From: Alexandru Elisei <alexandru.elisei@arm.com>
To: Andre Przywara <andre.przywara@arm.com>
Cc: kvm@vger.kernel.org,
Julien Thierry <julien.thierry.kdev@gmail.com>,
Suzuki Poulose <suzuki.poulose@arm.com>,
Julien Grall <julien.grall.oss@gmail.com>,
Marc Zyngier <maz@kernel.org>, Will Deacon <will@kernel.org>
Subject: Re: [PATCH kvmtool 01/16] arm: Allow use of hugepage with 16K pagesize host
Date: Wed, 6 Nov 2019 17:29:15 +0000 [thread overview]
Message-ID: <de4ec309-f6b8-c2fc-0936-0e6266756900@arm.com> (raw)
In-Reply-To: <20191106164711.77673d43@donnerap.cambridge.arm.com>
Hi,
On 11/6/19 4:47 PM, Andre Przywara wrote:
> On Mon, 23 Sep 2019 14:35:07 +0100
> Alexandru Elisei <alexandru.elisei@arm.com> wrote:
>
> Hi,
>
>> From: Suzuki K Poulose <suzuki.poulose@arm.com>
>>
>> With 16K pagesize, the hugepage size is 32M. Align the guest
>> memory to the hugepagesize for 16K.
>>
>> To query the host page size, we use sysconf(_SC_PAGESIZE) instead of
>> getpagesize, as suggested by man 2 getpagesize for portable applications.
>> Also use the sysconf function instead of getpagesize when setting
>> kvm->ram_pagesize.
>>
>> Cc: Marc Zyngier <marc.zyngier@arm.com>
>> Cc: Andre Przywara <andre.przywara@arm.com>
>> Cc: Will Deacon <will.deacon@arm.com>
>> Signed-off-by: Suzuki K Poulose <suzuki.poulose@arm.com>
>> Signed-off-by: Julien Grall <julien.grall@arm.com>
>> Tested-by: Alexandru Elisei <alexandru.elisei@arm.com>
>> Signed-off-by: Alexandru Elisei <alexandru.elisei@arm.com>
>> ---
>> arm/kvm.c | 36 +++++++++++++++++++++++++++++-------
>> builtin-run.c | 4 ++--
>> util/util.c | 2 +-
>> 3 files changed, 32 insertions(+), 10 deletions(-)
>>
>> diff --git a/arm/kvm.c b/arm/kvm.c
>> index 1f85fc60588f..1c5bdb8026bf 100644
>> --- a/arm/kvm.c
>> +++ b/arm/kvm.c
>> @@ -59,14 +59,33 @@ void kvm__arch_set_cmdline(char *cmdline, bool video)
>>
>> void kvm__arch_init(struct kvm *kvm, const char *hugetlbfs_path, u64 ram_size)
>> {
>> + unsigned long alignment;
>> +
>> /*
>> - * Allocate guest memory. We must align our buffer to 64K to
>> - * correlate with the maximum guest page size for virtio-mmio.
>> - * If using THP, then our minimal alignment becomes 2M.
>> - * 2M trumps 64K, so let's go with that.
>> + * Allocate guest memory. If the user wants to use hugetlbfs, then the
>> + * specified guest memory size must be a multiple of the host huge page
>> + * size in order for the allocation to succeed. The mmap return adress
>> + * is naturally aligned to the huge page size, so in this case we don't
>> + * need to perform any alignment.
>> + *
>> + * Otherwise, we must align our buffer to 64K to correlate with the
>> + * maximum guest page size for virtio-mmio. If using THP, then our
>> + * minimal alignment becomes 2M with a 4K page size. With a 16K page
>> + * size, the alignment becomes 32M. 32M and 2M trump 64K, so let's go
>> + * with the largest alignment supported by the host.
