DPDK patches and discussions
 help / color / mirror / Atom feed
From: Harman Kalra <hkalra@marvell.com>
To: David Marchand <david.marchand@redhat.com>,
	Dmitry Kozlyuk <dmitry.kozliuk@gmail.com>
Cc: Stephen Hemminger <stephen@networkplumber.org>,
	Thomas Monjalon <thomas@monjalon.net>,
	"dev@dpdk.org" <dev@dpdk.org>, Ray Kinsella <mdr@ashroe.eu>
Subject: Re: [dpdk-dev] [EXT] Re: [PATCH v3 2/7] eal/interrupts: implement get set APIs
Date: Thu, 21 Oct 2021 16:05:58 +0000
Message-ID: <BL1PR18MB41971B56AB1E73266CE6DBE5C5BF9@BL1PR18MB4197.namprd18.prod.outlook.com> (raw)
In-Reply-To: <CAJFAV8xN8cWeviqQCHfHj_r5HBLo6+UBL+W_as+KzWTEKPoZpg@mail.gmail.com>

Hi Dmitry, David

Please find responses inline.

> -----Original Message-----
> From: David Marchand <david.marchand@redhat.com>
> Sent: Thursday, October 21, 2021 7:03 PM
> To: Dmitry Kozlyuk <dmitry.kozliuk@gmail.com>; Harman Kalra
> <hkalra@marvell.com>
> Cc: Stephen Hemminger <stephen@networkplumber.org>; Thomas
> Monjalon <thomas@monjalon.net>; dev@dpdk.org; Ray Kinsella
> <mdr@ashroe.eu>
> Subject: Re: [EXT] Re: [dpdk-dev] [PATCH v3 2/7] eal/interrupts: implement
> get set APIs
> 
> On Thu, Oct 21, 2021 at 2:33 PM Dmitry Kozlyuk <dmitry.kozliuk@gmail.com>
> wrote:
> > > Hi All,
> > >
> > > I came across 2 issues introduced with auto detection mechanism.
> > > 1. In case of primary secondary model.  Primary application is
> > > started which makes lots of allocations via
> > > rte_malloc*
> > >
> > >     Secondary side:
> > >     a. Secondary starts, in its "rte_eal_init()" it makes some
> > > allocation via rte_*, and in one of the allocation request for heap expand
> is made as current memseg got exhausted. (malloc_heap_alloc_on_heap_id
> ()->
> > >    alloc_more_mem_on_socket()->try_expand_heap())
> > >    b. A request to primary for heap expand is sent. Please note
> > > secondary holds the spinlock while making the request.
> > > (malloc_heap_alloc_on_heap_id ()->rte_spinlock_lock(&(heap->lock));)
> > >
> > >    Primary side:
> > >    a. Primary receives the request, install a new hugepage and setups up
> the heap (handle_alloc_request())
> > >    b. To inform all the secondaries about the new memseg, primary
> > > sends a sync notice where it sets up an alarm (rte_mp_request_async ()-
> >mp_request_async()).
> > >    c. Inside alarm setup API, we register an interrupt callback.
> > >    d. Inside rte_intr_callback_register(), a new interrupt instance allocation
> is requested for "src->intr_handle"
> > >    e. Since memory management is detected as up, inside
> > > "rte_intr_instance_alloc()", call to "rte_zmalloc" for allocating
> > > memory and further inside "malloc_heap_alloc_on_heap_id()", primary
> will experience a deadlock while taking up the spinlock because this spinlock
> is already hold by secondary.
> > >
> > >
> > > 2. "eal_flags_file_prefix_autotest" is failing because the spawned
> > > process by this tests are expected to cleanup their hugepage traces from
> respective directories (eg /dev/hugepage).
> > > a. Inside eal_cleanup, rte_free()->malloc_heap_free(), where element
> > > to be freed is added to the free list and checked if nearby elements can
> be joined together and form a big free chunk (malloc_elem_free()).
> > > b. If this free chunk is big enough than the hugepage size,
> > > respective hugepage can be uninstalled after making sure no
> > > allocation from this hugepage exists.
> > > (malloc_heap_free()->malloc_heap_free_pages()-
> >eal_memalloc_free_seg
> > > ())
> > >
> > > But because of interrupt allocations made for pci intr handles (used
> > > for VFIO) and other driver specific interrupt handles are not cleaned up in
> "rte_eal_cleanup()", these hugepage files are not removed and test fails.
> >
> > Sad to hear. But it's a great and thorough analysis.

