* [PATCH] cxl: Set the valid bit in PE for dedicated mode
@ 2017-08-28 4:15 Vaibhav Jain
2017-08-28 4:25 ` Andrew Donnellan
` (3 more replies)
0 siblings, 4 replies; 7+ messages in thread
From: Vaibhav Jain @ 2017-08-28 4:15 UTC (permalink / raw)
To: linuxppc-dev, Frederic Barrat
Cc: Vaibhav Jain, Andrew Donnellan, Christophe Lombard,
Philippe Bergheaud, Alastair D'Silva
Make sure to set the valid-bit in software-state field of the
populated PE. This was earlier missing for dedicated mode AFUs, hence
was causing a PSL freeze when the AFU was activated.
Signed-off-by: Vaibhav Jain <vaibhav@linux.vnet.ibm.com>
---
drivers/misc/cxl/native.c | 4 ++++
1 file changed, 4 insertions(+)
diff --git a/drivers/misc/cxl/native.c b/drivers/misc/cxl/native.c
index 4a82c313cf71..0e748c682ee9 100644
--- a/drivers/misc/cxl/native.c
+++ b/drivers/misc/cxl/native.c
@@ -897,6 +897,10 @@ int cxl_attach_dedicated_process_psl9(struct cxl_context *ctx, u64 wed, u64 amr)
if (ctx->afu->adapter->native->sl_ops->update_dedicated_ivtes)
afu->adapter->native->sl_ops->update_dedicated_ivtes(ctx);
+ ctx->elem->software_state = cpu_to_be32(CXL_PE_SOFTWARE_STATE_V);
+ /* Make sure the changes to the PE are visible to the card */
+ smp_wmb();
+
result = cxl_ops->afu_reset(afu);
if (result)
return result;
--
2.13.5
^ permalink raw reply related [flat|nested] 7+ messages in thread
* Re: [PATCH] cxl: Set the valid bit in PE for dedicated mode
2017-08-28 4:15 [PATCH] cxl: Set the valid bit in PE for dedicated mode Vaibhav Jain
@ 2017-08-28 4:25 ` Andrew Donnellan
2017-08-28 8:58 ` christophe lombard
` (2 subsequent siblings)
3 siblings, 0 replies; 7+ messages in thread
From: Andrew Donnellan @ 2017-08-28 4:25 UTC (permalink / raw)
To: Vaibhav Jain, linuxppc-dev, Frederic Barrat
Cc: Christophe Lombard, Philippe Bergheaud, Alastair D'Silva
On 28/08/17 14:15, Vaibhav Jain wrote:
> Make sure to set the valid-bit in software-state field of the
> populated PE. This was earlier missing for dedicated mode AFUs, hence
> was causing a PSL freeze when the AFU was activated.
