diff mbox

[v2] iwlwifi: pcie: move rx workqueue initialization to iwl_trans_pcie_alloc()

Message ID 20170822073729.4856-1-luca@coelho.fi (mailing list archive)
State Accepted
Commit 10a54d8196d11f6cc0db2f71249f93854cb9fe55
Delegated to: Kalle Valo
Headers show

Commit Message

Luca Coelho Aug. 22, 2017, 7:37 a.m. UTC
From: Luca Coelho <luciano.coelho@intel.com>

Work queues cannot be allocated when a mutex is held because the mutex
may be in use and that would make it sleep.  Doing so generates the
following splat with 4.13+:

[   19.513298] ======================================================
[   19.513429] WARNING: possible circular locking dependency detected
[   19.513557] 4.13.0-rc5+ #6 Not tainted
[   19.513638] ------------------------------------------------------
[   19.513767] cpuhp/0/12 is trying to acquire lock:
[   19.513867]  (&tz->lock){+.+.+.}, at: [<ffffffff924afebb>] thermal_zone_get_temp+0x5b/0xb0
[   19.514047]
[   19.514047] but task is already holding lock:
[   19.514166]  (cpuhp_state){+.+.+.}, at: [<ffffffff91cc4baa>] cpuhp_thread_fun+0x3a/0x210
[   19.514338]
[   19.514338] which lock already depends on the new lock.

This lock dependency already existed with previous kernel versions,
but it was not detected until commit 49dfe2a67797 ("cpuhotplug: Link
lock stacks for hotplug callbacks") was introduced.

Reported-by: David Weinehall <david.weinehall@intel.com>
Reported-by: Jiri Kosina <jikos@kernel.org>
Signed-off-by: Luca Coelho <luciano.coelho@intel.com>
---
In v2:
   - updated the commit message to a new version, with a grammar fix
     and the actual commit that exposed the problem;

 drivers/net/wireless/intel/iwlwifi/pcie/internal.h |  2 ++
 drivers/net/wireless/intel/iwlwifi/pcie/rx.c       | 10 +---------
 drivers/net/wireless/intel/iwlwifi/pcie/trans.c    |  9 +++++++++
 3 files changed, 12 insertions(+), 9 deletions(-)

Comments

Luca Coelho Aug. 24, 2017, 6 a.m. UTC | #1
On Tue, 2017-08-22 at 10:37 +0300, Luca Coelho wrote:
> From: Luca Coelho <luciano.coelho@intel.com>
> 
> Work queues cannot be allocated when a mutex is held because the mutex
> may be in use and that would make it sleep.  Doing so generates the
> following splat with 4.13+:
> 
> [   19.513298] ======================================================
> [   19.513429] WARNING: possible circular locking dependency detected
> [   19.513557] 4.13.0-rc5+ #6 Not tainted
> [   19.513638] ------------------------------------------------------
> [   19.513767] cpuhp/0/12 is trying to acquire lock:
> [   19.513867]  (&tz->lock){+.+.+.}, at: [<ffffffff924afebb>] thermal_zone_get_temp+0x5b/0xb0
> [   19.514047]
> [   19.514047] but task is already holding lock:
> [   19.514166]  (cpuhp_state){+.+.+.}, at: [<ffffffff91cc4baa>] cpuhp_thread_fun+0x3a/0x210
> [   19.514338]
> [   19.514338] which lock already depends on the new lock.
> 
> This lock dependency already existed with previous kernel versions,
> but it was not detected until commit 49dfe2a67797 ("cpuhotplug: Link
> lock stacks for hotplug callbacks") was introduced.
> 
> Reported-by: David Weinehall <david.weinehall@intel.com>
> Reported-by: Jiri Kosina <jikos@kernel.org>
> Signed-off-by: Luca Coelho <luciano.coelho@intel.com>

Jiri, did you have a chance to try this out? I'm about to ask Kalle to
merge this so it gets in in time for 4.13-rc7.

