From: Bjorn Helgaas <bhelgaas@google.com>
To: Yinghai Lu <yinghai@kernel.org>
Cc: David Miller <davem@davemloft.net>,
Benjamin Herrenschmidt <benh@kernel.crashing.org>,
Wei Yang <weiyang@linux.vnet.ibm.com>, TJ <linux@iam.tj>,
Yijing Wang <wangyijing@huawei.com>,
Andrew Morton <akpm@linux-foundation.org>,
linux-pci@vger.kernel.org, linux-kernel@vger.kernel.org
Subject: Re: [PATCH v3 33/51] resources: Make allocate_resource return just fit resource
Date: Mon, 17 Aug 2015 23:21:35 -0500 [thread overview]
Message-ID: <20150818042135.GZ26431@google.com> (raw)
In-Reply-To: <1438039809-24957-34-git-send-email-yinghai@kernel.org>
On Mon, Jul 27, 2015 at 04:29:51PM -0700, Yinghai Lu wrote:
> Find all suitable empty slots and pick one just fit, so we could save
> the big slot for needed ones later when we have several pcie switches
> and some bridges get assigned bios and we need to assign others in kernel.
By "just fit," do you mean "best fit"? "Best fit" is a well-known term, so
if that's what you mean, let's use it.
I couldn't quite parse the PCIe switch stuff here. How is that relevant to
this change?
> Signed-off-by: Yinghai Lu <yinghai@kernel.org>
> ---
> kernel/resource.c | 81 ++++++++++++++++++++++++++++++++++++++++++++++---------
> 1 file changed, 68 insertions(+), 13 deletions(-)
>
> diff --git a/kernel/resource.c b/kernel/resource.c
> index 67b58a5..697b8ca 100644
> --- a/kernel/resource.c
> +++ b/kernel/resource.c
> @@ -48,6 +48,7 @@ struct resource_constraint {
> resource_size_t (*alignf)(void *, const struct resource *,
> resource_size_t, resource_size_t);
> void *alignf_data;
> + bool fit;
> };
>
> static DEFINE_RWLOCK(resource_lock);
> @@ -554,12 +555,15 @@ static void resource_clip(struct resource *res, resource_size_t min,
> * alignment constraints
> */
> static int __find_resource(struct resource *root, struct resource *old,
> - struct resource *new,
> + struct resource *new, struct resource *avail,
> resource_size_t size,
> struct resource_constraint *constraint)
> {
> struct resource *this = root->child;
> - struct resource tmp = *new, avail, alloc;
> + struct resource tmp = *new, availx, alloc;
> +
> + if (!avail || avail == new)
> + avail = &availx;
>
> tmp.start = root->start;
> /*
> @@ -583,15 +587,16 @@ static int __find_resource(struct resource *root, struct resource *old,
> arch_remove_reservations(&tmp);
>
> /* Check for overflow after ALIGN() */
> - avail.start = ALIGN(tmp.start, constraint->align);
> - avail.end = tmp.end;
> - avail.flags = new->flags & ~IORESOURCE_UNSET;
> - if (avail.start >= tmp.start) {
> - alloc.flags = avail.flags;
> - alloc.start = constraint->alignf(constraint->alignf_data, &avail,
> + avail->start = ALIGN(tmp.start, constraint->align);
> + avail->end = tmp.end;
> + avail->flags = new->flags & ~IORESOURCE_UNSET;
> + if (avail->start >= tmp.start) {
> + alloc.flags = avail->flags;
> + alloc.start = constraint->alignf(
> + constraint->alignf_data, avail,
> size, constraint->align);
> alloc.end = alloc.start + size - 1;
> - if (resource_contains(&avail, &alloc)) {
> + if (resource_contains(avail, &alloc)) {
> new->start = alloc.start;
> new->end = alloc.end;
> return 0;
> @@ -608,6 +613,11 @@ next: if (!this || this->end == root->end)
> return -EBUSY;
> }
>
> +struct good_resource {
> + struct list_head list;
> + struct resource avail;
> + struct resource new;
> +};
> /*
> * Find empty slot in the resource tree given range and alignment.
