* [PATCH net-next 1/4] pds_core: Don't assign interrupt index/bound_intr to notifyq
2024-02-02 19:59 [PATCH net-next 0/4] pds_core: Various improvements/cleanups Brett Creeley
@ 2024-02-02 19:59 ` Brett Creeley
2024-02-02 19:59 ` [PATCH net-next 2/4] pds_core: Unmask adminq interrupt in work thread Brett Creeley
` (3 subsequent siblings)
4 siblings, 0 replies; 6+ messages in thread
From: Brett Creeley @ 2024-02-02 19:59 UTC (permalink / raw)
To: davem, edumazet, kuba, pabeni, netdev, linux-kernel
Cc: shannon.nelson, brett.creeley
The notifyq rides on the adminq's interrupt, so there's
no need to setup and/or access the notifyq's interrupt
index or bound_intr. The driver sets the bound_intr
using qcq->intx = -1 for the notifyq, but luckily
nothing accesses that field for notifyq. Instead of
expecting that remains the case, just clean up
the notifyq's interrupt index and bound_intr fields.
Signed-off-by: Brett Creeley <brett.creeley@amd.com>
Reviewed-by: Shannon Nelson <shannon.nelson@amd.com>
Reviewed-by: Przemek Kitszel <przemyslaw.kitszel@intel.com>
---
drivers/net/ethernet/amd/pds_core/core.c | 5 +----
drivers/net/ethernet/amd/pds_core/debugfs.c | 3 ++-
2 files changed, 3 insertions(+), 5 deletions(-)
diff --git a/drivers/net/ethernet/amd/pds_core/core.c b/drivers/net/ethernet/amd/pds_core/core.c
index 7658a7286767..41507ade3570 100644
--- a/drivers/net/ethernet/amd/pds_core/core.c
+++ b/drivers/net/ethernet/amd/pds_core/core.c
@@ -129,6 +129,7 @@ static int pdsc_qcq_intr_alloc(struct pdsc *pdsc, struct pdsc_qcq *qcq)
if (index < 0)
return index;
qcq->intx = index;
+ qcq->cq.bound_intr = &pdsc->intr_info[index];
return 0;
}
@@ -222,7 +223,6 @@ int pdsc_qcq_alloc(struct pdsc *pdsc, unsigned int type, unsigned int index,
goto err_out_free_irq;
}
- qcq->cq.bound_intr = &pdsc->intr_info[qcq->intx];
qcq->cq.num_descs = num_descs;
qcq->cq.desc_size = cq_desc_size;
qcq->cq.tail_idx = 0;
@@ -430,9 +430,6 @@ int pdsc_setup(struct pdsc *pdsc, bool init)
if (err)
goto err_out_teardown;
- /* NotifyQ rides on the AdminQ interrupt */
- pdsc->notifyqcq.intx = pdsc->adminqcq.intx;
-
/* Set up the Core with the AdminQ and NotifyQ info */
err = pdsc_core_init(pdsc);
if (err)
diff --git a/drivers/net/ethernet/amd/pds_core/debugfs.c b/drivers/net/ethernet/amd/pds_core/debugfs.c
index 4e8579ca1c8c..ba592dbeef14 100644
--- a/drivers/net/ethernet/amd/pds_core/debugfs.c
+++ b/drivers/net/ethernet/amd/pds_core/debugfs.c
@@ -109,7 +109,6 @@ void pdsc_debugfs_add_qcq(struct pdsc *pdsc, struct pdsc_qcq *qcq)
struct dentry *qcq_dentry, *q_dentry, *cq_dentry;
struct dentry *intr_dentry;
struct debugfs_regset32 *intr_ctrl_regset;
- struct pdsc_intr_info *intr = &pdsc->intr_info[qcq->intx];
struct pdsc_queue *q = &qcq->q;
struct pdsc_cq *cq = &qcq->cq;
@@ -147,6 +146,8 @@ void pdsc_debugfs_add_qcq(struct pdsc *pdsc, struct pdsc_qcq *qcq)
debugfs_create_u16("tail", 0400, cq_dentry, &cq->tail_idx);
if (qcq->flags & PDS_CORE_QCQ_F_INTR) {
+ struct pdsc_intr_info *intr = &pdsc->intr_info[qcq->intx];
+
intr_dentry = debugfs_create_dir("intr", qcq->dentry);
if (IS_ERR_OR_NULL(intr_dentry))
return;
--
2.17.1
^ permalink raw reply related [flat|nested] 6+ messages in thread* [PATCH net-next 2/4] pds_core: Unmask adminq interrupt in work thread
