h
> >> +++ b/lib/test_hmm_uapi.h
> >> @@ -65,6 +65,7 @@ enum {
> >>   enum {
> >>   	/* 0 is reserved to catch uninitialized type fields */
> >>   	HMM_DMIRROR_MEMORY_DEVICE_PRIVATE = 1,
> >> +	HMM_DMIRROR_MEMORY_DEVICE_PUBLIC,
> >>   };
> >>   
> >>   #endif /* _LIB_TEST_HMM_UAPI_H */
> >>
> >
> >
> >
> 





  reply	other threads:[~2021-09-21  5:14 UTC|newest]

Thread overview: 20+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2021-09-13 16:15 [PATCH v2 00/12] MEMORY_DEVICE_PUBLIC for CPU-accessible coherent device memory Alex Sierra
2021-09-13 16:15 ` [PATCH v2 01/12] ext4/xfs: add page refcount helper Alex Sierra
2021-09-13 16:15 ` [PATCH v2 02/12] mm: remove extra ZONE_DEVICE struct page refcount Alex Sierra
2021-09-15  5:32   ` Ralph Campbell
2021-09-13 16:15 ` [PATCH v2 03/12] mm: add zone device public type memory support Alex Sierra
2021-09-13 16:15 ` [PATCH v2 04/12] mm: add device public vma selection for memory migration Alex Sierra
2021-09-13 16:15 ` [PATCH v2 05/12] drm/amdkfd: ref count init for device pages Alex Sierra
2021-09-13 16:15 ` [PATCH v2 06/12] drm/amdkfd: add SPM support for SVM Alex Sierra
2021-09-13 16:15 ` [PATCH v2 07/12] drm/amdkfd: public type as sys mem on migration to ram Alex Sierra
2021-09-13 16:16 ` [PATCH v2 08/12] lib: test_hmm add ioctl to get zone device type Alex Sierra
2021-09-13 16:16 ` [PATCH v2 09/12] lib: test_hmm add module param for " Alex Sierra
2021-09-20  8:53   ` Alistair Popple
2021-09-20 20:05     ` Sierra Guiza, Alejandro (Alex)
2021-09-21  5:14       ` Alistair Popple [this message]
2021-09-23 15:52         ` Sierra Guiza, Alejandro (Alex)
2021-10-01  0:28           ` Alistair Popple
2021-09-13 16:16 ` [PATCH v2 10/12] lib: add support for device public type in test_hmm Alex Sierra
2021-10-01  1:32   ` Alistair Popple
2021-09-13 16:16 ` [PATCH v2 11/12] tools: update hmm-test to support device public type Alex Sierra
2021-09-13 16:16 ` [PATCH v2 12/12] tools: update test_hmm script to support SP config Alex Sierra

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=2161903.HsYN06obEU@nvdebian \
    --to=apopple@nvidia.com \
    --cc=Felix.Kuehling@amd.com \
    --cc=akpm@linux-foundation.org \
    --cc=alex.sierra@amd.com \
    --cc=amd-gfx@lists.freedesktop.org \
    --cc=dri-devel@lists.freedesktop.org \
    --cc=hch@lst.de \
    --cc=jgg@nvidia.com \
    --cc=jglisse@redhat.com \
    --cc=linux-ext4@vger.kernel.org \
    --cc=linux-mm@kvack.org \
    --cc=linux-xfs@vger.kernel.org \
    --cc=rcampbell@nvidia.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).