From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S1758078AbcFAMB4 (ORCPT ); Wed, 1 Jun 2016 08:01:56 -0400 Received: from p3plsmtps2ded01.prod.phx3.secureserver.net ([208.109.80.58]:40643 "EHLO p3plsmtps2ded01.prod.phx3.secureserver.net" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1757518AbcFAMBx (ORCPT ); Wed, 1 Jun 2016 08:01:53 -0400 x-originating-ip: 72.167.245.219 From: Dexuan Cui To: gregkh@linuxfoundation.org, linux-kernel@vger.kernel.org, driverdev-devel@linuxdriverproject.org, olaf@aepfle.de, apw@canonical.com, jasowang@redhat.com, kys@microsoft.com, vkuznets@redhat.com Cc: haiyangz@microsoft.com, rolf.neugebauer@docker.com, dave.scott@docker.com, ian.campbell@docker.com Subject: [PATCH v4] Drivers: hv: vmbus: fix the race when querying & updating the percpu list Date: Wed, 1 Jun 2016 06:43:13 -0700 Message-Id: <1464788593-873-1-git-send-email-decui@microsoft.com> X-Mailer: git-send-email 1.7.4.1 X-CMAE-Envelope: MS4wfPAMV3W3sOLZUVikrmMCMan1qx2kDXcw/FMDSyXvalOmKXeiFo8cTigTpczZVdWKMUDde0OOKlqrD6nFvom0s9bu486YVM9deFvePXRw/lpc9eOeqqO2 RTqkIBje/HiNQXRmO3tNWHVVxPriV+PstKfg+PXrdaFrvsrBXoHWrW0gSKwU+cXkwCU7wjsy0Jz686/Z12pL+btSgpLr9FqG3Ndg7eUTKBjSFsqKDJlP19wS mnXwwpKoJDGk33FUpjw0ooajswjTWuPctunfQh95SgUkCLhHrcD4SSqt6zl94u4Bfk17oXvQeNSbuNny3LNyzd7ZzJRDL+llGi8LKsO14EZJ0cwc5VtW9G64 nfuq4OrX9xPl1vOEiQKWqvt8D+pN48vGiaZgtdZSMx7vjD+knwFA1PLU3GzLZ92pPR8SRsMnU/iQ5LrXJxZyIiulenvotEVey5c8b6F1Ny8UnkO8DtG6q4TE 2BNHqd4kiLb+yz8UHtBl3Uq6Mduj6MXp8+ulGrMCR6HZbZ3pqZQt1BLSH3PR2amaPo7lj5FunqVQhvNE/yvOyBbSYtTTBe5ByOG8IQ== Sender: linux-kernel-owner@vger.kernel.org List-ID: X-Mailing-List: linux-kernel@vger.kernel.org There is a rare race when we remove an entry from the global list hv_context.percpu_list[cpu] in hv_process_channel_removal() -> percpu_channel_deq() -> list_del(): at this time, if vmbus_on_event() -> process_chn_event() -> pcpu_relid2channel() is trying to query the list, we can get the kernel fault. Similarly, we also have the issue in the code path: vmbus_process_offer() -> percpu_channel_enq(). We can resolve the issue by disabling the tasklet when updating the list. The patch also moves vmbus_release_relid() to a later place where the channel has been removed from the per-cpu and the global lists. Reported-by: Rolf Neugebauer Signed-off-by: Dexuan Cui --- v2: added tasklet_schedule() after tasklet_enable(). Thanks, Vitaly! v3: I shouldn't have moved percpu_channel_deq() from hv_process_channel_removal() to vmbus_close_internal(): the channel couldn't be removed from the per-cpu list, if the channel state was not CHANNEL_OPENED_STATE. v3 fixed this. v3 also added 2 wrapper functions for the disable/enable stuff. v3 also moved vmbus_release_relid() to a later safe place. v4: fixed a copy-and-paste bug in vmbus_process_offer(): "hv_event_tasklet_disable(channel) and hv_event_tasklet_enable(channel)" are buggy: the 'channel' parameter should be 'newchannel'. Thanks Vitaly for this! You can also get the patch on https://github.com/dcui/linux/commits/decui/hv_sock/net-next/20160601_v11 drivers/hv/channel.c | 6 ++---- drivers/hv/channel_mgmt.c | 32 ++++++++++++++++++++++++++++---- include/linux/hyperv.h | 3 +++ 3 files changed, 33 insertions(+), 8 deletions(-) diff --git a/drivers/hv/channel.c b/drivers/hv/channel.c index 56dd261..ef8e9b5 100644 --- a/drivers/hv/channel.c +++ b/drivers/hv/channel.c @@ -512,7 +512,6 @@ static void reset_channel_cb(void *arg) static int vmbus_close_internal(struct vmbus_channel *channel) { struct vmbus_channel_close_channel *msg; - struct tasklet_struct *tasklet; int ret; /* @@ -524,8 +523,7 @@ static int vmbus_close_internal(struct vmbus_channel *channel) * To resolve the race, we can serialize them by disabling the * tasklet when the latter is