qemu-devel.nongnu.org archive mirror
 help / color / mirror / Atom feed
From: Linu Cherian <linuc.decode@gmail.com>
To: Eric Auger <eric.auger@redhat.com>
Cc: eric.auger.pro@gmail.com, peter.maydell@linaro.org,
	qemu-arm@nongnu.org, qemu-devel@nongnu.org,
	prem.mallappa@gmail.com, alex.williamson@redhat.com,
	mohun106@gmail.com, drjones@redhat.com, tcain@qti.qualcomm.com,
	Radha.Chintakuntla@cavium.com, Sunil.Goutham@cavium.com,
	mst@redhat.com, jean-philippe.brucker@arm.com, tn@semihalf.com,
	will.deacon@arm.com, robin.murphy@arm.com, peterx@redhat.com,
	bharat.bhushan@nxp.com, christoffer.dall@linaro.org,
	wtownsen@redhat.com, linu.cherian@cavium.com
Subject: Re: [Qemu-devel] [Qemu-arm] [PATCH v7 19/20] hw/arm/smmuv3: [not for upstream] add SMMU_CMD_TLBI_NH_VA_AM handling
Date: Tue, 17 Oct 2017 20:36:18 +0530	[thread overview]
Message-ID: <20171017150618.GA6575@virtx40> (raw)
In-Reply-To: <1504286483-23327-20-git-send-email-eric.auger@redhat.com>

Hi Eric,

On Fri Sep 01, 2017 at 07:21:22PM +0200, Eric Auger wrote:
> SMMUV3 does not support any IOVA range TLBI command:
> SMMU_CMD_TLBI_NH_VA invalidates TLB entries by page.
> That's an issue when running DPDK on guest. DPDK uses
> hugepages but each time a hugepage is mapped on guest side,
> a storm of SMMU_CMD_TLBI_NH_VA commands get sent by the
> guest smmuv3 driver and trapped by QEMU for VFIO replay.
> 
> Let's get prepared to handle implementation defined commands,
> SMMU_CMD_TLBI_NH_VA_VM, which invalidate a range of IOVAs.
> 
> Upon this command, we notify the whole range in one host.
> 
> Signed-off-by: Eric Auger <eric.auger@redhat.com>
> ---
>  hw/arm/smmuv3-internal.h |  1 +
>  hw/arm/smmuv3.c          | 13 +++++++++++++
>  hw/arm/trace-events      |  1 +
>  3 files changed, 15 insertions(+)
> 
> diff --git a/hw/arm/smmuv3-internal.h b/hw/arm/smmuv3-internal.h
> index f9f95ae..e70cf76 100644
> --- a/hw/arm/smmuv3-internal.h
> +++ b/hw/arm/smmuv3-internal.h
> @@ -289,6 +289,7 @@ enum {
>      SMMU_CMD_RESUME          = 0x44,
>      SMMU_CMD_STALL_TERM,
>      SMMU_CMD_SYNC,          /* 0x46 */
> +    SMMU_CMD_TLBI_NH_VA_AM   = 0x8F, /* VIOMMU Impl Defined */
>  };
>  
>  static const char *cmd_stringify[] = {
> diff --git a/hw/arm/smmuv3.c b/hw/arm/smmuv3.c
> index 9c8640f..55dc80b 100644
> --- a/hw/arm/smmuv3.c
> +++ b/hw/arm/smmuv3.c
> @@ -880,6 +880,19 @@ static int smmuv3_cmdq_consume(SMMUV3State *s)
>              smmuv3_replay_iova_range(&s->smmu_state, addr, size);
>              break;
>          }
> +        case SMMU_CMD_TLBI_NH_VA_AM:
> +        {
> +            int asid = extract32(cmd.word[1], 16, 16);
> +            int am = extract32(cmd.word[1], 0, 16);
> +            uint64_t low = extract32(cmd.word[2], 12, 20);
> +            uint64_t high = cmd.word[3];
> +            uint64_t addr = high << 32 | (low << 12);
> +            size_t size = am << 12;
> 


While testing dpdk, observed that there are map requests coming to 
arm smmuv3 driver with size greater than 256M. Since the current 
code supports only 256M( 16 + 12 bits), had to abuse the asid field
to pass the extra bits for address mask to get things working.

diff --git a/drivers/iommu/arm-smmu-v3.c b/drivers/iommu/arm-smmu-v3.c
index 04e2d75..51b1d07 100644
--- a/drivers/iommu/arm-smmu-v3.c
+++ b/drivers/iommu/arm-smmu-v3.c
@@ -1418,7 +1418,10 @@ static void arm_smmu_tlb_inv_range_nosync(unsigned long iova, size_t size,
                if (smmu->options & ARM_SMMU_OPT_TLBI_ON_MAP) {
                        cmd.opcode      = CMDQ_OP_TLBI_NH_VA_AM;
                        cmd.tlbi.am   = size >> 12;
+                       cmd.tlbi.asid = size >> 28;
                        granule = size;


