public inbox for kvm@vger.kernel.org
 help / color / mirror / Atom feed
From: Farhan Ali <alifm@linux.ibm.com>
To: Eric Farman <farman@linux.ibm.com>, Cornelia Huck <cohuck@redhat.com>
Cc: Halil Pasic <pasic@linux.ibm.com>,
	linux-s390@vger.kernel.org, kvm@vger.kernel.org
Subject: Re: [RFC PATCH v1 1/5] vfio-ccw: Move guest_cp storage into common struct
Date: Wed, 19 Jun 2019 17:12:44 -0400	[thread overview]
Message-ID: <79904d5c-c6f7-dae0-8f11-d3db37337eea@linux.ibm.com> (raw)
In-Reply-To: <50cf003b-017d-f475-238d-83aef2df7200@linux.ibm.com>



On 06/19/2019 04:53 PM, Eric Farman wrote:
> 
> 
> On 6/19/19 4:13 PM, Farhan Ali wrote:
>>
>>
>> On 06/18/2019 04:23 PM, Eric Farman wrote:
>>> Rather than allocating/freeing a piece of memory every time
>>> we try to figure out how long a CCW chain is, let's use a piece
>>> of memory allocated for each device.
>>>
>>> The io_mutex added with commit 4f76617378ee9 ("vfio-ccw: protect
>>> the I/O region") is held for the duration of the VFIO_CCW_EVENT_IO_REQ
>>> event that accesses/uses this space, so there should be no race
>>> concerns with another CPU attempting an (unexpected) SSCH for the
>>> same device.
>>>
>>> Suggested-by: Cornelia Huck <cohuck@redhat.com>
>>> Signed-off-by: Eric Farman <farman@linux.ibm.com>
>>> ---
>>> Conny, your suggestion [1] did not go unnoticed.  :)
>>>
>>> [1] https://patchwork.kernel.org/comment/22312659/
>>> ---
>>>    drivers/s390/cio/vfio_ccw_cp.c  | 23 ++++-------------------
>>>    drivers/s390/cio/vfio_ccw_cp.h  |  7 +++++++
>>>    drivers/s390/cio/vfio_ccw_drv.c |  7 +++++++
>>>    3 files changed, 18 insertions(+), 19 deletions(-)
>>>
>>> diff --git a/drivers/s390/cio/vfio_ccw_cp.c
>>> b/drivers/s390/cio/vfio_ccw_cp.c
>>> index 90d86e1354c1..f358502376be 100644
>>> --- a/drivers/s390/cio/vfio_ccw_cp.c
>>> +++ b/drivers/s390/cio/vfio_ccw_cp.c
>>> @@ -16,12 +16,6 @@
>>>      #include "vfio_ccw_cp.h"
>>>    -/*
>>> - * Max length for ccw chain.
>>> - * XXX: Limit to 256, need to check more?
>>> - */
>>> -#define CCWCHAIN_LEN_MAX    256
>>> -
>>>    struct pfn_array {
>>>        /* Starting guest physical I/O address. */
>>>        unsigned long        pa_iova;
>>> @@ -386,7 +380,7 @@ static void ccwchain_cda_free(struct ccwchain
>>> *chain, int idx)
>>>     */
>>>    static int ccwchain_calc_length(u64 iova, struct channel_program *cp)
>>>    {
>>> -    struct ccw1 *ccw, *p;
>>> +    struct ccw1 *ccw = cp->guest_cp;
>>>        int cnt;
>>>          /*
>>> @@ -394,15 +388,9 @@ static int ccwchain_calc_length(u64 iova, struct
>>> channel_program *cp)
>>>         * Currently the chain length is limited to CCWCHAIN_LEN_MAX (256).
>>>         * So copying 2K is enough (safe).
>>>         */
>>> -    p = ccw = kcalloc(CCWCHAIN_LEN_MAX, sizeof(*ccw), GFP_KERNEL);
>>> -    if (!ccw)
>>> -        return -ENOMEM;
>>> -
>>>        cnt = copy_ccw_from_iova(cp, ccw, iova, CCWCHAIN_LEN_MAX);
>>
>> Just a minor concern, should we clear out cp->guest_cp memory before we
>> do the copying? Given that the ccwchain_calc_length will also call be
>> called during tic handling, it's possible there might be some garbage
>> data in guest_cp, no?
> 
> Yeah, they'll be garbage there, but I'm not sure it's a problem.  By the
> time we get here again (ccwchain_loop_tic() -> ccwchain_handle_ccw()),
> we'll have saved the relevant CCWs for the first segment.  And the
> second time through we'll be copying a fresh 2K from the target of the
> TIC to cp->guest_cp, overwriting all that stale data with new CCWs (and
> new garbage data).
> 

Yes, you are right. Please disregard my concern :)

