
This CL introduces multiple SiteInstances per SiteInstanceGroup for subframe data: URLs. This feature is behind the feature flag kSiteInstanceGroupsForDataUrls, which is disabled by default, so there is no behaviour change in this CL. Subframe data: URLs will now have their own SiteInstance that goes in the same SiteInstanceGroup as its initiator. Currently, sandboxed data: subframes are excluded, as they require computing a variation of the initiator SiteInstance, which is out of scope for this CL and will be added in a followup. Because the new data: subframe shares a SiteInstanceGroup with its initiator, the number of processes remains the same as before. Test: Added SiteInstanceGroup browsertests Change-Id: If784b21ceccd440e35c0020053823ae287cf931d Bug: 40269084 Reviewed-on: https://chromium-review.googlesource.com/c/chromium/src/+/4675093 Reviewed-by: Andrey Kosyakov <caseq@chromium.org> Reviewed-by: Elly FJ <ellyjones@chromium.org> Reviewed-by: Charlie Reis <creis@chromium.org> Commit-Queue: Sharon Yang <yangsharon@chromium.org> Cr-Commit-Position: refs/heads/main@{#1382080}
360 lines
14 KiB
C++
360 lines
14 KiB
C++
// Copyright 2012 The Chromium Authors
|
|
// Use of this source code is governed by a BSD-style license that can be
|
|
// found in the LICENSE file.
|
|
|
|
#include "content/browser/browsing_instance.h"
|
|
|
|
#include "base/check_op.h"
|
|
#include "base/command_line.h"
|
|
#include "base/containers/contains.h"
|
|
#include "content/browser/child_process_security_policy_impl.h"
|
|
#include "content/browser/origin_agent_cluster_isolation_state.h"
|
|
#include "content/browser/security/coop/coop_related_group.h"
|
|
#include "content/browser/site_info.h"
|
|
#include "content/browser/site_instance_group.h"
|
|
#include "content/browser/site_instance_impl.h"
|
|
#include "content/public/browser/browser_context.h"
|
|
#include "content/public/browser/browser_or_resource_context.h"
|
|
#include "content/public/browser/content_browser_client.h"
|
|
#include "content/public/browser/site_isolation_policy.h"
|
|
#include "content/public/common/content_features.h"
|
|
#include "content/public/common/content_switches.h"
|
|
#include "content/public/common/url_constants.h"
|
|
|
|
namespace content {
|
|
|
|
// Start the BrowsingInstance ID counter from 1 to avoid a conflict with the
|
|
// invalid BrowsingInstanceId value, which is 0 in its underlying IdType32.
|
|
int BrowsingInstance::next_browsing_instance_id_ = 1;
|
|
|
|
BrowsingInstance::BrowsingInstance(
|
|
BrowserContext* browser_context,
|
|
const WebExposedIsolationInfo& web_exposed_isolation_info,
|
|
bool is_guest,
|
|
bool is_fenced,
|
|
bool is_fixed_storage_partition,
|
|
const scoped_refptr<CoopRelatedGroup>& coop_related_group,
|
|
std::optional<url::Origin> common_coop_origin)
|
|
: isolation_context_(
|
|
BrowsingInstanceId::FromUnsafeValue(next_browsing_instance_id_++),
|
|
BrowserOrResourceContext(browser_context),
|
|
is_guest,
|
|
is_fenced,
|
|
OriginAgentClusterIsolationState::CreateForDefaultIsolation(
|
|
browser_context)),
|
|
active_contents_count_(0u),
|
|
default_site_instance_(nullptr),
|
|
web_exposed_isolation_info_(web_exposed_isolation_info),
|
|
coop_related_group_(coop_related_group),
|
|
common_coop_origin_(common_coop_origin),
|
|
is_fixed_storage_partition_(is_fixed_storage_partition) {
|
|
DCHECK(browser_context);
|
|
if (is_guest) {
|
|
CHECK(is_fixed_storage_partition);
|
|
}
|
|
|
|
// If we get passed an empty group, build a new one. This is the common case.
