xen-devel.lists.xenproject.org archive mirror
 help / color / mirror / Atom feed
From: Julien Grall <julien.grall@citrix.com>
To: vijay.kilari@gmail.com, Ian.Campbell@citrix.com,
	stefano.stabellini@eu.citrix.com, stefano.stabellini@citrix.com,
	tim@xen.org, xen-devel@lists.xen.org
Cc: Prasun.Kapoor@caviumnetworks.com,
	Vijaya Kumar K <Vijaya.Kumar@caviumnetworks.com>,
	manish.jaggi@caviumnetworks.com
Subject: Re: [PATCH v6 16/31] xen/arm: ITS: Add virtual ITS commands support
Date: Thu, 3 Sep 2015 16:07:09 +0100	[thread overview]
Message-ID: <55E8621D.6090601@citrix.com> (raw)
In-Reply-To: <1441019208-2764-17-git-send-email-vijay.kilari@gmail.com>

Hi Vijay,

This patch looks good to me. Mostly coding style comment and question
about your code/comments. See below.

On 31/08/15 12:06, vijay.kilari@gmail.com wrote:
> diff --git a/xen/arch/arm/vgic-v3-its.c b/xen/arch/arm/vgic-v3-its.c
> index 14c38b3..fabbad0 100644
> --- a/xen/arch/arm/vgic-v3-its.c
> +++ b/xen/arch/arm/vgic-v3-its.c

[...]

> +bool_t is_valid_collection(struct domain *d, uint32_t col)

You should have define the prototype of is_valid_collection within this
patch rather than in patch #13.

> +{
> +    return (col <= (d->max_vcpus + 1));

col <= (vtis_get_max_collections(d)) to avoid hardcoding the number of
collection in multiple place.

> +}
> +
> +static inline uint16_t vits_get_max_collections(struct domain *d)
> +{
> +    /*
> +     * ITS only supports upto 256 collections without

up to

> +     * provisioning external memory. As per vITS design, number of
> +     * vCPUS should not exceed max number of collections.
> +     */
> +    ASSERT(d->max_vcpus < 256);

The problem with ASSERT is it's disappearing with non-debug build. Do
you ensure somewhere that it's never possible to create a domain with
more than 256 CPUs when vITS is in use?

Otherwise, bad things would happen if we are trying to create a guest
with more than 256 vCPUs.

> +
> +    return (d->max_vcpus + 1);

As said on v5, please add a comment to explain why d->max_vcpus + 1. It
could be a reference to the public spec...

> +}
> +

[...]

> +static int vits_process_int(struct vcpu *v, struct vgic_its *vits,
> +                            its_cmd_block *virt_cmd)
> +{
> +    uint32_t event, dev_id ;

Spurious space after dev_id.

> +
> +    event = virt_cmd->int_cmd.cmd;
> +    dev_id = virt_cmd->int_cmd.devid;
> +
> +    DPRINTK("%pv: vITS: INT: Device 0x%"PRIx32" id %"PRIu32"\n",
> +            v, dev_id, event);
> +
> +    /* TODO: Inject LPI */
> +
> +    return 0;
> +}


[...]

> +static int vits_process_mapc(struct vcpu *v, struct vgic_its *vits,
> +                             its_cmd_block *virt_cmd)
> +{
> +    uint16_t vcol_id;
> +    uint64_t vta = 0;

Setting vta to 0 is not necessary as you override the value 2 lines below.

Note that you could have directly assign the value in the declaration of
the variables. It would have drop 2 lines.

> +
> +    vcol_id = virt_cmd->mapc.col;
> +    vta = virt_cmd->mapc.ta;
> +
> +    DPRINTK("%pv: vITS: MAPC: vCID %"PRIu16" vTA 0x%"PRIx64" valid %"PRIu8"\n",
> +            v, vcol_id, vta, virt_cmd->mapc.valid);
> +
> +    if ( !is_valid_collection(v->domain, vcol_id) )
> +        return -EINVAL;
> +
> +    if ( virt_cmd->mapc.valid )
> +    {
> +        if ( vta > v->domain->max_vcpus )
> +            return -EINVAL;
> +        vits->collections[vcol_id].target_address = vta;
> +    }
> +    else
> +        vits->collections[vcol_id].target_address = INVALID_PADDR;
> +
> +    return 0;
> +}

[...]