--
Cheers,
Luca.
Kalle Valo Aug. 24, 2017, 1:50 p.m. UTC | #2
Luciano Coelho <luca@coelho.fi> wrote:

> From: Luca Coelho <luciano.coelho@intel.com>
> 
> Work queues cannot be allocated when a mutex is held because the mutex
> may be in use and that would make it sleep.  Doing so generates the
> following splat with 4.13+:
> 
> [   19.513298] ======================================================
> [   19.513429] WARNING: possible circular locking dependency detected
> [   19.513557] 4.13.0-rc5+ #6 Not tainted
> [   19.513638] ------------------------------------------------------
> [   19.513767] cpuhp/0/12 is trying to acquire lock:
> [   19.513867]  (&tz->lock){+.+.+.}, at: [<ffffffff924afebb>] thermal_zone_get_temp+0x5b/0xb0
> [   19.514047]
> [   19.514047] but task is already holding lock:
> [   19.514166]  (cpuhp_state){+.+.+.}, at: [<ffffffff91cc4baa>] cpuhp_thread_fun+0x3a/0x210
> [   19.514338]
> [   19.514338] which lock already depends on the new lock.
> 
> This lock dependency already existed with previous kernel versions,
> but it was not detected until commit 49dfe2a67797 ("cpuhotplug: Link
> lock stacks for hotplug callbacks") was introduced.
> 
> Reported-by: David Weinehall <david.weinehall@intel.com>
> Reported-by: Jiri Kosina <jikos@kernel.org>
> Signed-off-by: Luca Coelho <luciano.coelho@intel.com>

Patch applied to wireless-drivers.git, thanks.

10a54d8196d1 iwlwifi: pcie: move rx workqueue initialization to iwl_trans_pcie_alloc()
Jiri Kosina Aug. 24, 2017, 7:56 p.m. UTC | #3
On Thu, 24 Aug 2017, Luca Coelho wrote:

> > Work queues cannot be allocated when a mutex is held because the mutex
> > may be in use and that would make it sleep.  Doing so generates the
> > following splat with 4.13+:
> > 
> > [   19.513298] ======================================================
> > [   19.513429] WARNING: possible circular locking dependency detected
> > [   19.513557] 4.13.0-rc5+ #6 Not tainted
> > [   19.513638] ------------------------------------------------------
> > [   19.513767] cpuhp/0/12 is trying to acquire lock:
> > [   19.513867]  (&tz->lock){+.+.+.}, at: [<ffffffff924afebb>] thermal_zone_get_temp+0x5b/0xb0
> > [   19.514047]
> > [   19.514047] but task is already holding lock:
> > [   19.514166]  (cpuhp_state){+.+.+.}, at: [<ffffffff91cc4baa>] cpuhp_thread_fun+0x3a/0x210
> > [   19.514338]
> > [   19.514338] which lock already depends on the new lock.
> > 
> > This lock dependency already existed with previous kernel versions,
> > but it was not detected until commit 49dfe2a67797 ("cpuhotplug: Link
> > lock stacks for hotplug callbacks") was introduced.
> > 
> > Reported-by: David Weinehall <david.weinehall@intel.com>
> > Reported-by: Jiri Kosina <jikos@kernel.org>
> > Signed-off-by: Luca Coelho <luciano.coelho@intel.com>
> 
> Jiri, did you have a chance to try this out? I'm about to ask Kalle to
> merge this so it gets in in time for 4.13-rc7.

Sorry, I am almost completely offline for one more week (vacation), and 
will not have access to the affected system before that. But this indeed 
looks like a correct fix to me, so feel free to add

	Acked-by: Jiri Kosina <jkosina@suse.cz>

I'll be able to provide my Tested-by: eventually only in ~10 days.