2024-02-02 19:59 [PATCH net-next 0/4] pds_core: Various improvements/cleanups Brett Creeley
2024-02-02 19:59 ` [PATCH net-next 1/4] pds_core: Don't assign interrupt index/bound_intr to notifyq Brett Creeley
@ 2024-02-02 19:59 ` Brett Creeley
2024-02-02 19:59 ` [PATCH net-next 3/4] pds_core: Fix up some minor issues Brett Creeley
` (2 subsequent siblings)
4 siblings, 0 replies; 6+ messages in thread
From: Brett Creeley @ 2024-02-02 19:59 UTC (permalink / raw)
To: davem, edumazet, kuba, pabeni, netdev, linux-kernel
Cc: shannon.nelson, brett.creeley
Unmasking the interrupt during the pdsc_adminq_isr
is a bit early and could cause unnecessary interrupts.
Instead always unmask after processing the adminq
and notifyq in pdsc_work_thread()->pdsc_process_adminq().
Also, since we are always unmasking, there's no need
for the local credits variable in pdsc_process_adminq().
Signed-off-by: Brett Creeley <brett.creeley@amd.com>
Reviewed-by: Shannon Nelson <shannon.nelson@amd.com>
Reviewed-by: Przemek Kitszel <przemyslaw.kitszel@intel.com>
---
drivers/net/ethernet/amd/pds_core/adminq.c | 10 +++-------
1 file changed, 3 insertions(+), 7 deletions(-)
diff --git a/drivers/net/ethernet/amd/pds_core/adminq.c b/drivers/net/ethernet/amd/pds_core/adminq.c
index ea773cfa0af6..c83a0a80d533 100644
--- a/drivers/net/ethernet/amd/pds_core/adminq.c
+++ b/drivers/net/ethernet/amd/pds_core/adminq.c
@@ -82,7 +82,6 @@ void pdsc_process_adminq(struct pdsc_qcq *qcq)
unsigned long irqflags;
int nq_work = 0;
int aq_work = 0;
- int credits;
/* Don't process AdminQ when it's not up */
if (!pdsc_adminq_inc_if_up(pdsc)) {
@@ -128,11 +127,9 @@ void pdsc_process_adminq(struct pdsc_qcq *qcq)
credits:
/* Return the interrupt credits, one for each completion */
- credits = nq_work + aq_work;
- if (credits)
- pds_core_intr_credits(&pdsc->intr_ctrl[qcq->intx],
- credits,
- PDS_CORE_INTR_CRED_REARM);
+ pds_core_intr_credits(&pdsc->intr_ctrl[qcq->intx],
+ nq_work + aq_work,
+ PDS_CORE_INTR_CRED_REARM);
refcount_dec(&pdsc->adminq_refcnt);
}
@@ -157,7 +154,6 @@ irqreturn_t pdsc_adminq_isr(int irq, void *data)
qcq = &pdsc->adminqcq;
queue_work(pdsc->wq, &qcq->work);
- pds_core_intr_mask(&pdsc->intr_ctrl[qcq->intx], PDS_CORE_INTR_MASK_CLEAR);
refcount_dec(&pdsc->adminq_refcnt);
return IRQ_HANDLED;
--
2.17.1
^ permalink raw reply related [flat|nested] 6+ messages in thread* [PATCH net-next 3/4] pds_core: Fix up some minor issues
2024-02-02 19:59 [PATCH net-next 0/4] pds_core: Various improvements/cleanups Brett Creeley
2024-02-02 19:59 ` [PATCH net-next 1/4] pds_core: Don't assign interrupt index/bound_intr to notifyq Brett Creeley
2024-02-02 19:59 ` [PATCH net-next 2/4] pds_core: Unmask adminq interrupt in work thread Brett Creeley
@ 2024-02-02 19:59 ` Brett Creeley
2024-02-02 19:59 ` [PATCH net-next 4/4] pds_core: Clean up init/uninit flows to be more readable Brett Creeley
2024-02-06 12:30 ` [PATCH net-next 0/4] pds_core: Various improvements/cleanups patchwork-bot+netdevbpf
4 siblings, 0 replies; 6+ messages in thread
From: Brett Creeley @ 2024-02-02 19:59 UTC (permalink / raw)
To: davem, edumazet, kuba, pabeni, netdev, linux-kernel
Cc: shannon.nelson, brett.creeley
Running xmastree.py against the driver found some
RCT issues, so fix them.
