Re: [PATCH] DSPBRIDGE: Fix to avoid possible recursive locking
Hi Deepak, On Wed, 2010-02-10 at 02:01 +0100, ext Deepak Chitriki wrote: This patch fixes possible recursive locking detection.The implementation in which the spinlock is acquired and released is rectified in WMD_MSG_Get() to avoid locking contention. Added SYNC_EnterCS() and SYNC_LeaveCS()in WMD_MSG_Get()function. Cc: Ameya Palande ameya.pala...@nokia.com Cc: Omar Ramirez Luna omar.rami...@ti.com Cc: Nishanth Menon n...@ti.com Signed-off-by: Deepak Chitriki deepak.chitr...@ti.com --- drivers/dsp/bridge/wmd/msg_sm.c |4 1 files changed, 4 insertions(+), 0 deletions(-) diff --git a/drivers/dsp/bridge/wmd/msg_sm.c b/drivers/dsp/bridge/wmd/msg_sm.c index 50201e5..8faf5ad 100644 --- a/drivers/dsp/bridge/wmd/msg_sm.c +++ b/drivers/dsp/bridge/wmd/msg_sm.c @@ -300,8 +300,10 @@ DSP_STATUS WMD_MSG_Get(struct MSG_QUEUE *hMsgQueue, if (LST_IsEmpty(hMsgQueue-msgUsedList)) SYNC_ResetEvent(hMsgQueue-hSyncEvent); else { + (void)SYNC_LeaveCS(hMsgMgr-hSyncCS); NTFY_Notify(hMsgQueue-hNtfy, DSP_NODEMESSAGEREADY); + (void)SYNC_EnterCS(hMsgMgr-hSyncCS); SYNC_SetEvent(hMsgQueue-hSyncEvent); } @@ -352,8 +354,10 @@ DSP_STATUS WMD_MSG_Get(struct MSG_QUEUE *hMsgQueue, hMsgQueue-refCount--; /* Reset the event if there are still queued messages */ if (!LST_IsEmpty(hMsgQueue-msgUsedList)) { + (void)SYNC_LeaveCS(hMsgMgr-hSyncCS); NTFY_Notify(hMsgQueue-hNtfy, DSP_NODEMESSAGEREADY); + (void)SYNC_EnterCS(hMsgMgr-hSyncCS); SYNC_SetEvent(hMsgQueue-hSyncEvent); } /* Exit critical section */ Can you explain the need of calling NTFY_Notify() in WMD_MSG_Get()? I can see that the InputMsg calls NTFY_Notify() already! Can we get rid of NTFY_Notify() from WMD_MSG_Get() all together? Cheers, Ameya. -- To unsubscribe from this list: send the line unsubscribe linux-omap in the body of a message to majord...@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html
RE: [PATCH] DSPBRIDGE: Fix to avoid possible recursive locking
Hi Ameya, Yes I agree with your comments. As I understand NTFY_Notify() is called again inside WMD_MSG_Get() to make sure that notification is done in case if message queue is not empty.Since notification is already done once in InputMsg() while copying message to message queue,it doesn't make sense to notify again in WMD_MSG_Get().I guess this is not needed.I did a quick sanity testing removing NTFY_Notify() from WMD_MSG_Get() and no issues. I will rework the patch and update. Thanks, Deepak __ From: Ameya Palande [ameya.pala...@nokia.com] Sent: Wednesday, February 10, 2010 6:57 PM To: Chitriki Rudramuni, Deepak Cc: linux-omap; Ramirez Luna, Omar; Menon, Nishanth Subject: Re: [PATCH] DSPBRIDGE: Fix to avoid possible recursive locking Hi Deepak, On Wed, 2010-02-10 at 02:01 +0100, ext Deepak Chitriki wrote: This patch fixes possible recursive locking detection.The implementation in which the spinlock is acquired and released is rectified in WMD_MSG_Get() to avoid locking contention. Added SYNC_EnterCS() and SYNC_LeaveCS()in WMD_MSG_Get()function. Cc: Ameya Palande ameya.pala...@nokia.com Cc: Omar Ramirez Luna omar.rami...@ti.com Cc: Nishanth Menon n...