|
|
if (!coop_related_group_) {
|
|
coop_related_group_ =
|
|
base::WrapRefCounted<CoopRelatedGroup>(new CoopRelatedGroup(
|
|
browser_context, is_guest, is_fenced, is_fixed_storage_partition_));
|
|
}
|
|
DCHECK(coop_related_group_);
|
|
|
|
coop_related_group_->RegisterBrowsingInstance(this);
|
|
}
|
|
|
|
BrowserContext* BrowsingInstance::GetBrowserContext() const {
|
|
return isolation_context_.browser_or_resource_context().ToBrowserContext();
|
|
}
|
|
|
|
bool BrowsingInstance::HasSiteInstance(const SiteInfo& site_info) {
|
|
return base::Contains(site_instance_map_, site_info);
|
|
}
|
|
|
|
scoped_refptr<SiteInstanceImpl> BrowsingInstance::GetSiteInstanceForURL(
|
|
const UrlInfo& url_info,
|
|
bool allow_default_instance) {
|
|
return GetSiteInstanceForURL(url_info, /*creation_group=*/nullptr,
|
|
allow_default_instance);
|
|
}
|
|
|
|
scoped_refptr<SiteInstanceImpl> BrowsingInstance::GetSiteInstanceForURL(
|
|
const UrlInfo& url_info,
|
|
SiteInstanceGroup* creation_group,
|
|
bool allow_default_instance) {
|
|
scoped_refptr<SiteInstanceImpl> site_instance =
|
|
GetSiteInstanceForURLHelper(url_info, allow_default_instance);
|
|
|
|
if (site_instance)
|
|
return site_instance;
|
|
|
|
// No current SiteInstance for this site, so let's create one.
|
|
scoped_refptr<SiteInstanceImpl> instance = new SiteInstanceImpl(this);
|
|
|
|
// Set the site of this new SiteInstance, which will register it with us.
|
|
// Some URLs should leave the SiteInstance's site unassigned, though if
|
|
// `instance` is for a guest, we should always set the site to ensure that it
|
|
// carries guest information contained within SiteInfo.
|
|
if (SiteInstanceImpl::ShouldAssignSiteForUrlInfo(url_info) ||
|
|
isolation_context_.is_guest()) {
|
|
instance->SetSite(url_info);
|
|
}
|
|
|
|
// Add the new SiteInstance to `group`, if it exists.
|
|
if (creation_group) {
|
|
creation_group->AddSiteInstance(instance.get());
|
|
instance->SetSiteInstanceGroup(creation_group);
|
|
}
|
|
|
|
return instance;
|
|
}
|
|
|
|
SiteInfo BrowsingInstance::GetSiteInfoForURL(const UrlInfo& url_info,
|
|
bool allow_default_instance) {
|
|
scoped_refptr<SiteInstanceImpl> site_instance =
|
|
GetSiteInstanceForURLHelper(url_info, allow_default_instance);
|
|
|
|
if (site_instance)
|
|
return site_instance->GetSiteInfo();
|
|
|
|
return ComputeSiteInfoForURL(url_info);
|
|
}
|
|
|
|
scoped_refptr<SiteInstanceImpl> BrowsingInstance::GetSiteInstanceForSiteInfo(
|
|
const SiteInfo& site_info) {
|
|
auto i = site_instance_map_.find(site_info);
|
|
if (i != site_instance_map_.end())
|
|
return i->second.get();
|
|
|
|
scoped_refptr<SiteInstanceImpl> instance = new SiteInstanceImpl(this);
|
|
instance->SetSite(site_info);
|
|
return instance;
|
|
}
|
|
|
|
scoped_refptr<SiteInstanceImpl>
|
|
BrowsingInstance::GetMaybeGroupRelatedSiteInstanceForURL(
|
|
const UrlInfo& url_info,
|
|
SiteInstanceGroup* creation_group) {
|
|
CHECK(creation_group);
|
|
scoped_refptr<SiteInstanceImpl> instance = GetSiteInstanceForURL(
|
|
url_info, creation_group, /*allow_default_instance=*/false);
|
|
return instance;
|
|
}
|
|
|
|
scoped_refptr<SiteInstanceImpl>
|
|
BrowsingInstance::GetCoopRelatedSiteInstanceForURL(
|
|
const UrlInfo& url_info,
|
|
bool allow_default_instance) {
|
|
return coop_related_group_->GetCoopRelatedSiteInstanceForURL(
|
|
url_info, allow_default_instance);
|
|
}
|
|
|
|
scoped_refptr<SiteInstanceImpl> BrowsingInstance::GetSiteInstanceForURLHelper(
|
|
const UrlInfo& url_info,
|
|
bool allow_default_instance) {
|
|
const SiteInfo site_info = ComputeSiteInfoForURL(url_info);
|
|
auto i = site_instance_map_.find(site_info);
|
|
if (i != site_instance_map_.end())
|
|
return i->second.get();
|
|
|
|
// Check to see if we can use the default SiteInstance for sites that don't
|
|
// need to be isolated in their own process.
|
|
if (allow_default_instance &&
|
|
SiteInstanceImpl::CanBePlacedInDefaultSiteInstance(
|
|
isolation_context_, url_info.url, site_info)) {
|
|
scoped_refptr<SiteInstanceImpl> site_instance =
|
|
default_site_instance_.get();
|
|
if (!site_instance) {
|
|
site_instance = new SiteInstanceImpl(this);
|
|
|
|
// Note: |default_site_instance_| will get set inside this call
|
|
// via RegisterSiteInstance().