Thanks,
Luca Coelho Aug. 24, 2017, 8 p.m. UTC | #4
On Thu, 2017-08-24 at 21:56 +0200, Jiri Kosina wrote:
> On Thu, 24 Aug 2017, Luca Coelho wrote:
> 
> > > Work queues cannot be allocated when a mutex is held because the mutex
> > > may be in use and that would make it sleep.  Doing so generates the
> > > following splat with 4.13+:
> > > 
> > > [   19.513298] ======================================================
> > > [   19.513429] WARNING: possible circular locking dependency detected
> > > [   19.513557] 4.13.0-rc5+ #6 Not tainted
> > > [   19.513638] ------------------------------------------------------
> > > [   19.513767] cpuhp/0/12 is trying to acquire lock:
> > > [   19.513867]  (&tz->lock){+.+.+.}, at: [<ffffffff924afebb>] thermal_zone_get_temp+0x5b/0xb0
> > > [   19.514047]
> > > [   19.514047] but task is already holding lock:
> > > [   19.514166]  (cpuhp_state){+.+.+.}, at: [<ffffffff91cc4baa>] cpuhp_thread_fun+0x3a/0x210
> > > [   19.514338]
> > > [   19.514338] which lock already depends on the new lock.
> > > 
> > > This lock dependency already existed with previous kernel versions,
> > > but it was not detected until commit 49dfe2a67797 ("cpuhotplug: Link
> > > lock stacks for hotplug callbacks") was introduced.
> > > 
> > > Reported-by: David Weinehall <david.weinehall@intel.com>
> > > Reported-by: Jiri Kosina <jikos@kernel.org>
> > > Signed-off-by: Luca Coelho <luciano.coelho@intel.com>
> > 
> > Jiri, did you have a chance to try this out? I'm about to ask Kalle to
> > merge this so it gets in in time for 4.13-rc7.
> 
> Sorry, I am almost completely offline for one more week (vacation), and 
> will not have access to the affected system before that.

Sounds good! Enjoy! ;)


>  But this indeed 
> looks like a correct fix to me, so feel free to add
> 
> 	Acked-by: Jiri Kosina <jkosina@suse.cz>
> 
> I'll be able to provide my Tested-by: eventually only in ~10 days.


Kalle already picked it up in wireless-drivers and this should make it's
way to 4.13-rc7 (we hope).

In any case, thanks for reporting and the help debugging it.

--
Cheers,
Luca.
Kalle Valo Aug. 29, 2017, 8:19 a.m. UTC | #5
Luca Coelho <luca@coelho.fi> writes:

> On Thu, 2017-08-24 at 21:56 +0200, Jiri Kosina wrote:
>> On Thu, 24 Aug 2017, Luca Coelho wrote:
>> 
>> > > Work queues cannot be allocated when a mutex is held because the mutex
>> > > may be in use and that would make it sleep.  Doing so generates the
>> > > following splat with 4.13+:
>> > > 
>> > > [   19.513298] ======================================================
>> > > [   19.513429] WARNING: possible circular locking dependency detected
>> > > [   19.513557] 4.13.0-rc5+ #6 Not tainted
>> > > [   19.513638] ------------------------------------------------------
>> > > [   19.513767] cpuhp/0/12 is trying to acquire lock:
>> > > [ 19.513867] (&tz->lock){+.+.+.}, at: [<ffffffff924afebb>]
>> > > thermal_zone_get_temp+0x5b/0xb0
>> > > [   19.514047]
>> > > [   19.514047] but task is already holding lock:
>> > > [ 19.514166] (cpuhp_state){+.+.+.}, at: [<ffffffff91cc4baa>]
>> > > cpuhp_thread_fun+0x3a/0x210
>> > > [   19.514338]
>> > > [   19.514338] which lock already depends on the new lock.
>> > > 
>> > > This lock dependency already existed with previous kernel versions,
>> > > but it was not detected until commit 49dfe2a67797 ("cpuhotplug: Link
>> > > lock stacks for hotplug callbacks") was introduced.
>> > > 
>> > > Reported-by: David Weinehall <david.weinehall@intel.com>
>> > > Reported-by: Jiri Kosina <jikos@kernel.org>
>> > > Signed-off-by: Luca Coelho <luciano.coelho@intel.com>
>> > 
>> > Jiri, did you have a chance to try this out? I'm about to ask Kalle to
>> > merge this so it gets in in time for 4.13-rc7.
>> 
>> Sorry, I am almost completely offline for one more week (vacation), and 
>> will not have access to the affected system before that.
>
> Sounds good! Enjoy! ;)
>
>
>>  But this indeed 
>> looks like a correct fix to me, so feel free to add
>> 
>> 	Acked-by: Jiri Kosina <jkosina@suse.cz>
>> 
>> I'll be able to provide my Tested-by: eventually only in ~10 days.
>
>
> Kalle already picked it up in wireless-drivers and this should make it's
> way to 4.13-rc7 (we hope).