Sad but a good learning, atleast we identified areas to be worked upon.

> >
> > > There could be more such issues, I think we should firstly fix the DPDK.
> > > 1. Memory management should be made independent and should be the
> > > first thing to come up in rte_eal_init()
> >
> > As I have explained, buses must be able to report IOVA requirement at
> > this point (`get_iommu_class()` bus method).
> > Either `scan()` must complete before that or `get_iommu_class()` must
> > be able to work before `scan()` is called.
> >
> > > 2. rte_eal_cleanup() should be exactly opposite to rte_eal_init(),
> > > just like bus_probe, we should have bus_remove to clean up all the
> memory allocations.
> >
> > Yes. For most buses it will be just "unplug each device".
> > In fact, EAL could do it with `unplug()`, but it is not mandatory.

I implemented a rough bus_remove which was similar to unplug, faced
some issue. Not sure but some drivers might not be supporting hotplug, for
them unplug might be a challenge.


> >
> > >
> > > Regarding this IRQ series, I would like to fall back to our original
> > > design i.e. rte_intr_instance_alloc() should take an argument whether its
> memory should be allocated using glibc malloc or rte_malloc*.
> >
> > Seems there's no other option to make it on time.
> 
> - Sorry, my memory is too short, did we describe where we need to share
> rte_intr_handle objects?

Intr handle objects are shared in very few drivers.

> 
> I spent some time looking at uses of rte_intr_handle objects.
> 
> In many cases intr_handle objects are referenced in malloc() objects.
> The cases where rte_intr_handle are shared is in per device private bits in
> drivers.
> 

Yes, in V2 design I allocated memory using glibc malloc for such instances by
passing respective flag.

> A intr_handle often contains fds.
> For them to be used in mp setups, there needs to be a big machinery with
> SCM_RIGHTS but I see only 3 drivers which actually reference this.
> So if intr_handle fds are accessed by multiple processes, their content
> probably makes no sense wrt fds.

Those drivers will allocate using SHARED flag.

> 
> 
> From these two hints, I think we are going backwards, and the main usecase
> is that those rte_intr_instance objects are not used in mp.
> I even think they are never accessed from other processes.
> But I am not sure.
> 
> 
> - Seeing how time it short for rc1, I am ok with
> rte_intr_instance_alloc() taking a flag argument.
> And we can still go back on this API later.


Sure, I will revert back to original design and send V5 by tomorrow.

> 
> Can we agree on the flag name?
> rte_malloc() interest is that it makes objects shared for mp, so how about
> RTE_INTR_INSTANCE_F_SHARED ?

Yeah, it sounds good:
RTE_INTR_INSTANCE_F_SHARED  - rte_malloc
RTE_INTR_INSTANCE_F_PRIVATE - malloc


Thanks David, Dmitry, Thomas, Stephan for reviewing the series thoroughly and providing
inputs to improvise it.


Thanks
Harman

> 
> 
> --
> David Marchand


  reply	other threads:[~2021-10-21 16:06 UTC|newest]