>
> Signed-off-by: Vaibhav Jain <vaibhav@linux.vnet.ibm.com>
Acked-by: Andrew Donnellan <andrew.donnellan@au1.ibm.com>
> ---
> drivers/misc/cxl/native.c | 4 ++++
> 1 file changed, 4 insertions(+)
>
> diff --git a/drivers/misc/cxl/native.c b/drivers/misc/cxl/native.c
> index 4a82c313cf71..0e748c682ee9 100644
> --- a/drivers/misc/cxl/native.c
> +++ b/drivers/misc/cxl/native.c
> @@ -897,6 +897,10 @@ int cxl_attach_dedicated_process_psl9(struct cxl_context *ctx, u64 wed, u64 amr)
> if (ctx->afu->adapter->native->sl_ops->update_dedicated_ivtes)
> afu->adapter->native->sl_ops->update_dedicated_ivtes(ctx);
>
> + ctx->elem->software_state = cpu_to_be32(CXL_PE_SOFTWARE_STATE_V);
> + /* Make sure the changes to the PE are visible to the card */
> + smp_wmb();
> +
> result = cxl_ops->afu_reset(afu);
> if (result)
> return result;
>
--
Andrew Donnellan OzLabs, ADL Canberra
andrew.donnellan@au1.ibm.com IBM Australia Limited
^ permalink raw reply [flat|nested] 7+ messages in thread
* Re: [PATCH] cxl: Set the valid bit in PE for dedicated mode
2017-08-28 4:15 [PATCH] cxl: Set the valid bit in PE for dedicated mode Vaibhav Jain
2017-08-28 4:25 ` Andrew Donnellan
@ 2017-08-28 8:58 ` christophe lombard
2017-08-28 9:42 ` Frederic Barrat
2017-08-29 6:53 ` Michael Ellerman
3 siblings, 0 replies; 7+ messages in thread
From: christophe lombard @ 2017-08-28 8:58 UTC (permalink / raw)
To: Vaibhav Jain, linuxppc-dev, Frederic Barrat
Cc: Philippe Bergheaud, Alastair D'Silva, Andrew Donnellan,
Christophe Lombard
Le 28/08/2017 à 06:15, Vaibhav Jain a écrit :
> Make sure to set the valid-bit in software-state field of the
> populated PE. This was earlier missing for dedicated mode AFUs, hence
> was causing a PSL freeze when the AFU was activated.
Acked-by: Christophe Lombard <clombard@linux.vnet.ibm.com>
> Signed-off-by: Vaibhav Jain <vaibhav@linux.vnet.ibm.com>
> ---
> drivers/misc/cxl/native.c | 4 ++++
> 1 file changed, 4 insertions(+)
>
> diff --git a/drivers/misc/cxl/native.c b/drivers/misc/cxl/native.c
> index 4a82c313cf71..0e748c682ee9 100644
> --- a/drivers/misc/cxl/native.c
> +++ b/drivers/misc/cxl/native.c
> @@ -897,6 +897,10 @@ int cxl_attach_dedicated_process_psl9(struct cxl_context *ctx, u64 wed, u64 amr)
> if (ctx->afu->adapter->native->sl_ops->update_dedicated_ivtes)
> afu->adapter->native->sl_ops->update_dedicated_ivtes(ctx);
>
> + ctx->elem->software_state = cpu_to_be32(CXL_PE_SOFTWARE_STATE_V);
> + /* Make sure the changes to the PE are visible to the card */
> + smp_wmb();
> +
> result = cxl_ops->afu_reset(afu);
> if (result)
> return result;
^ permalink raw reply [flat|nested] 7+ messages in thread
* Re: [PATCH] cxl: Set the valid bit in PE for dedicated mode
2017-08-28 4:15 [PATCH] cxl: Set the valid bit in PE for dedicated mode Vaibhav Jain
2017-08-28 4:25 ` Andrew Donnellan
2017-08-28 8:58 ` christophe lombard
@ 2017-08-28 9:42 ` Frederic Barrat
2017-08-29 6:53 ` Michael Ellerman
3 siblings, 0 replies; 7+ messages in thread
From: Frederic Barrat @ 2017-08-28 9:42 UTC (permalink / raw)
To: Vaibhav Jain, linuxppc-dev
Cc: Andrew Donnellan, Christophe Lombard, Philippe Bergheaud,
Alastair D'Silva
Le 28/08/2017 à 06:15, Vaibhav Jain a écrit :
> Make sure to set the valid-bit in software-state field of the
> populated PE. This was earlier missing for dedicated mode AFUs, hence
> was causing a PSL freeze when the AFU was activated.
>
> Signed-off-by: Vaibhav Jain <vaibhav@linux.vnet.ibm.com>
> ---
Acked-by: Frederic Barrat <fbarrat@linux.vnet.ibm.com>
Thanks!