> */
> @@ -615,7 +625,49 @@ static int find_resource(struct resource *root, struct resource *new,
> resource_size_t size,
> struct resource_constraint *constraint)
> {
> - return __find_resource(root, NULL, new, size, constraint);
> + int ret = -1;
> + LIST_HEAD(head);
> + struct good_resource *good, *tmp;
> + resource_size_t avail_size = (resource_size_t)-1ULL;
> +
> + if (!constraint->fit)
> + return __find_resource(root, NULL, new, NULL, size,
> + constraint);
> +
> + /* find all suitable ones and add to the list */
> + for (;;) {
> + good = kzalloc(sizeof(*good), GFP_KERNEL);
> + if (!good)
> + break;
> +
> + good->new.start = new->start;
> + good->new.end = new->end;
> + good->new.flags = new->flags;
> + ret = __find_resource(root, NULL, &good->new, &good->avail,
> + size, constraint);
> + if (ret || __request_resource(root, &good->avail)) {
> + ret = -EBUSY;
> + kfree(good);
> + break;
> + }
> +
> + list_add(&good->list, &head);
> + }
Allocating memory and building a list in a function that allocates space
seems like a little bit of a hack. I think we're holding resource_lock
anyway; can't we just find a candidate, reserve it, look for another one,
reserve it, release the larger one, and repeat?
> + /* pick up the smallest one and delete the list */
> + list_for_each_entry_safe(good, tmp, &head, list) {
> + if (resource_size(&good->avail) < avail_size) {
> + avail_size = resource_size(&good->avail);
> + new->start = good->new.start;
> + new->end = good->new.end;
> + ret = 0;
> + }
> + list_del(&good->list);
> + __release_resource(&good->avail);
> + kfree(good);
> + }
> +
> + return ret;
> }
>
> /**
> @@ -636,7 +688,8 @@ static int __reallocate_resource(struct resource *root, struct resource *old,
> struct resource new = *old;
> struct resource *conflict;
>
> - if ((err = __find_resource(root, old, &new, newsize, constraint)))
> + err = __find_resource(root, old, &new, NULL, newsize, constraint);
> + if (err)
> goto out;
>
> if (resource_contains(&new, old)) {
> @@ -675,6 +728,7 @@ out:
> * @align: alignment requested, in bytes
> * @alignf: alignment function, optional, called if not NULL
> * @alignf_data: arbitrary data to pass to the @alignf function
> + * @fit: only allocate fit range.
> *
> * Caller need to hold resource_lock if needed.
> */
> @@ -685,7 +739,7 @@ static int __allocate_resource(struct resource *root, struct resource *new,
> const struct resource *,
> resource_size_t,
> resource_size_t),
> - void *alignf_data)
> + void *alignf_data, bool fit)
> {
> int err;
> struct resource_constraint constraint;
> @@ -698,6 +752,7 @@ static int __allocate_resource(struct resource *root, struct resource *new,
> constraint.align = align;
> constraint.alignf = alignf;
> constraint.alignf_data = alignf_data;
> + constraint.fit = fit;
>
> if (new->parent) {
> /* resource is already allocated, try reallocating with
> @@ -738,7 +793,7 @@ int allocate_resource(struct resource *root, struct resource *new,
>
> write_lock(&resource_lock);
> ret = __allocate_resource(root, new, size, min, max, align,
> - alignf, alignf_data);
> + alignf, alignf_data, true);
> write_unlock(&resource_lock);
>
> return ret;
> --
> 1.8.4.5
>
next prev parent reply other threads:[~2015-08-18 4:21 UTC|newest]
Thread overview: 81+ messages / expand[flat|nested] mbox.gz Atom feed top
