qemu-devel.nongnu.org archive mirror
 help / color / mirror / Atom feed
From: "Philippe Mathieu-Daudé" <philmd@linaro.org>
To: "Michael S. Tsirkin" <mst@redhat.com>,
	CLEMENT MATHIEU--DRIF <clement.mathieu--drif@eviden.com>
Cc: "qemu-devel@nongnu.org" <qemu-devel@nongnu.org>,
	"jasowang@redhat.com" <jasowang@redhat.com>,
	"zhenzhong.duan@intel.com" <zhenzhong.duan@intel.com>,
	"kevin.tian@intel.com" <kevin.tian@intel.com>,
	"yi.l.liu@intel.com" <yi.l.liu@intel.com>,
	"peterx@redhat.com" <peterx@redhat.com>
Subject: Re: [PATCH] intel_iommu: Take the bql before registering a new address space
Date: Tue, 15 Apr 2025 10:03:59 +0200	[thread overview]
Message-ID: <b0a898ef-93cf-4ce9-8737-cac94ed4f731@linaro.org> (raw)
In-Reply-To: <20250415034112-mutt-send-email-mst@kernel.org>

On 15/4/25 09:42, Michael S. Tsirkin wrote:
> On Tue, Apr 15, 2025 at 07:28:34AM +0000, CLEMENT MATHIEU--DRIF wrote:
>>
>>
>> On 15/04/2025 8:53 am, Philippe Mathieu-Daudé wrote:
>>> Caution: External email. Do not open attachments or click links, unless
>>> this email comes from a known sender and you know the content is safe.
>>>
>>>
>>> On 15/4/25 08:18, CLEMENT MATHIEU--DRIF wrote:
>>>> Address space creation might end up being called without holding the
>>>> bql as it is exposed through the IOMMU ops.
>>>>
>>>> Signed-off-by: Clement Mathieu--Drif <clement.mathieu--drif@eviden.com>
>>>> ---
>>>>    hw/i386/intel_iommu.c | 10 ++++++++++
>>>>    1 file changed, 10 insertions(+)
>>>>
>>>> diff --git a/hw/i386/intel_iommu.c b/hw/i386/intel_iommu.c
>>>> index dffd7ee885..fea2220013 100644
>>>> --- a/hw/i386/intel_iommu.c
>>>> +++ b/hw/i386/intel_iommu.c
>>>> @@ -4216,6 +4216,7 @@ VTDAddressSpace *vtd_find_add_as(IntelIOMMUState
>>>> *s, PCIBus *bus,
>>>>        vtd_dev_as = g_hash_table_lookup(s->vtd_address_spaces, &key);
>>>>        if (!vtd_dev_as) {
>>>>            struct vtd_as_key *new_key = g_malloc(sizeof(*new_key));
>>>> +        bool take_bql = !bql_locked();
>>>>
>>>>            new_key->bus = bus;
>>>>            new_key->devfn = devfn;
>>>> @@ -4238,6 +4239,11 @@ VTDAddressSpace
>>>> *vtd_find_add_as(IntelIOMMUState *s, PCIBus *bus,
>>>>            vtd_dev_as->context_cache_entry.context_cache_gen = 0;
>>>>            vtd_dev_as->iova_tree = iova_tree_new();
>>>>
>>>> +        /* Some functions in this branch require the bql, make sure
>>>> we own it */
>>>> +        if (take_bql) {
>>>> +            bql_lock();
>>>> +        }
>>>> +
>>>>            memory_region_init(&vtd_dev_as->root, OBJECT(s), name,
>>>> UINT64_MAX);
>>>>            address_space_init(&vtd_dev_as->as, &vtd_dev_as->root, "vtd-
>>>> root");
>>>>
>>>> @@ -4305,6 +4311,10 @@ VTDAddressSpace
>>>> *vtd_find_add_as(IntelIOMMUState *s, PCIBus *bus,
>>>>
>>>>            vtd_switch_address_space(vtd_dev_as);
>>>
>>> Would it help clarifying to propagate this argument down?
>>> vtd_switch_address_space(VTDAddressSpace *as, bool need_lock);
>>
>> Hi phil, vtd_switch_address_space already does the same kind of check
>>
>>>
>>>>
>>>> +        if (take_bql) {
>>>> +            bql_unlock();
>>>> +        }
>>>> +
>>>>            g_hash_table_insert(s->vtd_address_spaces, new_key,
>>>> vtd_dev_as);
>>>>        }
>>>>        return vtd_dev_as;
>>>
> 
> 
> As an apropos, I think any caller of bql_lock really should call
> bql_lock_impl so we know who took BQL. Or just use BQL_LOCK_GUARD.
> But, that's an unrelated cleanup.
> 

Yeah unrelated cleanup. Although I don't understand why these
code paths don't use memory_region_transaction_begin/commit and
have to access BQL.