> drivers/misc/cxl/native.c | 4 ++++
> 1 file changed, 4 insertions(+)
>
> diff --git a/drivers/misc/cxl/native.c b/drivers/misc/cxl/native.c
> index 4a82c313cf71..0e748c682ee9 100644
> --- a/drivers/misc/cxl/native.c
> +++ b/drivers/misc/cxl/native.c
> @@ -897,6 +897,10 @@ int cxl_attach_dedicated_process_psl9(struct cxl_context *ctx, u64 wed, u64 amr)
> if (ctx->afu->adapter->native->sl_ops->update_dedicated_ivtes)
> afu->adapter->native->sl_ops->update_dedicated_ivtes(ctx);
>
> + ctx->elem->software_state = cpu_to_be32(CXL_PE_SOFTWARE_STATE_V);
> + /* Make sure the changes to the PE are visible to the card */
> + smp_wmb();
> +
> result = cxl_ops->afu_reset(afu);
> if (result)
> return result;
>
^ permalink raw reply [flat|nested] 7+ messages in thread
* Re: [PATCH] cxl: Set the valid bit in PE for dedicated mode
2017-08-28 4:15 [PATCH] cxl: Set the valid bit in PE for dedicated mode Vaibhav Jain
` (2 preceding siblings ...)
2017-08-28 9:42 ` Frederic Barrat
@ 2017-08-29 6:53 ` Michael Ellerman
2017-08-30 12:04 ` Vaibhav Jain
3 siblings, 1 reply; 7+ messages in thread
From: Michael Ellerman @ 2017-08-29 6:53 UTC (permalink / raw)
To: Vaibhav Jain, linuxppc-dev, Frederic Barrat
Cc: Philippe Bergheaud, Alastair D'Silva, Vaibhav Jain,
Andrew Donnellan, Christophe Lombard
Vaibhav Jain <vaibhav@linux.vnet.ibm.com> writes:
> Make sure to set the valid-bit in software-state field of the
> populated PE. This was earlier missing for dedicated mode AFUs, hence
> was causing a PSL freeze when the AFU was activated.
>
> Signed-off-by: Vaibhav Jain <vaibhav@linux.vnet.ibm.com>
> ---
> drivers/misc/cxl/native.c | 4 ++++
> 1 file changed, 4 insertions(+)
>
> diff --git a/drivers/misc/cxl/native.c b/drivers/misc/cxl/native.c
> index 4a82c313cf71..0e748c682ee9 100644
> --- a/drivers/misc/cxl/native.c
> +++ b/drivers/misc/cxl/native.c
> @@ -897,6 +897,10 @@ int cxl_attach_dedicated_process_psl9(struct cxl_context *ctx, u64 wed, u64 amr)
> if (ctx->afu->adapter->native->sl_ops->update_dedicated_ivtes)
> afu->adapter->native->sl_ops->update_dedicated_ivtes(ctx);
>
> + ctx->elem->software_state = cpu_to_be32(CXL_PE_SOFTWARE_STATE_V);
> + /* Make sure the changes to the PE are visible to the card */
A barrier orders something vs something else. So what's the something
else in this case? Is it the afu_reset() below, what does that actually do?
> + smp_wmb();
> +
> result = cxl_ops->afu_reset(afu);
> if (result)
> return result;
cheers
^ permalink raw reply [flat|nested] 7+ messages in thread
* Re: [PATCH] cxl: Set the valid bit in PE for dedicated mode
2017-08-29 6:53 ` Michael Ellerman
@ 2017-08-30 12:04 ` Vaibhav Jain
2017-08-30 12:28 ` Michael Ellerman
0 siblings, 1 reply; 7+ messages in thread
From: Vaibhav Jain @ 2017-08-30 12:04 UTC (permalink / raw)
To: Michael Ellerman, linuxppc-dev, Frederic Barrat
Cc: Philippe Bergheaud, Alastair D'Silva, Andrew Donnellan,
Christophe Lombard
Hi Mpe,
Thanks for reviewing the patch
Michael Ellerman <mpe@ellerman.id.au> writes:
>> + ctx->elem->software_state = cpu_to_be32(CXL_PE_SOFTWARE_STATE_V);
>> + /* Make sure the changes to the PE are visible to the card */
>
> A barrier orders something vs something else. So what's the something
> else in this case? Is it the afu_reset() below, what does that actually do?