|
|
site_instance->SetSiteInfoToDefault(site_info.storage_partition_config());
|
|
DCHECK_EQ(default_site_instance_, site_instance.get());
|
|
}
|
|
|
|
// Add |site_info| to the set so we can keep track of all the sites the
|
|
// the default SiteInstance has been returned for.
|
|
site_instance->AddSiteInfoToDefault(site_info);
|
|
return site_instance;
|
|
}
|
|
|
|
return nullptr;
|
|
}
|
|
|
|
void BrowsingInstance::RegisterSiteInstance(SiteInstanceImpl* site_instance) {
|
|
DCHECK(site_instance->browsing_instance_.get() == this);
|
|
DCHECK(site_instance->HasSite());
|
|
|
|
// Verify that the SiteInstance's StoragePartitionConfig matches this
|
|
// BrowsingInstance's StoragePartitionConfig if it already has one.
|
|
const StoragePartitionConfig& storage_partition_config =
|
|
site_instance->GetSiteInfo().storage_partition_config();
|
|
if (storage_partition_config_.has_value()) {
|
|
// We should only use a single StoragePartition within a BrowsingInstance.
|
|
// If we're attempting to use multiple, something has gone wrong with the
|
|
// logic at upper layers. Similarly, whether this StoragePartition is for
|
|
// a guest should remain constant over a BrowsingInstance's lifetime.
|
|
CHECK_EQ(storage_partition_config_.value(), storage_partition_config);
|
|
CHECK_EQ(isolation_context_.is_guest(), site_instance->IsGuest());
|
|
} else {
|
|
storage_partition_config_ = storage_partition_config;
|
|
}
|
|
|
|
// Explicitly prevent the default SiteInstance from being added since
|
|
// the map is only supposed to contain instances that map to a single site.
|
|
if (site_instance->IsDefaultSiteInstance()) {
|
|
CHECK(!default_site_instance_);
|
|
default_site_instance_ = site_instance;
|
|
return;
|
|
}
|
|
|
|
const SiteInfo& site_info = site_instance->GetSiteInfo();
|
|
|
|
// Only register if we don't have a SiteInstance for this site already.
|
|
// It's possible to have two SiteInstances point to the same site if two
|
|
// tabs are navigated there at the same time. (We don't call SetSite or
|
|
// register them until DidNavigate.) If there is a previously existing
|
|
// SiteInstance for this site, we just won't register the new one.
|
|
auto i = site_instance_map_.find(site_info);
|
|
if (i == site_instance_map_.end()) {
|
|
// Not previously registered, so register it.
|
|
site_instance_map_[site_info] = site_instance;
|
|
}
|
|
}
|
|
|
|
void BrowsingInstance::UnregisterSiteInstance(SiteInstanceImpl* site_instance) {
|
|
DCHECK(site_instance->browsing_instance_.get() == this);
|
|
DCHECK(site_instance->HasSite());
|
|
|
|
if (site_instance == default_site_instance_) {
|
|
// The last reference to the default SiteInstance is being destroyed.
|
|
default_site_instance_ = nullptr;
|
|
}
|
|
|
|
// Only unregister the SiteInstance if it is the same one that is registered
|
|
// for the site. (It might have been an unregistered SiteInstance. See the
|
|
// comments in RegisterSiteInstance.)
|
|
auto i = site_instance_map_.find(site_instance->GetSiteInfo());
|
|
if (i != site_instance_map_.end() && i->second == site_instance) {
|
|
// Matches, so erase it.
|
|
site_instance_map_.erase(i);
|
|
}
|
|
}
|
|
|
|
// static
|
|
BrowsingInstanceId BrowsingInstance::NextBrowsingInstanceId() {
|
|
return BrowsingInstanceId::FromUnsafeValue(next_browsing_instance_id_);
|
|
}
|
|
|
|
BrowsingInstance::~BrowsingInstance() {
|
|
// We should only be deleted when all of the SiteInstances that refer to
|
|
// us are gone.
|
|
DCHECK(site_instance_map_.empty());
|
|
DCHECK_EQ(0u, active_contents_count_);
|
|
DCHECK(!default_site_instance_);
|
|
|
|
// Remove any origin isolation opt-ins related to this instance.
|
|
ChildProcessSecurityPolicyImpl* policy =
|
|
ChildProcessSecurityPolicyImpl::GetInstance();
|
|
policy->RemoveOptInIsolatedOriginsForBrowsingInstance(
|
|
isolation_context_.browsing_instance_id());
|
|
|
|
coop_related_group_->UnregisterBrowsingInstance(this);
|
|
}
|
|
|
|
SiteInfo BrowsingInstance::ComputeSiteInfoForURL(
|
|
const UrlInfo& url_info) const {
|
|
// If a StoragePartitionConfig is specified in both `url_info` and this
|
|
// BrowsingInstance, make sure they match.