Also, if allocating pdsc->intr_info in pdsc_dev_init()
fails the driver still tries to free pdsc->intr_info.
Fix this by just returning -ENOMEM since there's
nothing to free at this point of failure.
Signed-off-by: Brett Creeley <brett.creeley@amd.com>
Reviewed-by: Shannon Nelson <shannon.nelson@amd.com>
---
drivers/net/ethernet/amd/pds_core/debugfs.c | 5 ++---
drivers/net/ethernet/amd/pds_core/dev.c | 6 ++----
2 files changed, 4 insertions(+), 7 deletions(-)
diff --git a/drivers/net/ethernet/amd/pds_core/debugfs.c b/drivers/net/ethernet/amd/pds_core/debugfs.c
index ba592dbeef14..6bdd02b7aa6d 100644
--- a/drivers/net/ethernet/amd/pds_core/debugfs.c
+++ b/drivers/net/ethernet/amd/pds_core/debugfs.c
@@ -32,8 +32,8 @@ void pdsc_debugfs_del_dev(struct pdsc *pdsc)
static int identity_show(struct seq_file *seq, void *v)
{
- struct pdsc *pdsc = seq->private;
struct pds_core_dev_identity *ident;
+ struct pdsc *pdsc = seq->private;
int vt;
ident = &pdsc->dev_ident;
@@ -106,8 +106,7 @@ static const struct debugfs_reg32 intr_ctrl_regs[] = {
void pdsc_debugfs_add_qcq(struct pdsc *pdsc, struct pdsc_qcq *qcq)
{
- struct dentry *qcq_dentry, *q_dentry, *cq_dentry;
- struct dentry *intr_dentry;
+ struct dentry *qcq_dentry, *q_dentry, *cq_dentry, *intr_dentry;
struct debugfs_regset32 *intr_ctrl_regset;
struct pdsc_queue *q = &qcq->q;
struct pdsc_cq *cq = &qcq->cq;
diff --git a/drivers/net/ethernet/amd/pds_core/dev.c b/drivers/net/ethernet/amd/pds_core/dev.c
index e65a1632df50..7dc102a31185 100644
--- a/drivers/net/ethernet/amd/pds_core/dev.c
+++ b/drivers/net/ethernet/amd/pds_core/dev.c
@@ -341,10 +341,8 @@ int pdsc_dev_init(struct pdsc *pdsc)
/* Get intr_info struct array for tracking */
pdsc->intr_info = kcalloc(nintrs, sizeof(*pdsc->intr_info), GFP_KERNEL);
- if (!pdsc->intr_info) {
- err = -ENOMEM;
- goto err_out;
- }
+ if (!pdsc->intr_info)
+ return -ENOMEM;
err = pci_alloc_irq_vectors(pdsc->pdev, nintrs, nintrs, PCI_IRQ_MSIX);
if (err != nintrs) {
--
2.17.1
^ permalink raw reply related [flat|nested] 6+ messages in thread* [PATCH net-next 4/4] pds_core: Clean up init/uninit flows to be more readable
2024-02-02 19:59 [PATCH net-next 0/4] pds_core: Various improvements/cleanups Brett Creeley
` (2 preceding siblings ...)