> +static int vits_read_virt_cmd(struct vcpu *v, struct vgic_its *vits,
> +                              its_cmd_block *virt_cmd)
> +{
> +    paddr_t maddr;
> +    struct domain *d = v->domain;
> +    int ret;
> +
> +    ASSERT(spin_is_locked(&vits->lock));
> +
> +    if ( !(vits->cmd_base & GITS_CBASER_VALID) )
> +    {
> +        dprintk(XENLOG_G_ERR, "%pv: vITS: Invalid CBASER\n", v);
> +        return 0;
> +    }
> +
> +    /* Map only the page that is required */

IHMO the "CMD Q can be more than 1 page" was valid and useful. Sorry if
I wasn't enough clear that I wanted the typo fixed on the second sentence.

> +    maddr = (vits->cmd_base & GITS_CBASER_PA_MASK) +
> +             atomic_read(&vits->cmd_read);
> +
> +    DPRINTK("%pv: vITS: Mapping CMD Q maddr 0x%"PRIx64" read 0x%"PRIx32"\n",
> +            v, maddr, atomic_read(&vits->cmd_read));
> +
> +    ret = vits_access_guest_table(d, maddr, (void *)virt_cmd,
> +                                  sizeof(its_cmd_block), 0);
> +    if ( ret )
> +    {
> +        dprintk(XENLOG_G_ERR,
> +                "%pv: vITS: Failed to get command page @page 0x%"PRIx32"\n",
> +                v, atomic_read(&vits->cmd_read));
> +        return -EINVAL;
> +    }
> +
> +    /* No command queue is created by vits to check on Q full */

I don't understand this comment. What do you mean?

> +    atomic_add(sizeof(its_cmd_block), &vits->cmd_read);
> +    if ( atomic_read(&vits->cmd_read) == vits->cmd_qsize )
> +    {
> +         DPRINTK("%pv: vITS: Reset read @ 0x%"PRIx32" qsize 0x%"PRIx64"\n",
> +                 v, atomic_read(&vits->cmd_read), vits->cmd_qsize);
> +
> +         atomic_set(&vits->cmd_read, 0);
> +    }
> +
> +    return 0;
> +}

[...]

> +int vits_domain_init(struct domain *d)
> +{
> +    struct vgic_its *vits;
> +    int i;
> +
> +    ASSERT(is_hardware_domain(d));
> +
> +    d->arch.vgic.vits = xzalloc(struct vgic_its);
> +    if ( !d->arch.vgic.vits )
> +        return -ENOMEM;
> +
> +    vits = d->arch.vgic.vits;
> +
> +    spin_lock_init(&vits->lock);
> +
> +    vits->collections = xzalloc_array(struct its_collection,
> +                                      vits_get_max_collections(d));
> +    if ( !vits->collections )
> +        return -ENOMEM;
> +
> +    for ( i = 0; i < vits_get_max_collections(d); i++ )
> +        vits->collections[i].target_address = ~0UL;

You are using 2 different values for the invalid address: ~0UL and
INVALID_PADDR.

They are not defined the same way and even though they may give the same
result for target_address by luck you should always use the same
definition everywhere.

I would prefer to see INVALID_PADDR even though a define
VITS_TARGET_INVALID (or similar) would have been nice. Although, I'm
fine if you don't add the define.

> +
> +    return 0;
> +}

[...]

> diff --git a/xen/include/asm-arm/gic-its.h b/xen/include/asm-arm/gic-its.h
> index 42f6551..4327ba2 100644
> --- a/xen/include/asm-arm/gic-its.h
> +++ b/xen/include/asm-arm/gic-its.h
> @@ -21,6 +21,7 @@
>  #include <asm/gic_v3_defs.h>
>  #include <xen/rbtree.h>
>  
> +#define MAPC_ITT_IPA_SHIFT              8

I would much prefer to see this define with the definition of the
command structure.