@ti.com Signed-off-by: Deepak Chitriki deepak.chitr...@ti.com --- drivers/dsp/bridge/wmd/msg_sm.c |4 1 files changed, 4 insertions(+), 0 deletions(-) diff --git a/drivers/dsp/bridge/wmd/msg_sm.c b/drivers/dsp/bridge/wmd/msg_sm.c index 50201e5..8faf5ad 100644 --- a/drivers/dsp/bridge/wmd/msg_sm.c +++ b/drivers/dsp/bridge/wmd/msg_sm.c @@ -300,8 +300,10 @@ DSP_STATUS WMD_MSG_Get(struct MSG_QUEUE *hMsgQueue, if (LST_IsEmpty(hMsgQueue-msgUsedList)) SYNC_ResetEvent(hMsgQueue-hSyncEvent); else { + (void)SYNC_LeaveCS(hMsgMgr-hSyncCS); NTFY_Notify(hMsgQueue-hNtfy, DSP_NODEMESSAGEREADY); + (void)SYNC_EnterCS(hMsgMgr-hSyncCS); SYNC_SetEvent(hMsgQueue-hSyncEvent); } @@ -352,8 +354,10 @@ DSP_STATUS WMD_MSG_Get(struct MSG_QUEUE *hMsgQueue, hMsgQueue-refCount--; /* Reset the event if there are still queued messages */ if (!LST_IsEmpty(hMsgQueue-msgUsedList)) { + (void)SYNC_LeaveCS(hMsgMgr-hSyncCS); NTFY_Notify(hMsgQueue-hNtfy, DSP_NODEMESSAGEREADY); + (void)SYNC_EnterCS(hMsgMgr-hSyncCS); SYNC_SetEvent(hMsgQueue-hSyncEvent); } /* Exit critical section */ Can you explain the need of calling NTFY_Notify() in WMD_MSG_Get()? I can see that the InputMsg calls NTFY_Notify() already! Can we get rid of NTFY_Notify() from WMD_MSG_Get() all together? Cheers, Ameya.-- To unsubscribe from this list: send the line unsubscribe linux-omap in the body of a message to majord...@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html
[RESEND] [PATCH] DSPBRIDGE: Fix to avoid possible recursive locking
Please discard this patch.The rework patch will be available in v1. Thanks, Deepak From: linux-omap-ow...@vger.kernel.org [linux-omap-ow...@vger.kernel.org] On Behalf Of Chitriki Rudramuni, Deepak Sent: Thursday, February 11, 2010 4:44 AM To: Ameya Palande Cc: linux-omap; Ramirez Luna, Omar; Menon, Nishanth Subject: RE: [PATCH] DSPBRIDGE: Fix to avoid possible recursive locking Hi Ameya, Yes I agree with your comments. As I understand NTFY_Notify() is called again inside WMD_MSG_Get() to make sure that notification is done in case if message queue is not empty.Since notification is already done once in InputMsg() while copying message to message queue,it doesn't make sense to notify again in WMD_MSG_Get().I guess this is not needed.I did a quick sanity testing removing NTFY_Notify() from WMD_MSG_Get() and no issues. I will rework the patch and update. Thanks, Deepak __ From: Ameya Palande [ameya.pala...@nokia.com] Sent: Wednesday, February 10, 2010 6:57 PM To: Chitriki Rudramuni, Deepak Cc: linux-omap; Ramirez Luna, Omar; Menon, Nishanth Subject: Re: [PATCH] DSPBRIDGE: Fix to avoid possible recursive locking Hi Deepak, On Wed, 2010-02-10 at 02:01 +0100, ext Deepak Chitriki wrote: This patch fixes possible recursive locking detection.The implementation in which the spinlock is acquired and released is rectified in WMD_MSG_Get() to avoid locking contention. Added SYNC_EnterCS() and SYNC_LeaveCS()in WMD_MSG_Get()function. Cc: Ameya Palande ameya.pala...@nokia.com Cc: Omar Ramirez Luna omar.rami...@ti.com Cc: Nishanth Menon n...@ti.com Signed-off-by: Deepak Chitriki deepak.chitr...