On Qemu side,

  if (cfg.disabled || cfg.bypassed) {
@@ -884,12 +899,15 @@ static int smmuv3_cmdq_consume(SMMUV3State *s)
         case SMMU_CMD_TLBI_NH_VA_AM:
         {
             int asid = extract32(cmd.word[1], 16, 16);
             int am = extract32(cmd.word[1], 0, 16);
             uint64_t low = extract32(cmd.word[2], 12, 20);
             uint64_t high = cmd.word[3];
             uint64_t addr = high << 32 | (low << 12);
             size_t size = am << 12;
 
+           am = am | asid << 16;
+           size = am << 12;        
+



+
> +            trace_smmuv3_cmdq_tlbi_nh_va_am(asid, am, addr, size);
> +            smmuv3_replay_iova_range(&s->smmu_state, addr, size);
> +            break;
> +        }
>          case SMMU_CMD_TLBI_NH_VAA:
>          case SMMU_CMD_TLBI_EL3_ALL:
>          case SMMU_CMD_TLBI_EL3_VA:
> diff --git a/hw/arm/trace-events b/hw/arm/trace-events
> index 15f84d6..fba33ac 100644
> --- a/hw/arm/trace-events
> +++ b/hw/arm/trace-events
> @@ -26,6 +26,7 @@ smmuv3_cmdq_opcode(const char *opcode) "<--- %s"
>  smmuv3_cmdq_cfgi_ste(int streamid) "     |_ streamid =%d"
>  smmuv3_cmdq_cfgi_ste_range(int start, int end) "     |_ start=0x%d - end=0x%d"
>  smmuv3_cmdq_tlbi_nh_va(int asid, int vmid, uint64_t addr) "     |_ asid =%d vmid =%d addr=0x%"PRIx64
> +smmuv3_cmdq_tlbi_nh_va_am(int asid, int am, size_t size, uint64_t addr) "     |_ asid =%d am =%d size=0x%lx addr=0x%"PRIx64
>  smmuv3_cmdq_consume_out(uint8_t prod_wrap, uint32_t prod, uint8_t cons_wrap, uint32_t cons) "prod_wrap:%d, prod:0x%x cons_wrap:%d cons:0x%x"
>  smmuv3_update(bool is_empty, uint32_t prod, uint32_t cons, uint8_t prod_wrap, uint8_t cons_wrap) "q empty:%d prod:%d cons:%d p.wrap:%d p.cons:%d"
>  smmuv3_update_check_cmd(int error) "cmdq not enabled or error :0x%x"
> -- 
> 2.5.5
> 
> 

-- 
Linu cherian

  parent reply	other threads:[~2017-10-17 15:06 UTC|newest]