>
The issue is with call to afu_enable() after the call to afu_reset that
would start the AFU. If this load gets reordered and PSL doesnt see the
valid bit set for this structure then it will result in PSL entering a
freeze-state.
Though on second thoughts afu_enable() is grabbing a spin-lock before
doing an mmio to start the AFU that would be forcing a barrier
anyways. But since that spans the function boundary hence to be safe
have added a write barrier after populating the process element.
Lastly function is not performance critical as it will be usually called
in the life time of a process only once. So the impact smp_wmb() is
having would be minimal.
--
Vaibhav Jain <vaibhav@linux.vnet.ibm.com>
Linux Technology Center, IBM India Pvt. Ltd.
^ permalink raw reply [flat|nested] 7+ messages in thread
* Re: [PATCH] cxl: Set the valid bit in PE for dedicated mode
2017-08-30 12:04 ` Vaibhav Jain
@ 2017-08-30 12:28 ` Michael Ellerman
0 siblings, 0 replies; 7+ messages in thread
From: Michael Ellerman @ 2017-08-30 12:28 UTC (permalink / raw)
To: Vaibhav Jain, linuxppc-dev, Frederic Barrat
Cc: Philippe Bergheaud, Alastair D'Silva, Andrew Donnellan,
Christophe Lombard
Vaibhav Jain <vaibhav@linux.vnet.ibm.com> writes:
> Hi Mpe,
>
> Thanks for reviewing the patch
>
> Michael Ellerman <mpe@ellerman.id.au> writes:
>
>>> + ctx->elem->software_state = cpu_to_be32(CXL_PE_SOFTWARE_STATE_V);
>>> + /* Make sure the changes to the PE are visible to the card */
>>
>> A barrier orders something vs something else. So what's the something
>> else in this case? Is it the afu_reset() below, what does that actually do?
>>
>
> The issue is with call to afu_enable() after the call to afu_reset that
> would start the AFU. If this load gets reordered and PSL doesnt see the
> valid bit set for this structure then it will result in PSL entering a
> freeze-state.
OK, so it's ordering the store above to ctx->elem->software_state vs the
store to the AFU in afu_enable().
> Though on second thoughts afu_enable() is grabbing a spin-lock before
> doing an mmio to start the AFU that would be forcing a barrier
> anyways. But since that spans the function boundary hence to be safe
> have added a write barrier after populating the process element.
The spin lock doesn't help you, stores are allowed to leak into the
locked region.
But the MMIO is preceeded by a sync.
> Lastly function is not performance critical as it will be usually called
> in the life time of a process only once. So the impact smp_wmb() is
> having would be minimal.
Sure. Performance is not the issue, barriers are subtle so it's
important that they're well documented.
So I don't think you need the barrier, because the out_be64() will do it
for you. But if you really want to add one, I don't mind.
But, you should use wmb(), not smp_wmb(), because the ordering is still
required on non-SMP systems. And please update the comment to capture
all of the above discussion.
cheers
^ permalink raw reply [flat|nested] 7+ messages in thread
end of thread, other threads:[~2017-08-30 12:28 UTC | newest]
Thread overview: 7+ messages (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
2017-08-28 4:15 [PATCH] cxl: Set the valid bit in PE for dedicated mode Vaibhav Jain
2017-08-28 4:25 ` Andrew Donnellan
2017-08-28 8:58 ` christophe lombard
2017-08-28 9:42 ` Frederic Barrat
2017-08-29 6:53 ` Michael Ellerman
2017-08-30 12:04 ` Vaibhav Jain
2017-08-30 12:28 ` Michael Ellerman
This is an external index of several public inboxes,
see mirroring instructions on how to clone and mirror
all data and code used by this external index.