@ti.com --- drivers/dsp/bridge/wmd/msg_sm.c |4 1 files changed, 4 insertions(+), 0 deletions(-) diff --git a/drivers/dsp/bridge/wmd/msg_sm.c b/drivers/dsp/bridge/wmd/msg_sm.c index 50201e5..8faf5ad 100644 --- a/drivers/dsp/bridge/wmd/msg_sm.c +++ b/drivers/dsp/bridge/wmd/msg_sm.c @@ -300,8 +300,10 @@ DSP_STATUS WMD_MSG_Get(struct MSG_QUEUE *hMsgQueue, if (LST_IsEmpty(hMsgQueue-msgUsedList)) SYNC_ResetEvent(hMsgQueue-hSyncEvent); else { + (void)SYNC_LeaveCS(hMsgMgr-hSyncCS); NTFY_Notify(hMsgQueue-hNtfy, DSP_NODEMESSAGEREADY); + (void)SYNC_EnterCS(hMsgMgr-hSyncCS); SYNC_SetEvent(hMsgQueue-hSyncEvent); } @@ -352,8 +354,10 @@ DSP_STATUS WMD_MSG_Get(struct MSG_QUEUE *hMsgQueue, hMsgQueue-refCount--; /* Reset the event if there are still queued messages */ if (!LST_IsEmpty(hMsgQueue-msgUsedList)) { + (void)SYNC_LeaveCS(hMsgMgr-hSyncCS); NTFY_Notify(hMsgQueue-hNtfy, DSP_NODEMESSAGEREADY); + (void)SYNC_EnterCS(hMsgMgr-hSyncCS); SYNC_SetEvent(hMsgQueue-hSyncEvent); } /* Exit critical section */ Can you explain the need of calling NTFY_Notify() in WMD_MSG_Get()? I can see that the InputMsg calls NTFY_Notify() already! Can we get rid of NTFY_Notify() from WMD_MSG_Get() all together? Cheers, Ameya.-- To unsubscribe from this list: send the line unsubscribe linux-omap in the body of a message to majord...@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html-- To unsubscribe from this list: send the line unsubscribe linux-omap in the body of a message to majord...@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html
[PATCH] DSPBRIDGE: Fix to avoid possible recursive locking
This patch fixes possible recursive locking detection.The implementation in which the spinlock is acquired and released is rectified in WMD_MSG_Get() to avoid locking contention. Added SYNC_EnterCS() and SYNC_LeaveCS()in WMD_MSG_Get()function. Cc: Ameya Palande ameya.pala...@nokia.com Cc: Omar Ramirez Luna omar.rami...@ti.com Cc: Nishanth Menon n...@ti.com Signed-off-by: Deepak Chitriki deepak.chitr...@ti.com --- drivers/dsp/bridge/wmd/msg_sm.c |4 1 files changed, 4 insertions(+), 0 deletions(-) diff --git a/drivers/dsp/bridge/wmd/msg_sm.c b/drivers/dsp/bridge/wmd/msg_sm.c index 50201e5..8faf5ad 100644 --- a/drivers/dsp/bridge/wmd/msg_sm.c +++ b/drivers/dsp/bridge/wmd/msg_sm.c @@ -300,8 +300,10 @@ DSP_STATUS WMD_MSG_Get(struct MSG_QUEUE *hMsgQueue, if (LST_IsEmpty(hMsgQueue-msgUsedList)) SYNC_ResetEvent(hMsgQueue-hSyncEvent); else { + (void)SYNC_LeaveCS(hMsgMgr-hSyncCS); NTFY_Notify(hMsgQueue-hNtfy, DSP_NODEMESSAGEREADY); + (void)SYNC_EnterCS(hMsgMgr-hSyncCS); SYNC_SetEvent(hMsgQueue-hSyncEvent); } @@ -352,8 +354,10 @@ DSP_STATUS WMD_MSG_Get(struct MSG_QUEUE *hMsgQueue, hMsgQueue-refCount--; /* Reset the event if there are still queued messages */ if (!LST_IsEmpty(hMsgQueue-msgUsedList)) { + (void)SYNC_LeaveCS(hMsgMgr-hSyncCS); NTFY_Notify(hMsgQueue-hNtfy, DSP_NODEMESSAGEREADY); + (void)SYNC_EnterCS(hMsgMgr-hSyncCS); SYNC_SetEvent(hMsgQueue-hSyncEvent); } /* Exit critical section */ -- 1.6.3.3 -- To unsubscribe from this list: send the line unsubscribe linux-omap in the body of a message to majord...@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html