Thread overview: 152+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2021-08-26 14:57 [dpdk-dev] [RFC 0/7] make rte_intr_handle internal Harman Kalra
2021-08-26 14:57 ` [dpdk-dev] [RFC 1/7] eal: interrupt handle API prototypes Harman Kalra
2021-08-31 15:52   ` Kinsella, Ray
2021-08-26 14:57 ` [dpdk-dev] [RFC 2/7] eal/interrupts: implement get set APIs Harman Kalra
2021-08-31 15:53   ` Kinsella, Ray
2021-08-26 14:57 ` [dpdk-dev] [RFC 3/7] eal/interrupts: avoid direct access to interrupt handle Harman Kalra
2021-08-26 14:57 ` [dpdk-dev] [RFC 4/7] test/interrupt: apply get set interrupt handle APIs Harman Kalra
2021-08-26 14:57 ` [dpdk-dev] [RFC 5/7] drivers: remove direct access to interrupt handle fields Harman Kalra
2021-08-26 14:57 ` [dpdk-dev] [RFC 6/7] eal/interrupts: make interrupt handle structure opaque Harman Kalra
2021-08-26 14:57 ` [dpdk-dev] [RFC 7/7] eal/alarm: introduce alarm fini routine Harman Kalra
2021-09-03 12:40 ` [dpdk-dev] [PATCH v1 0/7] make rte_intr_handle internal Harman Kalra
2021-09-03 12:40   ` [dpdk-dev] [PATCH v1 1/7] eal: interrupt handle API prototypes Harman Kalra
2021-09-03 12:40   ` [dpdk-dev] [PATCH v1 2/7] eal/interrupts: implement get set APIs Harman Kalra
2021-09-28 15:46     ` David Marchand
2021-10-04  8:51       ` [dpdk-dev] [EXT] " Harman Kalra
2021-10-04  9:57         ` David Marchand
2021-10-12 15:22           ` Thomas Monjalon
2021-10-13 17:54             ` Harman Kalra
2021-10-13 17:57               ` Harman Kalra
2021-10-13 18:52                 ` Thomas Monjalon
2021-10-14  8:22                   ` Thomas Monjalon
2021-10-14  9:31                     ` Harman Kalra
2021-10-14  9:37                       ` David Marchand
2021-10-14  9:41                       ` Thomas Monjalon
2021-10-14 10:31                         ` Harman Kalra
2021-10-14 10:35                           ` Thomas Monjalon
2021-10-14 10:44                             ` Harman Kalra
2021-10-14 12:04                               ` Thomas Monjalon
2021-10-14 10:25                       ` Dmitry Kozlyuk
2021-10-03 18:05     ` [dpdk-dev] " Dmitry Kozlyuk
2021-10-04 10:37       ` [dpdk-dev] [EXT] " Harman Kalra
2021-10-04 11:18         ` Dmitry Kozlyuk
2021-10-04 14:03           ` Harman Kalra
2021-09-03 12:40   ` [dpdk-dev] [PATCH v1 3/7] eal/interrupts: avoid direct access to interrupt handle Harman Kalra
2021-09-03 12:40   ` [dpdk-dev] [PATCH v1 4/7] test/interrupt: apply get set interrupt handle APIs Harman Kalra
2021-09-03 12:41   ` [dpdk-dev] [PATCH v1 5/7] drivers: remove direct access to interrupt handle fields Harman Kalra
2021-09-03 12:41   ` [dpdk-dev] [PATCH v1 6/7] eal/interrupts: make interrupt handle structure opaque Harman Kalra
2021-10-03 18:16     ` Dmitry Kozlyuk
2021-10-04 14:09       ` [dpdk-dev] [EXT] " Harman Kalra
2021-09-03 12:41   ` [dpdk-dev] [PATCH v1 7/7] eal/alarm: introduce alarm fini routine Harman Kalra
2021-09-15 14:13   ` [dpdk-dev] [PATCH v1 0/7] make rte_intr_handle internal Harman Kalra
2021-09-23  8:20   ` David Marchand