It (10a54d8196d1) didn't make it to -rc7 but it's in net tree now and
should make it to the next release on Sunday (either -rc8 or the final).
David Weinehall Aug. 30, 2017, 2:57 p.m. UTC | #6
On Tue, Aug 22, 2017 at 10:37:29AM +0300, Luca Coelho wrote:
> From: Luca Coelho <luciano.coelho@intel.com>
> 
> Work queues cannot be allocated when a mutex is held because the mutex
> may be in use and that would make it sleep.  Doing so generates the
> following splat with 4.13+:
> 
> [   19.513298] ======================================================
> [   19.513429] WARNING: possible circular locking dependency detected
> [   19.513557] 4.13.0-rc5+ #6 Not tainted
> [   19.513638] ------------------------------------------------------
> [   19.513767] cpuhp/0/12 is trying to acquire lock:
> [   19.513867]  (&tz->lock){+.+.+.}, at: [<ffffffff924afebb>] thermal_zone_get_temp+0x5b/0xb0
> [   19.514047]
> [   19.514047] but task is already holding lock:
> [   19.514166]  (cpuhp_state){+.+.+.}, at: [<ffffffff91cc4baa>] cpuhp_thread_fun+0x3a/0x210
> [   19.514338]
> [   19.514338] which lock already depends on the new lock.
> 
> This lock dependency already existed with previous kernel versions,
> but it was not detected until commit 49dfe2a67797 ("cpuhotplug: Link
> lock stacks for hotplug callbacks") was introduced.
> 
> Reported-by: David Weinehall <david.weinehall@intel.com>
> Reported-by: Jiri Kosina <jikos@kernel.org>
> Signed-off-by: Luca Coelho <luciano.coelho@intel.com>

With this patch I no longer get the lockdep warning,
and the driver seems to work just as well as before.

Thanks!