2024-02-02 19:59 ` [PATCH net-next 3/4] pds_core: Fix up some minor issues Brett Creeley
@ 2024-02-02 19:59 ` Brett Creeley
2024-02-06 12:30 ` [PATCH net-next 0/4] pds_core: Various improvements/cleanups patchwork-bot+netdevbpf
4 siblings, 0 replies; 6+ messages in thread
From: Brett Creeley @ 2024-02-02 19:59 UTC (permalink / raw)
To: davem, edumazet, kuba, pabeni, netdev, linux-kernel
Cc: shannon.nelson, brett.creeley
The setup and teardown flows are somewhat hard to follow regarding
pdsc_core_init()/pdsc_dev_init() and their corresponding teardown
flows being in pdsc_teardown(). Improve the readability by adding
new pdsc_core_uninit()/pdsc_dev_unint() functions that mirror their
init counterparts. Also, move the notify and admin qcq allocations
into pdsc_core_init(), so they can be freed in pdsc_core_uninit().
Signed-off-by: Brett Creeley <brett.creeley@amd.com>
Reviewed-by: Shannon Nelson <shannon.nelson@amd.com>
---
drivers/net/ethernet/amd/pds_core/core.c | 87 ++++++++++++------------
drivers/net/ethernet/amd/pds_core/core.h | 1 +
drivers/net/ethernet/amd/pds_core/dev.c | 16 +++++
3 files changed, 61 insertions(+), 43 deletions(-)
diff --git a/drivers/net/ethernet/amd/pds_core/core.c b/drivers/net/ethernet/amd/pds_core/core.c
index 41507ade3570..1234a4a8a4ae 100644
--- a/drivers/net/ethernet/amd/pds_core/core.c
+++ b/drivers/net/ethernet/amd/pds_core/core.c
@@ -300,6 +300,17 @@ int pdsc_qcq_alloc(struct pdsc *pdsc, unsigned int type, unsigned int index,
return err;
}
+static void pdsc_core_uninit(struct pdsc *pdsc)
+{
+ pdsc_qcq_free(pdsc, &pdsc->notifyqcq);
+ pdsc_qcq_free(pdsc, &pdsc->adminqcq);
+
+ if (pdsc->kern_dbpage) {
+ iounmap(pdsc->kern_dbpage);
+ pdsc->kern_dbpage = NULL;
+ }
+}
+
static int pdsc_core_init(struct pdsc *pdsc)
{
union pds_core_dev_comp comp = {};
@@ -310,9 +321,32 @@ static int pdsc_core_init(struct pdsc *pdsc)
struct pds_core_dev_init_data_in cidi;
u32 dbid_count;
u32 dbpage_num;
+ int numdescs;
size_t sz;
int err;
+ /* Scale the descriptor ring length based on number of CPUs and VFs */
+ numdescs = max_t(int, PDSC_ADMINQ_MIN_LENGTH, num_online_cpus());
+ numdescs += 2 * pci_sriov_get_totalvfs(pdsc->pdev);
+ numdescs = roundup_pow_of_two(numdescs);
+ err = pdsc_qcq_alloc(pdsc, PDS_CORE_QTYPE_ADMINQ, 0, "adminq",
+ PDS_CORE_QCQ_F_CORE | PDS_CORE_QCQ_F_INTR,
+ numdescs,
+ sizeof(union pds_core_adminq_cmd),
+ sizeof(union pds_core_adminq_comp),
+ 0, &pdsc->adminqcq);
+ if (err)
+ return err;
+
+ err = pdsc_qcq_alloc(pdsc, PDS_CORE_QTYPE_NOTIFYQ, 0, "notifyq",
+ PDS_CORE_QCQ_F_NOTIFYQ,
+ PDSC_NOTIFYQ_LENGTH,
+ sizeof(struct pds_core_notifyq_cmd),