>  /*
>   * ITS registers, offsets from ITS_base
>   */
> @@ -59,6 +60,7 @@
>  #define GITS_CBASER_InnerShareable      (1UL << 10)
>  #define GITS_CBASER_SHAREABILITY_MASK   (3UL << 10)
>  #define GITS_CBASER_CACHEABILITY_MASK   (7UL << 59)
> +#define GITS_CBASER_PA_MASK             (0xfffffffff000UL)
>  
>  #define GITS_BASER_NR_REGS              8
>  

Regards,

-- 
Julien Grall

  reply	other threads:[~2015-09-03 15:07 UTC|newest]

Thread overview: 87+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2015-08-31 11:06 [PATCH v6 00/31] Add ITS support vijay.kilari
2015-08-31 11:06 ` [PATCH v6 01/31] xen/dt: Handle correctly node with interrupt-map in dt_for_each_irq_map vijay.kilari
2015-08-31 14:20   ` Julien Grall
2015-09-02 15:28     ` Ian Campbell
2015-09-02 15:30       ` Wei Liu
2015-09-02 15:45       ` Julien Grall
2015-09-02 15:52         ` Ian Campbell
2015-09-04 14:41         ` Ian Campbell
2015-08-31 11:06 ` [PATCH v6 02/31] xen/arm: Add bitmap_find_next_zero_area helper function vijay.kilari
2015-08-31 11:06 ` [PATCH v6 03/31] xen: Add log2 functionality vijay.kilari
2015-08-31 11:21   ` Jan Beulich
2015-08-31 11:06 ` [PATCH v6 04/31] xen/arm: Set nr_cpu_ids to available number of cpus vijay.kilari
2015-08-31 14:25   ` Julien Grall
2015-09-09 12:48     ` Ian Campbell
2015-08-31 11:06 ` [PATCH v6 05/31] xen/arm: Rename NR_IRQs and vgic_num_irqs helper function vijay.kilari
2015-08-31 14:40   ` Julien Grall
2015-09-09 13:08     ` Ian Campbell
2015-09-09 13:23       ` Julien Grall
2015-08-31 11:06 ` [PATCH v6 06/31] xen/arm: ITS: Port ITS driver to Xen vijay.kilari
2015-08-31 15:41   ` Julien Grall
2015-09-03 17:02   ` Julien Grall
2015-08-31 11:06 ` [PATCH v6 07/31] xen/arm: ITS: Add helper functions to manage its_devices vijay.kilari
2015-08-31 11:06 ` [PATCH v6 08/31] xen/arm: ITS: Introduce msi_desc for LPIs vijay.kilari
2015-08-31 16:20   ` Julien Grall
2015-09-09 13:16   ` Ian Campbell
2015-09-09 13:28     ` Julien Grall
2015-08-31 11:06 ` [PATCH v6 09/31] xen/arm: ITS: Add APIs to add and assign device vijay.kilari
2015-09-03 17:34   ` Julien Grall
2015-09-09 13:28     ` Ian Campbell
2015-09-09 13:44       ` Julien Grall
2015-09-09 15:07         ` Ian Campbell
2015-09-09 16:19           ` Julien Grall
2015-08-31 11:06 ` [PATCH v6 10/31] xen/arm: ITS: Introduce gic_is_lpi helper function vijay.kilari
2015-08-31 16:49   ` Julien Grall
2015-09-01  9:02     ` Vijay Kilari
2015-09-01 11:40       ` Julien Grall
2015-09-01 11:56         ` Vijay Kilari
2015-09-01 13:02           ` Julien Grall
2015-09-03  6:32             ` Vijay Kilari
2015-09-03  9:48               ` Julien Grall
2015-08-31 11:06 ` [PATCH v6 11/31] xen/arm: ITS: Enable compilation of physical ITS driver vijay.kilari
2015-08-31 11:06 ` [PATCH v6 12/31] xen/arm: Move vgic locking inside get_irq_priority callback vijay.kilari
2015-08-31 16:34   ` Julien Grall
2015-08-31 11:06 ` [PATCH v6 13/31] xen/arm: ITS: implement hw_irq_controller for LPIs vijay.kilari
2015-08-31 17:53   ` Julien Grall