2021-10-05 12:14 ` [dpdk-dev] [PATCH v2 0/6] " Harman Kalra
2021-10-05 12:14   ` [dpdk-dev] [PATCH v2 1/6] eal/interrupts: implement get set APIs Harman Kalra
2021-10-14  0:58     ` Dmitry Kozlyuk
2021-10-14 17:15       ` [dpdk-dev] [EXT] " Harman Kalra
2021-10-14 17:53         ` Dmitry Kozlyuk
2021-10-15  7:53           ` Thomas Monjalon
2021-10-14  7:31     ` [dpdk-dev] " David Marchand
2021-10-05 12:14   ` [dpdk-dev] [PATCH v2 2/6] eal/interrupts: avoid direct access to interrupt handle Harman Kalra
2021-10-14  0:59     ` Dmitry Kozlyuk
2021-10-14 17:31       ` [dpdk-dev] [EXT] " Harman Kalra
2021-10-14 17:53         ` Dmitry Kozlyuk
2021-10-05 12:14   ` [dpdk-dev] [PATCH v2 3/6] test/interrupt: apply get set interrupt handle APIs Harman Kalra
2021-10-05 12:15   ` [dpdk-dev] [PATCH v2 4/6] drivers: remove direct access to interrupt handle Harman Kalra
2021-10-05 12:15   ` [dpdk-dev] [PATCH v2 5/6] eal/interrupts: make interrupt handle structure opaque Harman Kalra
2021-10-05 12:15   ` [dpdk-dev] [PATCH v2 6/6] eal/alarm: introduce alarm fini routine Harman Kalra
2021-10-05 16:07 ` [dpdk-dev] [RFC 0/7] make rte_intr_handle internal Stephen Hemminger
2021-10-07 10:57   ` [dpdk-dev] [EXT] " Harman Kalra
2021-10-18 19:37 ` [dpdk-dev] [PATCH v3 " Harman Kalra
2021-10-18 19:37   ` [dpdk-dev] [PATCH v3 1/7] malloc: introduce malloc is ready API Harman Kalra
2021-10-19 15:53     ` Thomas Monjalon
2021-10-18 19:37   ` [dpdk-dev] [PATCH v3 2/7] eal/interrupts: implement get set APIs Harman Kalra
2021-10-18 22:07     ` Dmitry Kozlyuk
2021-10-19  8:50       ` [dpdk-dev] [EXT] " Harman Kalra
2021-10-19 18:44         ` Harman Kalra
2021-10-18 22:56     ` [dpdk-dev] " Stephen Hemminger
2021-10-19  8:32       ` [dpdk-dev] [EXT] " Harman Kalra
2021-10-19 15:58         ` Thomas Monjalon
2021-10-20 15:30         ` Dmitry Kozlyuk
2021-10-21  9:16           ` Harman Kalra
2021-10-21 12:33             ` Dmitry Kozlyuk
2021-10-21 13:32               ` David Marchand
2021-10-21 16:05                 ` Harman Kalra [this message]
2021-10-18 19:37   ` [dpdk-dev] [PATCH v3 3/7] eal/interrupts: avoid direct access to interrupt handle Harman Kalra
2021-10-18 19:37   ` [dpdk-dev] [PATCH v3 4/7] test/interrupt: apply get set interrupt handle APIs Harman Kalra
2021-10-18 19:37   ` [dpdk-dev] [PATCH v3 5/7] drivers: remove direct access to interrupt handle Harman Kalra
2021-10-18 19:37   ` [dpdk-dev] [PATCH v3 6/7] eal/interrupts: make interrupt handle structure opaque Harman Kalra
2021-10-18 19:37   ` [dpdk-dev] [PATCH v3 7/7] eal/alarm: introduce alarm fini routine Harman Kalra
2021-10-19 18:35 ` [dpdk-dev] [PATCH v4 0/7] make rte_intr_handle internal Harman Kalra
2021-10-19 18:35   ` [dpdk-dev] [PATCH v4 1/7] malloc: introduce malloc is ready API Harman Kalra
2021-10-19 22:01     ` Dmitry Kozlyuk
2021-10-19 22:04       ` Dmitry Kozlyuk
2021-10-20  9:01         ` [dpdk-dev] [EXT] " Harman Kalra
2021-10-19 18:35   ` [dpdk-dev] [PATCH v4 2/7] eal/interrupts: implement get set APIs Harman Kalra
2021-10-20  6:14     ` David Marchand