Tested-by: David Weinehall <david.weinehall@linux.intel.com>

> ---
> In v2:
>    - updated the commit message to a new version, with a grammar fix
>      and the actual commit that exposed the problem;
> 
>  drivers/net/wireless/intel/iwlwifi/pcie/internal.h |  2 ++
>  drivers/net/wireless/intel/iwlwifi/pcie/rx.c       | 10 +---------
>  drivers/net/wireless/intel/iwlwifi/pcie/trans.c    |  9 +++++++++
>  3 files changed, 12 insertions(+), 9 deletions(-)
> 
> diff --git a/drivers/net/wireless/intel/iwlwifi/pcie/internal.h b/drivers/net/wireless/intel/iwlwifi/pcie/internal.h
> index fa315d84e98e..a1ea9ef97ed9 100644
> --- a/drivers/net/wireless/intel/iwlwifi/pcie/internal.h
> +++ b/drivers/net/wireless/intel/iwlwifi/pcie/internal.h
> @@ -787,6 +787,8 @@ int iwl_pci_fw_enter_d0i3(struct iwl_trans *trans);
>  
>  void iwl_pcie_enable_rx_wake(struct iwl_trans *trans, bool enable);
>  
> +void iwl_pcie_rx_allocator_work(struct work_struct *data);
> +
>  /* common functions that are used by gen2 transport */
>  void iwl_pcie_apm_config(struct iwl_trans *trans);
>  int iwl_pcie_prepare_card_hw(struct iwl_trans *trans);
> diff --git a/drivers/net/wireless/intel/iwlwifi/pcie/rx.c b/drivers/net/wireless/intel/iwlwifi/pcie/rx.c
> index 351c4423125a..942736d3fa75 100644
> --- a/drivers/net/wireless/intel/iwlwifi/pcie/rx.c
> +++ b/drivers/net/wireless/intel/iwlwifi/pcie/rx.c
> @@ -597,7 +597,7 @@ static void iwl_pcie_rx_allocator_get(struct iwl_trans *trans,
>  	rxq->free_count += RX_CLAIM_REQ_ALLOC;
>  }
>  
> -static void iwl_pcie_rx_allocator_work(struct work_struct *data)
> +void iwl_pcie_rx_allocator_work(struct work_struct *data)
>  {
>  	struct iwl_rb_allocator *rba_p =
>  		container_of(data, struct iwl_rb_allocator, rx_alloc);
> @@ -900,10 +900,6 @@ static int _iwl_pcie_rx_init(struct iwl_trans *trans)
>  			return err;
>  	}
>  	def_rxq = trans_pcie->rxq;
> -	if (!rba->alloc_wq)
> -		rba->alloc_wq = alloc_workqueue("rb_allocator",
> -						WQ_HIGHPRI | WQ_UNBOUND, 1);
> -	INIT_WORK(&rba->rx_alloc, iwl_pcie_rx_allocator_work);
>  
>  	spin_lock(&rba->lock);
>  	atomic_set(&rba->req_pending, 0);
> @@ -1017,10 +1013,6 @@ void iwl_pcie_rx_free(struct iwl_trans *trans)
>  	}
>  
>  	cancel_work_sync(&rba->rx_alloc);
> -	if (rba->alloc_wq) {
> -		destroy_workqueue(rba->alloc_wq);
> -		rba->alloc_wq = NULL;
> -	}
>  
>  	iwl_pcie_free_rbs_pool(trans);
>  
> diff --git a/drivers/net/wireless/intel/iwlwifi/pcie/trans.c b/drivers/net/wireless/intel/iwlwifi/pcie/trans.c
> index f95eec52508e..3927bbf04f72 100644
> --- a/drivers/net/wireless/intel/iwlwifi/pcie/trans.c
> +++ b/drivers/net/wireless/intel/iwlwifi/pcie/trans.c
> @@ -1786,6 +1786,11 @@ void iwl_trans_pcie_free(struct iwl_trans *trans)
>  		iwl_pcie_tx_free(trans);
>  	iwl_pcie_rx_free(trans);
>  
> +	if (trans_pcie->rba.alloc_wq) {
> +		destroy_workqueue(trans_pcie->rba.alloc_wq);
> +		trans_pcie->rba.alloc_wq = NULL;
> +	}
> +
>  	if (trans_pcie->msix_enabled) {
>  		for (i = 0; i < trans_pcie->alloc_vecs; i++) {
>  			irq_set_affinity_hint(
> @@ -3169,6 +3174,10 @@ struct iwl_trans *iwl_trans_pcie_alloc(struct pci_dev *pdev,
>  		trans_pcie->inta_mask = CSR_INI_SET_MASK;
>  	 }
>  
> +	trans_pcie->rba.alloc_wq = alloc_workqueue("rb_allocator",
> +						   WQ_HIGHPRI | WQ_UNBOUND, 1);
> +	INIT_WORK(&trans_pcie->rba.rx_alloc, iwl_pcie_rx_allocator_work);
> +
>  #ifdef CONFIG_IWLWIFI_PCIE_RTPM
>  	trans->runtime_pm_mode = IWL_PLAT_PM_MODE_D0I3;
>  #else
> -- 
> 2.14.1
>
Luca Coelho Aug. 31, 2017, 6:11 a.m. UTC | #7
On Wed, 2017-08-30 at 17:57 +0300, David Weinehall wrote:
> On Tue, Aug 22, 2017 at 10:37:29AM +0300, Luca Coelho wrote:
> > From: Luca Coelho <luciano.coelho@intel.com>
> > 
> > Work queues cannot be allocated when a mutex is held because the mutex
> > may be in use and that would make it sleep.  Doing so generates the
> > following splat with 4.13+:
> > 
> > [   19.513298] ======================================================
> > [   19.513429] WARNING: possible circular locking dependency detected
> > [   19.513557] 4.13.0-rc5+ #6 Not tainted
> > [   19.513638] ------------------------------------------------------
> > [   19.513767] cpuhp/0/12 is trying to acquire lock:
> > [   19.513867]  (&tz->lock){+.+.+.}, at: [<ffffffff924afebb>] thermal_zone_get_temp+0x5b/0xb0
> > [   19.514047]
> > [   19.514047] but task is already holding lock:
> > [   19.514166]  (cpuhp_state){+.+.+.}, at: [<ffffffff91cc4baa>] cpuhp_thread_fun+0x3a/0x210
> > [   19.514338]
> > [   19.514338] which lock already depends on the new lock.
> > 
> > This lock dependency already existed with previous kernel versions,
> > but it was not detected until commit 49dfe2a67797 ("cpuhotplug: Link
> > lock stacks for hotplug callbacks") was introduced.
> > 
> > Reported-by: David Weinehall <david.weinehall@intel.com>
> > Reported-by: Jiri Kosina <jikos@kernel.org>
> > Signed-off-by: Luca Coelho <luciano.coelho@intel.com>
> 
> With this patch I no longer get the lockdep warning,
> and the driver seems to work just as well as before.