2015-08-31 11:06 ` [PATCH v6 14/31] xen/arm: ITS: Initialize physical ITS and export lpi support vijay.kilari
2015-08-31 18:35   ` Julien Grall
2015-08-31 11:06 ` [PATCH v6 15/31] xen/arm: ITS: Add virtual ITS driver vijay.kilari
2015-09-02 17:20   ` Julien Grall
2015-08-31 11:06 ` [PATCH v6 16/31] xen/arm: ITS: Add virtual ITS commands support vijay.kilari
2015-09-03 15:07   ` Julien Grall [this message]
2015-08-31 11:06 ` [PATCH v6 17/31] xen/arm: ITS: Store LPIs allocated and IRQ ID bits per domain vijay.kilari
2015-09-03 16:25   ` Julien Grall
2015-09-07  6:59     ` Vijay Kilari
2015-09-07 10:56       ` Julien Grall
2015-08-31 11:06 ` [PATCH v6 18/31] xen/arm: ITS: Enable virtual ITS driver vijay.kilari
2015-08-31 11:06 ` [PATCH v6 19/31] xen/arm: ITS: Export ITS info to Virtual ITS vijay.kilari
2015-09-03 16:48   ` Julien Grall
2015-08-31 11:06 ` [PATCH v6 20/31] xen/arm: ITS: Introduce helper to get number of event IDs vijay.kilari
2015-09-03 17:51   ` Julien Grall
2015-08-31 11:06 ` [PATCH v6 21/31] xen/arm: ITS: Add GITS registers emulation vijay.kilari
2015-09-07 13:14   ` Julien Grall
2015-08-31 11:06 ` [PATCH v6 22/31] xen/arm: ITS: Add virtual ITS availability check helper vijay.kilari
2015-09-07 13:41   ` Julien Grall
2015-08-31 11:06 ` [PATCH v6 23/31] xen/arm: ITS: Add 32-bit access to GICR_TYPER vijay.kilari
2015-08-31 16:06   ` Julien Grall
2015-08-31 11:06 ` [PATCH v6 24/31] xen/arm: ITS: Add GICR register emulation vijay.kilari
2015-09-07 14:20   ` Julien Grall
2015-09-07 15:26     ` Vijay Kilari
2015-09-09 13:55       ` Ian Campbell
2015-09-09 16:11         ` Julien Grall
2015-08-31 11:06 ` [PATCH v6 25/31] xen/arm: ITS: Allocate irq descriptors for LPIs vijay.kilari
2015-08-31 11:06 ` [PATCH v6 26/31] xen/arm: ITS: Allocate pending_lpi " vijay.kilari
2015-08-31 11:06 ` [PATCH v6 27/31] xen/arm: ITS: Route LPIs vijay.kilari
2015-08-31 11:06 ` [PATCH v6 28/31] xen/arm: ITS: Add domain specific ITS initialization vijay.kilari
2015-08-31 11:06 ` [PATCH v6 29/31] xen/arm: ITS: Map ITS translation space vijay.kilari
2015-08-31 19:07   ` Julien Grall
2015-08-31 11:06 ` [PATCH v6 30/31] xen/arm: ITS: Generate ITS node for Dom0 vijay.kilari
2015-08-31 11:06 ` [PATCH v6 31/31] xen/arm: ITS: Add pci devices in ThunderX vijay.kilari
2015-09-09 15:22   ` Ian Campbell
2015-09-02 15:38 ` [PATCH v6 00/31] Add ITS support Ian Campbell
2015-09-02 15:52   ` Ian Campbell
2015-09-03 16:45 ` Julien Grall
2015-09-09 15:29 ` Ian Campbell
2015-09-14 11:00   ` Vijay Kilari
2015-09-14 11:09     ` Julien Grall
2015-09-14 13:04       ` Vijay Kilari

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=55E8621D.6090601@citrix.com \
    --to=julien.grall@citrix.com \
    --cc=Ian.Campbell@citrix.com \
    --cc=Prasun.Kapoor@caviumnetworks.com \
    --cc=Vijaya.Kumar@caviumnetworks.com \
    --cc=manish.jaggi@caviumnetworks.com \
    --cc=stefano.stabellini@citrix.com \
    --cc=stefano.stabellini@eu.citrix.com \
    --cc=tim@xen.org \
    --cc=vijay.kilari@gmail.com \
    --cc=xen-devel@lists.xen.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).