2021-10-20 14:29       ` Dmitry Kozlyuk
2021-10-20 16:15     ` Dmitry Kozlyuk
2021-10-19 18:35   ` [dpdk-dev] [PATCH v4 3/7] eal/interrupts: avoid direct access to interrupt handle Harman Kalra
2021-10-19 21:27     ` Dmitry Kozlyuk
2021-10-20  9:25       ` [dpdk-dev] [EXT] " Harman Kalra
2021-10-20  9:52         ` Dmitry Kozlyuk
2021-10-19 18:35   ` [dpdk-dev] [PATCH v4 4/7] test/interrupt: apply get set interrupt handle APIs Harman Kalra
2021-10-19 18:35   ` [dpdk-dev] [PATCH v4 5/7] drivers: remove direct access to interrupt handle Harman Kalra
2021-10-20  1:57     ` Hyong Youb Kim (hyonkim)
2021-10-19 18:35   ` [dpdk-dev] [PATCH v4 6/7] eal/interrupts: make interrupt handle structure opaque Harman Kalra
2021-10-19 18:35   ` [dpdk-dev] [PATCH v4 7/7] eal/alarm: introduce alarm fini routine Harman Kalra
2021-10-19 21:39     ` Dmitry Kozlyuk
2021-10-22 20:49 ` [dpdk-dev] [PATCH v5 0/6] make rte_intr_handle internal Harman Kalra
2021-10-22 20:49   ` [dpdk-dev] [PATCH v5 1/6] eal/interrupts: implement get set APIs Harman Kalra
2021-10-22 23:33     ` Dmitry Kozlyuk
2021-10-22 20:49   ` [dpdk-dev] [PATCH v5 2/6] eal/interrupts: avoid direct access to interrupt handle Harman Kalra
2021-10-22 23:33     ` Dmitry Kozlyuk
2021-10-22 20:49   ` [dpdk-dev] [PATCH v5 3/6] test/interrupt: apply get set interrupt handle APIs Harman Kalra
2021-10-22 20:49   ` [dpdk-dev] [PATCH v5 4/6] drivers: remove direct access to interrupt handle Harman Kalra
2021-10-22 20:49   ` [dpdk-dev] [PATCH v5 5/6] eal/interrupts: make interrupt handle structure opaque Harman Kalra
2021-10-22 23:33     ` Dmitry Kozlyuk
2021-10-22 20:49   ` [dpdk-dev] [PATCH v5 6/6] eal/alarm: introduce alarm fini routine Harman Kalra
2021-10-22 23:33     ` Dmitry Kozlyuk
2021-10-22 23:37       ` Dmitry Kozlyuk
2021-10-24 20:04   ` [dpdk-dev] [PATCH v6 0/9] make rte_intr_handle internal David Marchand
2021-10-24 20:04     ` [dpdk-dev] [PATCH v6 1/9] interrupts: add allocator and accessors David Marchand
2021-10-24 20:04     ` [dpdk-dev] [PATCH v6 2/9] interrupts: remove direct access to interrupt handle David Marchand
2021-10-25  6:57       ` David Marchand
2021-10-24 20:04     ` [dpdk-dev] [PATCH v6 3/9] test/interrupts: " David Marchand
2021-10-24 20:04     ` [dpdk-dev] [PATCH v6 4/9] alarm: " David Marchand
2021-10-25 10:49       ` Dmitry Kozlyuk
2021-10-25 11:09         ` David Marchand
2021-10-24 20:04     ` [dpdk-dev] [PATCH v6 5/9] lib: " David Marchand
2021-10-24 20:04     ` [dpdk-dev] [PATCH v6 6/9] drivers: " David Marchand
2021-10-24 20:04     ` [dpdk-dev] [PATCH v6 7/9] interrupts: make interrupt handle structure opaque David Marchand
2021-10-24 20:04     ` [dpdk-dev] [PATCH v6 8/9] interrupts: rename device specific file descriptor David Marchand
2021-10-24 20:04     ` [dpdk-dev] [PATCH v6 9/9] interrupts: extend event list David Marchand
2021-10-25 10:49       ` Dmitry Kozlyuk
2021-10-25 11:11         ` David Marchand
2021-10-25 13:04   ` [dpdk-dev] [PATCH v5 0/6] make rte_intr_handle internal Raslan Darawsheh
2021-10-25 13:09     ` David Marchand