Great! Thanks for reporting and testing, David!


> Thanks!
> 
> Tested-by: David Weinehall <david.weinehall@linux.intel.com>

Thanks for this too, but unfortunately it's too late to add it, since
the patch is already in net tree.

--
Cheers,
Luca.
Jiri Kosina Sept. 4, 2017, 11:43 a.m. UTC | #8
On Thu, 24 Aug 2017, Luca Coelho wrote:

> > looks like a correct fix to me, so feel free to add
> > 
> > 	Acked-by: Jiri Kosina <jkosina@suse.cz>
> > 
> > I'll be able to provide my Tested-by: eventually only in ~10 days.
> 
> 
> Kalle already picked it up in wireless-drivers and this should make it's
> way to 4.13-rc7 (we hope).
> 
> In any case, thanks for reporting and the help debugging it.

I know it's pretty late for this to be added to the commit, but I don't 
want this to be left in the void, so for the sake of completness:

	Tested-by: Jiri Kosina <jkosina@suse.cz>

Thanks,
Luca Coelho Sept. 4, 2017, 11:44 a.m. UTC | #9
On Mon, 2017-09-04 at 13:43 +0200, Jiri Kosina wrote:
> On Thu, 24 Aug 2017, Luca Coelho wrote:
> 
> > > looks like a correct fix to me, so feel free to add
> > > 
> > > 	Acked-by: Jiri Kosina <jkosina@suse.cz>
> > > 
> > > I'll be able to provide my Tested-by: eventually only in ~10 days.
> > 
> > 
> > Kalle already picked it up in wireless-drivers and this should make it's
> > way to 4.13-rc7 (we hope).
> > 
> > In any case, thanks for reporting and the help debugging it.
> 
> I know it's pretty late for this to be added to the commit, but I don't 
> want this to be left in the void, so for the sake of completness:
> 
> 	Tested-by: Jiri Kosina <jkosina@suse.cz>

Thanks, Jiri, for reporting, debugging and testing!