Some untested idea to investigate:
-- >8 --
diff --git a/hw/i386/intel_iommu.c b/hw/i386/intel_iommu.c
index fea22200135..b2a809cb3db 100644
--- a/hw/i386/intel_iommu.c
+++ b/hw/i386/intel_iommu.c
@@ -1726,3 +1726,6 @@ static bool vtd_as_pt_enabled(VTDAddressSpace *as)

-/* Return whether the device is using IOMMU translation. */
+/*
+ * Return whether the device is using IOMMU translation.
+ * Called with BQL locked.
+ */
  static bool vtd_switch_address_space(VTDAddressSpace *as)
@@ -1730,4 +1733,2 @@ static bool 
vtd_switch_address_space(VTDAddressSpace *as)
      bool use_iommu, pt;
-    /* Whether we need to take the BQL on our own */
-    bool take_bql = !bql_locked();

@@ -1743,10 +1744,3 @@ static bool 
vtd_switch_address_space(VTDAddressSpace *as)

-    /*
-     * It's possible that we reach here without BQL, e.g., when called
-     * from vtd_pt_enable_fast_path(). However the memory APIs need
-     * it. We'd better make sure we have had it already, or, take it.
-     */
-    if (take_bql) {
-        bql_lock();
-    }
+    memory_region_transaction_begin();

@@ -1803,5 +1797,3 @@ static bool 
vtd_switch_address_space(VTDAddressSpace *as)

-    if (take_bql) {
-        bql_unlock();
-    }
+    memory_region_transaction_commit();

@@ -1905,2 +1897,4 @@ static void 
vtd_pt_enable_fast_path(IntelIOMMUState *s, uint16_t source_id)

+    BQL_LOCK_GUARD();
+
      if (vtd_switch_address_space(vtd_as) == false) {
@@ -4241,6 +4235,3 @@ VTDAddressSpace *vtd_find_add_as(IntelIOMMUState 
*s, PCIBus *bus,

-        /* Some functions in this branch require the bql, make sure we 
own it */
-        if (take_bql) {
-            bql_lock();
-        }
+        memory_region_transaction_begin();

@@ -4313,5 +4304,3 @@ VTDAddressSpace *vtd_find_add_as(IntelIOMMUState 
*s, PCIBus *bus,

-        if (take_bql) {
-            bql_unlock();
-        }
+        memory_region_transaction_commit();

---



  reply	other threads:[~2025-04-15  8:04 UTC|newest]

Thread overview: 23+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2025-04-15  6:18 [PATCH] intel_iommu: Take the bql before registering a new address space CLEMENT MATHIEU--DRIF
2025-04-15  6:53 ` Philippe Mathieu-Daudé
2025-04-15  7:28   ` CLEMENT MATHIEU--DRIF
2025-04-15  7:42     ` Michael S. Tsirkin
2025-04-15  8:03       ` Philippe Mathieu-Daudé [this message]
2025-04-15  8:36         ` Yi Liu
2025-04-15  6:55 ` Michael S. Tsirkin
2025-04-15  7:11 ` Michael S. Tsirkin
2025-04-15 12:33   ` Stefan Hajnoczi
2025-04-15 13:24     ` Paolo Bonzini
2025-04-15 14:03     ` Michael S. Tsirkin
2025-04-15 14:14     ` CLEMENT MATHIEU--DRIF
2025-04-15 14:31       ` Stefan Hajnoczi
2025-04-15  8:33 ` Yi Liu
2025-04-15  9:30 ` Paolo Bonzini
2025-04-15 11:49   ` Philippe Mathieu-Daudé
2025-04-15 11:52     ` Philippe Mathieu-Daudé
2025-04-15 11:50   ` CLEMENT MATHIEU--DRIF
2025-04-15 11:51     ` Paolo Bonzini
2025-04-15 11:55       ` Philippe Mathieu-Daudé
2025-04-15 12:59         ` Paolo Bonzini
2025-04-15 15:27           ` CLEMENT MATHIEU--DRIF
2025-04-15 13:08 ` Paolo Bonzini

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=b0a898ef-93cf-4ce9-8737-cac94ed4f731@linaro.org \
    --to=philmd@linaro.org \
    --cc=clement.mathieu--drif@eviden.com \
    --cc=jasowang@redhat.com \
    --cc=kevin.tian@intel.com \
    --cc=mst@redhat.com \
    --cc=peterx@redhat.com \
    --cc=qemu-devel@nongnu.org \
    --cc=yi.l.liu@intel.com \
    --cc=zhenzhong.duan@intel.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).