Thread overview: 72+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2017-09-01 17:21 [Qemu-devel] [PATCH v7 00/20] ARM SMMUv3 Emulation Support Eric Auger
2017-09-01 17:21 ` [Qemu-devel] [PATCH v7 01/20] hw/arm/smmu-common: smmu base device and datatypes Eric Auger
2017-09-27 17:38   ` Peter Maydell
2017-09-28  7:57     ` Auger Eric
2017-09-30  8:28     ` Prem Mallappa
2017-10-02  7:43       ` Auger Eric
2017-09-01 17:21 ` [Qemu-devel] [PATCH v7 02/20] hw/arm/smmu-common: IOMMU memory region and address space setup Eric Auger
2017-10-09 14:39   ` Peter Maydell
2017-09-01 17:21 ` [Qemu-devel] [PATCH v7 03/20] hw/arm/smmu-common: smmu_read/write_sysmem Eric Auger
2017-10-09 14:46   ` Peter Maydell
2017-09-01 17:21 ` [Qemu-devel] [PATCH v7 04/20] hw/arm/smmu-common: VMSAv8-64 page table walk Eric Auger
2017-10-09 15:36   ` Peter Maydell
2017-09-01 17:21 ` [Qemu-devel] [PATCH v7 05/20] hw/arm/smmuv3: Skeleton Eric Auger
2017-09-08 10:52   ` [Qemu-devel] [Qemu-arm] " Linu Cherian
2017-09-08 15:18     ` Auger Eric
2017-09-12  6:14       ` Linu Cherian
2017-10-09 16:17   ` [Qemu-devel] " Peter Maydell
2017-09-01 17:21 ` [Qemu-devel] [PATCH v7 06/20] hw/arm/smmuv3: Wired IRQ and GERROR helpers Eric Auger
2017-10-09 17:01   ` Peter Maydell
2017-09-01 17:21 ` [Qemu-devel] [PATCH v7 07/20] hw/arm/smmuv3: Queue helpers Eric Auger
2017-10-09 17:12   ` Peter Maydell
2017-09-01 17:21 ` [Qemu-devel] [PATCH v7 08/20] hw/arm/smmuv3: Implement MMIO write operations Eric Auger
2017-10-09 17:17   ` Peter Maydell
2017-09-01 17:21 ` [Qemu-devel] [PATCH v7 09/20] hw/arm/smmuv3: Event queue recording helper Eric Auger
2017-10-09 17:34   ` Peter Maydell
2017-09-01 17:21 ` [Qemu-devel] [PATCH v7 10/20] hw/arm/smmuv3: Implement translate callback Eric Auger
2017-10-09 17:45   ` Peter Maydell
2018-02-06 12:19     ` Auger Eric
2018-02-06 12:43       ` Peter Maydell
2018-02-06 12:56         ` Auger Eric
2017-09-01 17:21 ` [Qemu-devel] [PATCH v7 11/20] target/arm/kvm: Translate the MSI doorbell in kvm_arch_fixup_msi_route Eric Auger
2017-09-01 17:21 ` [Qemu-devel] [PATCH v7 12/20] hw/arm/smmuv3: Implement data structure and TLB invalidation notifications Eric Auger
2017-09-01 17:21 ` [Qemu-devel] [PATCH v7 13/20] hw/arm/smmuv3: Implement IOMMU memory region replay callback Eric Auger
2017-09-14  9:27   ` [Qemu-devel] [Qemu-arm] " Linu Cherian
2017-09-14 14:31     ` Tomasz Nowicki
2017-09-14 14:43       ` Tomasz Nowicki
2017-09-15  7:30         ` Auger Eric
2017-09-15  7:41           ` Auger Eric
2017-09-15 10:42           ` tn
2017-09-15 13:19             ` Auger Eric
2017-09-15 14:50             ` Auger Eric
2017-09-18  9:50               ` Tomasz Nowicki
2017-09-15  7:23     ` Auger Eric
2017-09-01 17:21 ` [Qemu-devel] [PATCH v7 14/20] hw/arm/virt: Store the PCI host controller dt phandle Eric Auger
2017-09-01 17:21 ` [Qemu-devel] [PATCH v7 15/20] hw/arm/sysbus-fdt: Pass the VirtMachineState to the node creation functions Eric Auger
2017-10-09 17:47   ` Peter Maydell
2017-11-13 13:00     ` Auger Eric
2017-11-13 13:08       ` Peter Maydell
2017-11-13 13:37         ` Auger Eric
2017-11-13 13:44           ` Peter Maydell
2017-11-13 13:59             ` Auger Eric
2017-09-01 17:21 ` [Qemu-devel] [PATCH v7 16/20] hw/arm/sysbus-fdt: Pass the platform bus base address in PlatformBusFDTData Eric Auger
2017-09-01 17:21 ` [Qemu-devel] [PATCH v7 17/20] hw/arm/sysbus-fdt: Allow smmuv3 dynamic instantiation Eric Auger
2017-09-01 17:21 ` [Qemu-devel] [PATCH v7 18/20] hw/arm/virt-acpi-build: Add smmuv3 node in IORT table Eric Auger
2017-09-01 17:21 ` [Qemu-devel] [PATCH v7 19/20] hw/arm/smmuv3: [not for upstream] add SMMU_CMD_TLBI_NH_VA_AM handling Eric Auger
2017-10-09 17:48   ` Peter Maydell
2017-10-17 15:06   ` Linu Cherian [this message]
2017-09-01 17:21 ` [Qemu-devel] [PATCH v7 20/20] hw/arm/smmuv3: [not for upstream] Add caching-mode option Eric Auger
2017-10-09 17:49   ` Peter Maydell
2017-09-07 12:39 ` [Qemu-devel] [PATCH v7 00/20] ARM SMMUv3 Emulation Support Peter Maydell
2017-09-08  8:35   ` Auger Eric
2017-09-08  5:47 ` Michael S. Tsirkin
2017-09-08  8:36   ` Auger Eric
2017-09-12  6:18 ` [Qemu-devel] [Qemu-arm] " Linu Cherian
2017-09-12  6:38   ` Auger Eric
2017-09-28  6:43 ` Linu Cherian
2017-09-28  7:13   ` Peter Xu
2017-09-28  7:54     ` Auger Eric
2017-09-28  9:21       ` Linu Cherian
2017-10-24  5:38 ` Linu Cherian
2017-10-24 10:20   ` Will Deacon
2017-10-24 17:06     ` Linu Cherian

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=20171017150618.GA6575@virtx40 \
    --to=linuc.decode@gmail.com \
    --cc=Radha.Chintakuntla@cavium.com \
    --cc=Sunil.Goutham@cavium.com \
    --cc=alex.williamson@redhat.com \
    --cc=bharat.bhushan@nxp.com \
    --cc=christoffer.dall@linaro.org \
    --cc=drjones@redhat.com \
    --cc=eric.auger.pro@gmail.com \
    --cc=eric.auger@redhat.com \
    --cc=jean-philippe.brucker@arm.com \
    --cc=linu.cherian@cavium.com \
    --cc=mohun106@gmail.com \
    --cc=mst@redhat.com \
    --cc=peter.maydell@linaro.org \
    --cc=peterx@redhat.com \
    --cc=prem.mallappa@gmail.com \
    --cc=qemu-arm@nongnu.org \
    --cc=qemu-devel@nongnu.org \
    --cc=robin.murphy@arm.com \
    --cc=tcain@qti.qualcomm.com \
    --cc=tn@semihalf.com \
    --cc=will.deacon@arm.com \
    --cc=wtownsen@redhat.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).