2015-07-27 23:29 [PATCH v3 00/51] PCI: Resource allocation cleanup for v4.3 Yinghai Lu
2015-07-27 23:29 ` [PATCH v3 01/51] PCI: Cleanup res_to_dev_res() printout for addon resources Yinghai Lu
2015-08-17 22:50 ` Bjorn Helgaas
2015-08-18 21:19 ` Yinghai Lu
2015-07-27 23:29 ` [PATCH v3 02/51] PCI: Reuse res_to_dev_res in reassign_resources_sorted Yinghai Lu
2015-07-27 23:29 ` [PATCH v3 03/51] PCI: Use correct align for optional only resources during sorting Yinghai Lu
2015-08-17 23:00 ` Bjorn Helgaas
2015-08-18 19:01 ` Yinghai Lu
2015-07-27 23:29 ` [PATCH v3 04/51] PCI: Optimize bus align/size calculation during sizing Yinghai Lu
2015-08-17 23:49 ` Bjorn Helgaas
2015-08-18 20:29 ` Yinghai Lu
2015-07-27 23:29 ` [PATCH v3 05/51] PCI: Optimize bus align/size calculation for optional " Yinghai Lu
2015-07-27 23:29 ` [PATCH v3 06/51] PCI: Don't add too much optional size for hotplug bridge mmio Yinghai Lu
2015-07-27 23:29 ` [PATCH v3 07/51] PCI: Reorder resources list for must/optional resources Yinghai Lu
2015-08-17 23:52 ` Bjorn Helgaas
2015-08-18 20:58 ` Yinghai Lu
2015-07-27 23:29 ` [PATCH v3 08/51] PCI: Remove duplicated code for resource sorting Yinghai Lu
2015-07-27 23:29 ` [PATCH v3 09/51] PCI: Rename pdev_sort_resources to pdev_check_resources Yinghai Lu
2015-08-17 23:53 ` Bjorn Helgaas
2015-08-18 21:36 ` Yinghai Lu
2015-08-18 21:45 ` Yinghai Lu
2015-07-27 23:29 ` [PATCH v3 10/51] PCI: Treat ROM resource as optional during realloc Yinghai Lu
2015-07-27 23:29 ` [PATCH v3 11/51] PCI: Add debug printout during releasing partial assigned resources Yinghai Lu
2015-07-27 23:29 ` [PATCH v3 12/51] PCI: Simplify res reference using in __assign_resourcs_sorted Yinghai Lu
2015-07-27 23:29 ` [PATCH v3 13/51] PCI: Separate realloc list checking after allocation Yinghai Lu
2015-08-17 23:54 ` Bjorn Helgaas
2015-08-18 21:58 ` Yinghai Lu
2015-07-27 23:29 ` [PATCH v3 14/51] PCI: Add __add_to_list() Yinghai Lu
2015-07-27 23:29 ` [PATCH v3 15/51] PCI: Cache window alignment value Yinghai Lu
2015-07-27 23:29 ` [PATCH v3 16/51] PCI: Check if resource is allocated before pci_assign Yinghai Lu
2015-07-27 23:29 ` [PATCH v3 17/51] PCI: Separate out save_resources/restore_resource Yinghai Lu
2015-07-27 23:29 ` [PATCH v3 18/51] PCI: Move comment to pci_need_to_release() Yinghai Lu
2015-07-27 23:29 ` [PATCH v3 19/51] PCI: Separate must+optional assigning to another function Yinghai Lu
2015-07-27 23:29 ` [PATCH v3 20/51] PCI: Skip must+optional if there is no optional addon Yinghai Lu
2015-08-17 23:56 ` Bjorn Helgaas
2015-08-18 22:39 ` Yinghai Lu
2015-07-27 23:29 ` [PATCH v3 21/51] PCI: Move saved required resource list out of must+optional assigning Yinghai Lu
2015-07-27 23:29 ` [PATCH v3 22/51] PCI: Add alt_size allocation support Yinghai Lu
2015-08-18 0:03 ` Bjorn Helgaas
2015-08-19 5:28 ` Yinghai Lu
2015-07-27 23:29 ` [PATCH v3 23/51] PCI: Add support for more than two alt_size under same bridge Yinghai Lu
2015-07-27 23:29 ` [PATCH v3 24/51] PCI: Better support for two alt_size Yinghai Lu
2015-07-27 23:29 ` [PATCH v3 25/51] PCI: Fix size calculation with old_size on rescan path Yinghai Lu
2015-08-18 4:09 ` Bjorn Helgaas
2015-08-19 6:25 ` Yinghai Lu
2015-07-27 23:29 ` [PATCH v3 26/51] PCI: Don't add too much optional size for hotplug bridge io Yinghai Lu