|
|
if (url_info.storage_partition_config.has_value() &&
|
|
storage_partition_config_.has_value()) {
|
|
CHECK_EQ(storage_partition_config_.value(),
|
|
url_info.storage_partition_config.value());
|
|
}
|
|
// If no StoragePartitionConfig was set in `url_info`, create a new UrlInfo
|
|
// that inherit's this BrowsingInstance's StoragePartitionConfig.
|
|
UrlInfo url_info_with_partition =
|
|
url_info.storage_partition_config.has_value()
|
|
? url_info
|
|
: UrlInfo(UrlInfoInit(url_info).WithStoragePartitionConfig(
|
|
storage_partition_config_));
|
|
|
|
// The WebExposedIsolationInfos must be compatible for this function to make
|
|
// sense.
|
|
DCHECK(WebExposedIsolationInfo::AreCompatible(
|
|
url_info.web_exposed_isolation_info, web_exposed_isolation_info_));
|
|
|
|
// If the passed in UrlInfo has a null WebExposedIsolationInfo, meaning that
|
|
// it is compatible with any isolation state, we reuse the isolation state of
|
|
// the BrowsingInstance.
|
|
url_info_with_partition.web_exposed_isolation_info =
|
|
url_info.web_exposed_isolation_info.value_or(web_exposed_isolation_info_);
|
|
return SiteInfo::Create(isolation_context_, url_info_with_partition);
|
|
}
|
|
|
|
int BrowsingInstance::EstimateOriginAgentClusterOverhead() {
|
|
DCHECK(SiteIsolationPolicy::IsProcessIsolationForOriginAgentClusterEnabled());
|
|
|
|
std::set<SiteInfo> site_info_set;
|
|
std::set<SiteInfo> site_info_set_no_oac;
|
|
|
|
// The following computes an estimate of how many additional processes have
|
|
// been created to deal with OriginAgentCluster (OAC) headers. When OAC
|
|
// headers forces an additional process, that corresponds to the SiteInfo's
|
|
// is_origin_keyed_ flag being set. To compute the estimate, we use the set of
|
|
// unique SiteInstances (each represented by a unique SiteInfo) in each
|
|
// BrowsingInstance as a proxy for the set of different RenderProcesses. We
|
|
// start with the total count of SiteInfos, then we create a new set of
|
|
// SiteInfos created by resetting the is_origin_keyed_ flag on each of the
|
|
// SiteInfos (along with any corresponding adjustments to the site_url_ and
|
|
// process_lock_url_ to reflect the possible conversion from origin to site).
|
|
// The assumption here is that SiteInfos that forced a new process due to OAC
|
|
// may no longer be unique once these values are reset, and as such the new
|
|
// set will have less elements than the original set, with the difference
|
|
// being the count of extra SiteInstances due to OAC. There are cases where
|
|
// ignoring the OAC header would still result in an extra process, e.g. when
|
|
// the SiteInfo's origin appears in the command-line origin isolation list.
|
|
//
|
|
// The estimate is computed using several simplifying assumptions:
|
|
// 1) We only consider HTTPS SiteInfos to compute the additional SiteInfos.
|
|
// This assumption should generally be valid, since we don't apply
|
|
// is_origin_keyed_ to non-HTTPS schemes.
|
|
// 2) We assume that SiteInfos from multiple BrowsingInstances aren't
|
|
// coalesced into a single RenderProcess. While this isn't true in general,
|
|
// it is difficult in practice to account for, so we don't try to.
|
|
for (auto& entry : site_instance_map_) {
|
|
const SiteInfo& site_info = entry.first;
|
|
GURL process_lock_url = site_info.process_lock_url();
|
|
if (!process_lock_url.SchemeIs(url::kHttpsScheme))
|
|
continue;
|
|
|
|
site_info_set.insert(site_info);
|
|
site_info_set_no_oac.insert(
|
|
site_info.GetNonOriginKeyedEquivalentForMetrics(isolation_context_));
|
|
}
|
|
DCHECK_GE(site_info_set.size(), site_info_set_no_oac.size());
|
|
int result = site_info_set.size() - site_info_set_no_oac.size();
|
|
return result;
|
|
}
|
|
|
|
size_t BrowsingInstance::GetCoopRelatedGroupActiveContentsCount() {
|
|
return coop_related_group_->active_contents_count();
|
|
}
|
|
|
|
void BrowsingInstance::IncrementActiveContentsCount() {
|
|
active_contents_count_++;
|
|
|
|
coop_related_group_->increment_active_contents_count();
|
|
}
|
|
|
|
void BrowsingInstance::DecrementActiveContentsCount() {
|
|
DCHECK_LT(0u, active_contents_count_);
|
|
active_contents_count_--;
|
|
|
|
coop_related_group_->decrement_active_contents_count();
|
|
}
|
|
|
|
} // namespace content
|