2021-10-25 13:34   ` [dpdk-dev] [PATCH v7 0/9] " David Marchand
2021-10-25 13:34     ` [dpdk-dev] [PATCH v7 1/9] interrupts: add allocator and accessors David Marchand
2021-10-25 13:34     ` [dpdk-dev] [PATCH v7 2/9] interrupts: remove direct access to interrupt handle David Marchand
2021-10-25 13:34     ` [dpdk-dev] [PATCH v7 3/9] test/interrupts: " David Marchand
2021-10-25 13:34     ` [dpdk-dev] [PATCH v7 4/9] alarm: " David Marchand
2021-10-25 13:34     ` [dpdk-dev] [PATCH v7 5/9] lib: " David Marchand
2021-10-28  6:14       ` Jiang, YuX
2021-10-25 13:34     ` [dpdk-dev] [PATCH v7 6/9] drivers: " David Marchand
2021-10-25 13:34     ` [dpdk-dev] [PATCH v7 7/9] interrupts: make interrupt handle structure opaque David Marchand
2021-10-25 13:34     ` [dpdk-dev] [PATCH v7 8/9] interrupts: rename device specific file descriptor David Marchand
2021-10-25 13:34     ` [dpdk-dev] [PATCH v7 9/9] interrupts: extend event list David Marchand
2021-10-25 14:27   ` [dpdk-dev] [PATCH v8 0/9] make rte_intr_handle internal David Marchand
2021-10-25 14:27     ` [dpdk-dev] [PATCH v8 1/9] interrupts: add allocator and accessors David Marchand
2021-10-25 14:27     ` [dpdk-dev] [PATCH v8 2/9] interrupts: remove direct access to interrupt handle David Marchand
2021-10-25 14:27     ` [dpdk-dev] [PATCH v8 3/9] test/interrupts: " David Marchand
2021-10-25 14:27     ` [dpdk-dev] [PATCH v8 4/9] alarm: " David Marchand
2021-10-25 14:27     ` [dpdk-dev] [PATCH v8 5/9] lib: " David Marchand
2021-10-25 14:27     ` [dpdk-dev] [PATCH v8 6/9] drivers: " David Marchand
2021-10-25 14:27     ` [dpdk-dev] [PATCH v8 7/9] interrupts: make interrupt handle structure opaque David Marchand
2021-10-25 14:27     ` [dpdk-dev] [PATCH v8 8/9] interrupts: rename device specific file descriptor David Marchand
2021-10-25 14:27     ` [dpdk-dev] [PATCH v8 9/9] interrupts: extend event list David Marchand
2021-10-28 15:58       ` Ji, Kai
2021-10-28 17:16         ` David Marchand
2021-10-25 14:32     ` [dpdk-dev] [PATCH v8 0/9] make rte_intr_handle internal Raslan Darawsheh
2021-10-25 19:24     ` David Marchand

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=BL1PR18MB41971B56AB1E73266CE6DBE5C5BF9@BL1PR18MB4197.namprd18.prod.outlook.com \
    --to=hkalra@marvell.com \
    --cc=david.marchand@redhat.com \
    --cc=dev@dpdk.org \
    --cc=dmitry.kozliuk@gmail.com \
    --cc=mdr@ashroe.eu \
    --cc=stephen@networkplumber.org \
    --cc=thomas@monjalon.net \
    /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

DPDK patches and discussions

This inbox may be cloned and mirrored by anyone:

	git clone --mirror http://inbox.dpdk.org/dev/0 dev/git/0.git

	# If you have public-inbox 1.1+ installed, you may
	# initialize and index your mirror using the following commands:
	public-inbox-init -V2 dev dev/ http://inbox.dpdk.org/dev \
		dev@dpdk.org
	public-inbox-index dev

Example config snippet for mirrors.
Newsgroup available over NNTP:
	nntp://inbox.dpdk.org/inbox.dpdk.dev


AGPL code for this site: git clone https://public-inbox.org/public-inbox.git