--
Cheers,
Luca.
diff mbox

Patch

diff --git a/drivers/net/wireless/intel/iwlwifi/pcie/internal.h b/drivers/net/wireless/intel/iwlwifi/pcie/internal.h
index fa315d84e98e..a1ea9ef97ed9 100644
--- a/drivers/net/wireless/intel/iwlwifi/pcie/internal.h
+++ b/drivers/net/wireless/intel/iwlwifi/pcie/internal.h
@@ -787,6 +787,8 @@  int iwl_pci_fw_enter_d0i3(struct iwl_trans *trans);
 
 void iwl_pcie_enable_rx_wake(struct iwl_trans *trans, bool enable);
 
+void iwl_pcie_rx_allocator_work(struct work_struct *data);
+
 /* common functions that are used by gen2 transport */
 void iwl_pcie_apm_config(struct iwl_trans *trans);
 int iwl_pcie_prepare_card_hw(struct iwl_trans *trans);
diff --git a/drivers/net/wireless/intel/iwlwifi/pcie/rx.c b/drivers/net/wireless/intel/iwlwifi/pcie/rx.c
index 351c4423125a..942736d3fa75 100644
--- a/drivers/net/wireless/intel/iwlwifi/pcie/rx.c
+++ b/drivers/net/wireless/intel/iwlwifi/pcie/rx.c
@@ -597,7 +597,7 @@  static void iwl_pcie_rx_allocator_get(struct iwl_trans *trans,
 	rxq->free_count += RX_CLAIM_REQ_ALLOC;
 }
 
-static void iwl_pcie_rx_allocator_work(struct work_struct *data)
+void iwl_pcie_rx_allocator_work(struct work_struct *data)
 {
 	struct iwl_rb_allocator *rba_p =
 		container_of(data, struct iwl_rb_allocator, rx_alloc);
@@ -900,10 +900,6 @@  static int _iwl_pcie_rx_init(struct iwl_trans *trans)
 			return err;
 	}
 	def_rxq = trans_pcie->rxq;
-	if (!rba->alloc_wq)
-		rba->alloc_wq = alloc_workqueue("rb_allocator",
-						WQ_HIGHPRI | WQ_UNBOUND, 1);
-	INIT_WORK(&rba->rx_alloc, iwl_pcie_rx_allocator_work);
 
 	spin_lock(&rba->lock);
 	atomic_set(&rba->req_pending, 0);
@@ -1017,10 +1013,6 @@  void iwl_pcie_rx_free(struct iwl_trans *trans)
 	}
 
 	cancel_work_sync(&rba->rx_alloc);
-	if (rba->alloc_wq) {
-		destroy_workqueue(rba->alloc_wq);
-		rba->alloc_wq = NULL;
-	}
 
 	iwl_pcie_free_rbs_pool(trans);
 
diff --git a/drivers/net/wireless/intel/iwlwifi/pcie/trans.c b/drivers/net/wireless/intel/iwlwifi/pcie/trans.c
index f95eec52508e..3927bbf04f72 100644
--- a/drivers/net/wireless/intel/iwlwifi/pcie/trans.c
+++ b/drivers/net/wireless/intel/iwlwifi/pcie/trans.c
@@ -1786,6 +1786,11 @@  void iwl_trans_pcie_free(struct iwl_trans *trans)
 		iwl_pcie_tx_free(trans);
 	iwl_pcie_rx_free(trans);
 
+	if (trans_pcie->rba.alloc_wq) {
+		destroy_workqueue(trans_pcie->rba.alloc_wq);
+		trans_pcie->rba.alloc_wq = NULL;
+	}
+
 	if (trans_pcie->msix_enabled) {
 		for (i = 0; i < trans_pcie->alloc_vecs; i++) {
 			irq_set_affinity_hint(
@@ -3169,6 +3174,10 @@  struct iwl_trans *iwl_trans_pcie_alloc(struct pci_dev *pdev,
 		trans_pcie->inta_mask = CSR_INI_SET_MASK;
 	 }
 
+	trans_pcie->rba.alloc_wq = alloc_workqueue("rb_allocator",
+						   WQ_HIGHPRI | WQ_UNBOUND, 1);
+	INIT_WORK(&trans_pcie->rba.rx_alloc, iwl_pcie_rx_allocator_work);
+
 #ifdef CONFIG_IWLWIFI_PCIE_RTPM
 	trans->runtime_pm_mode = IWL_PLAT_PM_MODE_D0I3;
 #else