2015-07-27 23:29 ` [PATCH v3 27/51] PCI: Move ISA ioport align out of calculate_iosize Yinghai Lu
2015-08-18 4:11 ` Bjorn Helgaas
2015-08-19 6:32 ` Yinghai Lu
2015-07-27 23:29 ` [PATCH v3 28/51] PCI: Unifiy calculate_size for io port and mmio Yinghai Lu
2015-08-18 4:13 ` Bjorn Helgaas
2015-08-19 6:37 ` Yinghai Lu
2015-07-27 23:29 ` [PATCH v3 29/51] PCI: Allow optional only io resource must size to be 0 Yinghai Lu
2015-07-27 23:29 ` [PATCH v3 30/51] PCI: Unify skip_ioresource_align() Yinghai Lu
2015-07-27 23:29 ` [PATCH v3 31/51] PCI: Kill macro checking for bus io port sizing Yinghai Lu
2015-07-27 23:29 ` [PATCH v3 32/51] resources: Split out __allocate_resource() Yinghai Lu
2015-08-18 4:14 ` Bjorn Helgaas
2015-08-19 6:58 ` Yinghai Lu
2015-07-27 23:29 ` [PATCH v3 33/51] resources: Make allocate_resource return just fit resource Yinghai Lu
2015-08-18 4:21 ` Bjorn Helgaas [this message]
2015-08-19 7:22 ` Yinghai Lu
2015-07-27 23:29 ` [PATCH v3 34/51] PCI: Check pref compatible bit for mem64 resource of pcie device Yinghai Lu
2015-07-27 23:29 ` [PATCH v3 35/51] PCI: Only treat non-pef mmio64 as pref if all bridges has MEM_64 Yinghai Lu
2015-07-27 23:29 ` [PATCH v3 36/51] PCI: Add has_mem64 for host_bridge Yinghai Lu
2015-07-27 23:29 ` [PATCH v3 37/51] PCI: Only treat non-pef mmio64 as pref if host-bridge has_mem64 Yinghai Lu
2015-07-27 23:29 ` [PATCH v3 38/51] PCI: Restore pref mmio allocation logic for hostbridge without mmio64 Yinghai Lu
2015-07-27 23:29 ` [PATCH v3 39/51] sparc/PCI: Add mem64 resource parsing for root bus Yinghai Lu
2015-07-27 23:29 ` [PATCH v3 40/51] sparc/PCI: Add IORESOURCE_MEM_64 for 64-bit resource in of parsing Yinghai Lu
2015-07-27 23:29 ` [PATCH v3 41/51] powerpc/PCI: " Yinghai Lu
2015-07-27 23:30 ` [PATCH v3 42/51] of/PCI: Add IORESOURCE_MEM_64 for 64-bit resource Yinghai Lu
2015-07-27 23:30 ` [PATCH v3 43/51] PCI: Treat optional as must in first try for bridge rescan Yinghai Lu
2015-07-27 23:30 ` [PATCH v3 44/51] PCI: Get new realloc size for bridge for last try Yinghai Lu
2015-07-27 23:30 ` [PATCH v3 45/51] PCI: Don't release sibiling bridge resources during hotplug Yinghai Lu
2015-07-27 23:30 ` [PATCH v3 46/51] PCI: Don't release fixed resource for realloc Yinghai Lu
2015-07-27 23:30 ` [PATCH v3 47/51] PCI: Claim fixed resource during remove/rescan path Yinghai Lu
2015-07-27 23:30 ` [PATCH v3 48/51] PCI: Set resource to FIXED for lsi devices Yinghai Lu
2015-07-27 23:30 ` [PATCH v3 49/51] PCI, x86: Add pci=assign_pref_bars to re-allocate pref bars Yinghai Lu
2015-07-27 23:30 ` [PATCH v3 50/51] PCI: Introduce resource_disabled() Yinghai Lu
2015-07-27 23:30 ` [PATCH v3 51/51] PCI: Don't set flags to 0 when assign resource fail Yinghai Lu
2015-08-17 22:48 ` [PATCH v3 00/51] PCI: Resource allocation cleanup for v4.3 Bjorn Helgaas
2015-08-18 18:43 ` Yinghai Lu
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=20150818042135.GZ26431@google.com \
--to=bhelgaas@google.com \
--cc=akpm@linux-foundation.org \
--cc=benh@kernel.crashing.org \
--cc=davem@davemloft.net \
--cc=linux-kernel@vger.kernel.org \
--cc=linux-pci@vger.kernel.org \
--cc=linux@iam.tj \
--cc=wangyijing@huawei.com \
--cc=weiyang@linux.vnet.ibm.com \
--cc=yinghai@kernel.org \
/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).