running here. */ - tasklet = hv_context.event_dpc[channel->target_cpu]; - tasklet_disable(tasklet); + hv_event_tasklet_disable(channel); /* * In case a device driver's probe() fails (e.g., @@ -591,7 +589,7 @@ static int vmbus_close_internal(struct vmbus_channel *channel) get_order(channel->ringbuffer_pagecount * PAGE_SIZE)); out: - tasklet_enable(tasklet); + hv_event_tasklet_enable(channel); return ret; } diff --git a/drivers/hv/channel_mgmt.c b/drivers/hv/channel_mgmt.c index 38b682ba..fd89c4c 100644 --- a/drivers/hv/channel_mgmt.c +++ b/drivers/hv/channel_mgmt.c @@ -21,6 +21,7 @@ #define pr_fmt(fmt) KBUILD_MODNAME ": " fmt #include +#include #include #include #include @@ -303,16 +304,32 @@ static void vmbus_release_relid(u32 relid) vmbus_post_msg(&msg, sizeof(struct vmbus_channel_relid_released)); } +void hv_event_tasklet_disable(struct vmbus_channel *channel) +{ + struct tasklet_struct *tasklet; + tasklet = hv_context.event_dpc[channel->target_cpu]; + tasklet_disable(tasklet); +} + +void hv_event_tasklet_enable(struct vmbus_channel *channel) +{ + struct tasklet_struct *tasklet; + tasklet = hv_context.event_dpc[channel->target_cpu]; + tasklet_enable(tasklet); + + /* In case there is any pending event */ + tasklet_schedule(tasklet); +} + void hv_process_channel_removal(struct vmbus_channel *channel, u32 relid) { unsigned long flags; struct vmbus_channel *primary_channel; - vmbus_release_relid(relid); - BUG_ON(!channel->rescind); BUG_ON(!mutex_is_locked(&vmbus_connection.channel_mutex)); + hv_event_tasklet_disable(channel); if (channel->target_cpu != get_cpu()) { put_cpu(); smp_call_function_single(channel->target_cpu, @@ -321,6 +338,7 @@ void hv_process_channel_removal(struct vmbus_channel *channel, u32 relid) percpu_channel_deq(channel); put_cpu(); } + hv_event_tasklet_enable(channel); if (channel->primary_channel == NULL) { list_del(&channel->listentry); @@ -341,6 +359,8 @@ void hv_process_channel_removal(struct vmbus_channel *channel, u32 relid) cpumask_clear_cpu(channel->target_cpu, &primary_channel->alloced_cpus_in_node); + vmbus_release_relid(relid); + free_channel(channel); } @@ -409,6 +429,7 @@ static void vmbus_process_offer(struct vmbus_channel *newchannel) init_vp_index(newchannel, dev_type); + hv_event_tasklet_disable(newchannel); if (newchannel->target_cpu != get_cpu()) { put_cpu(); smp_call_function_single(newchannel->target_cpu, @@ -418,6 +439,7 @@ static void vmbus_process_offer(struct vmbus_channel *newchannel) percpu_channel_enq(newchannel); put_cpu(); } + hv_event_tasklet_enable(newchannel); /* * This state is used to indicate a successful open @@ -463,12 +485,11 @@ static void vmbus_process_offer(struct vmbus_channel *newchannel) return; err_deq_chan: - vmbus_release_relid(newchannel->offermsg.child_relid); - mutex_lock(&vmbus_connection.channel_mutex); list_del(&newchannel->listentry); mutex_unlock(&vmbus_connection.channel_mutex); + hv_event_tasklet_disable(newchannel); if (newchannel->target_cpu != get_cpu()) { put_cpu(); smp_call_function_single(newchannel->target_cpu, @@ -477,6 +498,9 @@ err_deq_chan: percpu_channel_deq(newchannel); put_cpu(); } + hv_event_tasklet_enable(newchannel); + + vmbus_release_relid(newchannel->offermsg.child_relid); err_free_chan: free_channel(newchannel); diff --git a/include/linux/hyperv.h b/include/linux/hyperv.h index 7be7237..8070cda 100644 --- a/include/linux/hyperv.h +++ b/include/linux/hyperv.h @@ -1328,6 +1328,9 @@ extern bool vmbus_prep_negotiate_resp(struct icmsg_hdr *, struct icmsg_negotiate *, u8 *, int, int); +void hv_event_tasklet_disable(struct vmbus_channel *channel); +void hv_event_tasklet_enable(struct vmbus_channel *channel); + void hv_process_channel_removal(struct vmbus_channel *channel, u32 relid); /* -- 2.1.0