>> */
>> + if (hugetlbfs_path) {
>> + /* Don't do any alignment. */
>> + alignment = 0;
>> + } else {
>> + if (sysconf(_SC_PAGESIZE) == SZ_16K)
>> + alignment = SZ_32M;
>> + else
>> + alignment = SZ_2M;
>> + }
>> +
>> kvm->ram_size = min(ram_size, (u64)ARM_MAX_MEMORY(kvm));
>> - kvm->arch.ram_alloc_size = kvm->ram_size + SZ_2M;
>> + kvm->arch.ram_alloc_size = kvm->ram_size + alignment;
> So that means that on a 16K page size host we always allocate 32MB more memory than requested. In practise the pages before the new start should stay unpopulated, but I wonder if we should munmap that unused region before the new start?
> Just thinking that people tend to use kvmtool because of its smaller memory footprint ...
I don't think it matters, kvmtool will not touch that region, so the process'
resident set size will stay the same.
Thanks,
Alex
> Otherwise the code looks alright.
>
> Cheers,
> Andre.
>
>> kvm->arch.ram_alloc_start = mmap_anon_or_hugetlbfs(kvm, hugetlbfs_path,
>> kvm->arch.ram_alloc_size);
>>
>> @@ -74,8 +93,11 @@ void kvm__arch_init(struct kvm *kvm, const char *hugetlbfs_path, u64 ram_size)
>> die("Failed to map %lld bytes for guest memory (%d)",
>> kvm->arch.ram_alloc_size, errno);
>>
>> - kvm->ram_start = (void *)ALIGN((unsigned long)kvm->arch.ram_alloc_start,
>> - SZ_2M);
>> + kvm->ram_start = kvm->arch.ram_alloc_start;
>> + /* The result of aligning to 0 is 0. Let's avoid that. */
>> + if (alignment)
>> + kvm->ram_start = (void *)ALIGN((unsigned long)kvm->ram_start,
>> + alignment);
>>
>> madvise(kvm->arch.ram_alloc_start, kvm->arch.ram_alloc_size,
>> MADV_MERGEABLE);
>> diff --git a/builtin-run.c b/builtin-run.c
>> index f8dc6c7229b0..c867c8ba0892 100644
>> --- a/builtin-run.c
>> +++ b/builtin-run.c
>> @@ -127,8 +127,8 @@ void kvm_run_set_wrapper_sandbox(void)
>> "Run this script when booting into custom" \
>> " rootfs"), \
>> OPT_STRING('\0', "hugetlbfs", &(cfg)->hugetlbfs_path, "path", \
>> - "Hugetlbfs path"), \
>> - \
>> + "Hugetlbfs path. Memory size must be a multiple"\
>> + " of the huge page size"), \
>> OPT_GROUP("Kernel options:"), \
>> OPT_STRING('k', "kernel", &(cfg)->kernel_filename, "kernel", \
>> "Kernel to boot in virtual machine"), \
>> diff --git a/util/util.c b/util/util.c
>> index 1877105e3c08..217addd75e6f 100644
>> --- a/util/util.c
>> +++ b/util/util.c
>> @@ -127,7 +127,7 @@ void *mmap_anon_or_hugetlbfs(struct kvm *kvm, const char *hugetlbfs_path, u64 si
>> */
>> return mmap_hugetlbfs(kvm, hugetlbfs_path, size);
>> else {
>> - kvm->ram_pagesize = getpagesize();
>> + kvm->ram_pagesize = sysconf(_SC_PAGESIZE);
>> return mmap(NULL, size, PROT_RW, MAP_ANON_NORESERVE, -1, 0);
>> }
>> }
next prev parent reply other threads:[~2019-11-06 17:29 UTC|newest]
Thread overview: 40+ messages / expand[flat|nested] mbox.gz Atom feed top
2019-09-23 13:35 [PATCH kvmtool 00/16] arm: Allow the user to define the memory layout Alexandru Elisei
2019-09-23 13:35 ` [PATCH kvmtool 01/16] arm: Allow use of hugepage with 16K pagesize host Alexandru Elisei