>>
>>
>>> -    if (cnt) {
>>> -        kfree(ccw);
>>> +    if (cnt)
>>>            return cnt;
>>> -    }
>>>          cnt = 0;
>>>        do {
>>> @@ -413,10 +401,8 @@ static int ccwchain_calc_length(u64 iova, struct
>>> channel_program *cp)
>>>             * orb specified one of the unsupported formats, we defer
>>>             * checking for IDAWs in unsupported formats to here.
>>>             */
>>> -        if ((!cp->orb.cmd.c64 || cp->orb.cmd.i2k) && ccw_is_idal(ccw)) {
>>> -            kfree(p);
>>> +        if ((!cp->orb.cmd.c64 || cp->orb.cmd.i2k) && ccw_is_idal(ccw))
>>>                return -EOPNOTSUPP;
>>> -        }
>>>              /*
>>>             * We want to keep counting if the current CCW has the
>>> @@ -435,7 +421,6 @@ static int ccwchain_calc_length(u64 iova, struct
>>> channel_program *cp)
>>>        if (cnt == CCWCHAIN_LEN_MAX + 1)
>>>            cnt = -EINVAL;
>>>    -    kfree(p);
>>>        return cnt;
>>>    }
>>>    @@ -461,7 +446,7 @@ static int ccwchain_handle_ccw(u32 cda, struct
>>> channel_program *cp)
>>>        struct ccwchain *chain;
>>>        int len, ret;
>>>    -    /* Get chain length. */
>>> +    /* Copy the chain from cda to cp, and count the CCWs in it */
>>>        len = ccwchain_calc_length(cda, cp);
>>>        if (len < 0)
>>>            return len;
>>> diff --git a/drivers/s390/cio/vfio_ccw_cp.h
>>> b/drivers/s390/cio/vfio_ccw_cp.h
>>> index 3c20cd208da5..7cdc38049033 100644
>>> --- a/drivers/s390/cio/vfio_ccw_cp.h
>>> +++ b/drivers/s390/cio/vfio_ccw_cp.h
>>> @@ -16,6 +16,12 @@
>>>      #include "orb.h"
>>>    +/*
>>> + * Max length for ccw chain.
>>> + * XXX: Limit to 256, need to check more?
>>> + */
>>> +#define CCWCHAIN_LEN_MAX    256
>>> +
>>>    /**
>>>     * struct channel_program - manage information for channel program
>>>     * @ccwchain_list: list head of ccwchains
>>> @@ -32,6 +38,7 @@ struct channel_program {
>>>        union orb orb;
>>>        struct device *mdev;
>>>        bool initialized;
>>> +    struct ccw1 *guest_cp;
>>>    };
>>>      extern int cp_init(struct channel_program *cp, struct device *mdev,
>>> diff --git a/drivers/s390/cio/vfio_ccw_drv.c
>>> b/drivers/s390/cio/vfio_ccw_drv.c
>>> index 66a66ac1f3d1..34a9a5e3fd36 100644
>>> --- a/drivers/s390/cio/vfio_ccw_drv.c
>>> +++ b/drivers/s390/cio/vfio_ccw_drv.c
>>> @@ -129,6 +129,11 @@ static int vfio_ccw_sch_probe(struct subchannel
>>> *sch)
>>>        if (!private)
>>>            return -ENOMEM;
>>>    +    private->cp.guest_cp = kcalloc(CCWCHAIN_LEN_MAX, sizeof(struct
>>> ccw1),
>>> +                       GFP_KERNEL);
>>> +    if (!private->cp.guest_cp)
>>> +        goto out_free;
>>> +
>>>        private->io_region = kmem_cache_zalloc(vfio_ccw_io_region,
>>>                               GFP_KERNEL | GFP_DMA);
>>>        if (!private->io_region)
>>> @@ -169,6 +174,7 @@ static int vfio_ccw_sch_probe(struct subchannel *sch)
>>>            kmem_cache_free(vfio_ccw_cmd_region, private->cmd_region);
>>>        if (private->io_region)
>>>            kmem_cache_free(vfio_ccw_io_region, private->io_region);
>>> +    kfree(private->cp.guest_cp);
>>>        kfree(private);
>>>        return ret;
>>>    }
>>> @@ -185,6 +191,7 @@ static int vfio_ccw_sch_remove(struct subchannel
>>> *sch)
>>>          kmem_cache_free(vfio_ccw_cmd_region, private->cmd_region);
>>>        kmem_cache_free(vfio_ccw_io_region, private->io_region);
>>> +    kfree(private->cp.guest_cp);
>>>        kfree(private);
>>>          return 0;
>>>
> 

  reply	other threads:[~2019-06-19 21:14 UTC|newest]

Thread overview: 19+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2019-06-18 20:23 [RFC PATCH v1 0/5] s390: more vfio-ccw code rework Eric Farman
2019-06-18 20:23 ` [RFC PATCH v1 1/5] vfio-ccw: Move guest_cp storage into common struct Eric Farman
2019-06-19  8:14   ` Cornelia Huck
2019-06-19 20:13   ` Farhan Ali
2019-06-19 20:53     ` Eric Farman
2019-06-19 21:12       ` Farhan Ali [this message]
2019-06-18 20:23 ` [RFC PATCH v1 2/5] vfio-ccw: Skip second copy of guest cp to host Eric Farman
2019-06-19  8:17   ` Cornelia Huck
2019-06-18 20:23 ` [RFC PATCH v1 3/5] vfio-ccw: Copy CCW data outside length calculation Eric Farman
2019-06-19  8:18   ` Cornelia Huck
2019-06-18 20:23 ` [RFC PATCH v1 4/5] vfio-ccw: Factor out the ccw0-to-ccw1 transition Eric Farman
2019-06-19  8:22   ` Cornelia Huck
2019-06-18 20:23 ` [RFC PATCH v1 5/5] vfio-ccw: Remove copy_ccw_from_iova() Eric Farman
2019-06-19  8:23   ` Cornelia Huck
2019-06-19 21:13   ` Farhan Ali
2019-06-19  8:25 ` [RFC PATCH v1 0/5] s390: more vfio-ccw code rework Cornelia Huck
2019-06-19 11:11   ` Eric Farman
2019-06-19 21:15 ` Farhan Ali
2019-06-21 12:25 ` Cornelia Huck

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=79904d5c-c6f7-dae0-8f11-d3db37337eea@linux.ibm.com \
    --to=alifm@linux.ibm.com \
    --cc=cohuck@redhat.com \
    --cc=farman@linux.ibm.com \
    --cc=kvm@vger.kernel.org \
    --cc=linux-s390@vger.kernel.org \
    --cc=pasic@linux.ibm.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