2019-11-06 16:47 ` Andre Przywara
2019-11-06 17:29 ` Alexandru Elisei [this message]
2019-09-23 13:35 ` [PATCH kvmtool 02/16] kvm__arch_init: Don't pass hugetlbfs_path and ram_size in parameter Alexandru Elisei
2019-11-06 16:47 ` Andre Przywara
2019-11-07 10:03 ` Alexandru Elisei
2019-09-23 13:35 ` [PATCH kvmtool 03/16] virtio/scsi: Allow the use of multiple banks Alexandru Elisei
2019-11-06 16:48 ` Andre Przywara
2020-02-05 18:07 ` Suzuki Kuruppassery Poulose
2019-09-23 13:35 ` [PATCH kvmtool 04/16] kvmtool: Add helper to sanitize arch specific KVM configuration Alexandru Elisei
2019-11-06 16:48 ` Andre Przywara
2019-11-07 10:05 ` Alexandru Elisei
2020-02-05 18:16 ` Suzuki Kuruppassery Poulose
2019-09-23 13:35 ` [PATCH kvmtool 05/16] kvmtool: Use MB consistently Alexandru Elisei
2019-11-06 16:49 ` Andre Przywara
2020-02-05 18:17 ` Suzuki Kuruppassery Poulose
2019-09-23 13:35 ` [PATCH kvmtool 06/16] builtin-run.c: Always use ram_size in bytes Alexandru Elisei
2019-11-06 16:49 ` Andre Przywara
2019-11-07 10:08 ` Alexandru Elisei
2020-02-05 19:03 ` Suzuki Kuruppassery Poulose
2019-09-23 13:35 ` [PATCH kvmtool 07/16] arm: Remove redundant define ARM_PCI_CFG_SIZE Alexandru Elisei
2019-11-06 16:49 ` Andre Przywara
2020-02-06 11:49 ` Suzuki Kuruppassery Poulose
2019-09-23 13:35 ` [PATCH kvmtool 08/16] arm: Move anything related to RAM initialization in kvm__init_ram Alexandru Elisei
2019-11-07 13:46 ` Andre Przywara
2019-09-23 13:35 ` [PATCH kvmtool 09/16] arm: Allow the user to specify RAM base address Alexandru Elisei
2019-11-07 13:54 ` Andre Przywara
2020-02-06 12:20 ` Suzuki Kuruppassery Poulose
2019-09-23 13:35 ` [PATCH kvmtool 10/16] kvmtool: Allow standard size specifiers for memory Alexandru Elisei
2019-11-07 13:55 ` Andre Przywara
2019-09-23 13:35 ` [PATCH kvmtool 11/16] arm/pci: Remove unused ioports Alexandru Elisei
2019-09-23 13:35 ` [PATCH kvmtool 12/16] Fold kvm__init_ram call in kvm__arch_init and rename it Alexandru Elisei
2019-09-23 13:35 ` [PATCH kvmtool 13/16] arm: Allow any base address for RAM Alexandru Elisei
2019-09-23 13:35 ` [PATCH kvmtool 14/16] arm: Move memory related code to memory.c Alexandru Elisei
2019-09-23 13:35 ` [PATCH kvmtool 15/16] kvmtool: Make the size@addr option parser globally visible Alexandru Elisei
2019-09-23 13:35 ` [PATCH kvmtool 16/16] arm: Allow the user to define the MMIO regions Alexandru Elisei
2020-02-05 17:16 ` [PATCH kvmtool 00/16] arm: Allow the user to define the memory layout Will Deacon
2020-02-05 17:18 ` Alexandru Elisei
2020-02-06 9:20 ` Marc Zyngier
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=de4ec309-f6b8-c2fc-0936-0e6266756900@arm.com \
--to=alexandru.elisei@arm.com \
--cc=andre.przywara@arm.com \
--cc=julien.grall.oss@gmail.com \
--cc=julien.thierry.kdev@gmail.com \
--cc=kvm@vger.kernel.org \
--cc=maz@kernel.org \
--cc=suzuki.poulose@arm.com \
--cc=will@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).