+ sizeof(union pds_core_notifyq_comp),
+ 0, &pdsc->notifyqcq);
+ if (err)
+ goto err_out_uninit;
+
cidi.adminq_q_base = cpu_to_le64(pdsc->adminqcq.q_base_pa);
cidi.adminq_cq_base = cpu_to_le64(pdsc->adminqcq.cq_base_pa);
cidi.notifyq_cq_base = cpu_to_le64(pdsc->notifyqcq.cq.base_pa);
@@ -336,7 +370,7 @@ static int pdsc_core_init(struct pdsc *pdsc)
if (err) {
dev_err(pdsc->dev, "Device init command failed: %pe\n",
ERR_PTR(err));
- return err;
+ goto err_out_uninit;
}
pdsc->hw_index = le32_to_cpu(cido.core_hw_index);
@@ -346,7 +380,8 @@ static int pdsc_core_init(struct pdsc *pdsc)
pdsc->kern_dbpage = pdsc_map_dbpage(pdsc, dbpage_num);
if (!pdsc->kern_dbpage) {
dev_err(pdsc->dev, "Cannot map dbpage, aborting\n");
- return -ENOMEM;
+ err = -ENOMEM;
+ goto err_out_uninit;
}
pdsc->adminqcq.q.hw_type = cido.adminq_hw_type;
@@ -359,6 +394,10 @@ static int pdsc_core_init(struct pdsc *pdsc)
pdsc->last_eid = 0;
+ return 0;
+
+err_out_uninit:
+ pdsc_core_uninit(pdsc);
return err;
}
@@ -401,35 +440,12 @@ static int pdsc_viftypes_init(struct pdsc *pdsc)
int pdsc_setup(struct pdsc *pdsc, bool init)
{
- int numdescs;
int err;
err = pdsc_dev_init(pdsc);
if (err)
return err;
- /* Scale the descriptor ring length based on number of CPUs and VFs */
- numdescs = max_t(int, PDSC_ADMINQ_MIN_LENGTH, num_online_cpus());
- numdescs += 2 * pci_sriov_get_totalvfs(pdsc->pdev);
- numdescs = roundup_pow_of_two(numdescs);
- err = pdsc_qcq_alloc(pdsc, PDS_CORE_QTYPE_ADMINQ, 0, "adminq",
- PDS_CORE_QCQ_F_CORE | PDS_CORE_QCQ_F_INTR,
- numdescs,
- sizeof(union pds_core_adminq_cmd),
- sizeof(union pds_core_adminq_comp),
- 0, &pdsc->adminqcq);
- if (err)
- goto err_out_teardown;
-
- err = pdsc_qcq_alloc(pdsc, PDS_CORE_QTYPE_NOTIFYQ, 0, "notifyq",
- PDS_CORE_QCQ_F_NOTIFYQ,
- PDSC_NOTIFYQ_LENGTH,
- sizeof(struct pds_core_notifyq_cmd),
- sizeof(union pds_core_notifyq_comp),
- 0, &pdsc->notifyqcq);
- if (err)
- goto err_out_teardown;
-
/* Set up the Core with the AdminQ and NotifyQ info */
err = pdsc_core_init(pdsc);
if (err)
@@ -455,35 +471,20 @@ int pdsc_setup(struct pdsc *pdsc, bool init)
void pdsc_teardown(struct pdsc *pdsc, bool removing)
{
- int i;
-
if (!pdsc->pdev->is_virtfn)
pdsc_devcmd_reset(pdsc);
if (pdsc->adminqcq.work.func)
cancel_work_sync(&pdsc->adminqcq.work);
- pdsc_qcq_free(pdsc, &pdsc->notifyqcq);
- pdsc_qcq_free(pdsc, &pdsc->adminqcq);
+
+ pdsc_core_uninit(pdsc);
if (removing) {
kfree(pdsc->viftype_status);
pdsc->viftype_status = NULL;
}
- if (pdsc->intr_info) {
- for (i = 0; i < pdsc->nintrs; i++)
- pdsc_intr_free(pdsc, i);
-
- kfree(pdsc->intr_info);
- pdsc->intr_info = NULL;
- pdsc->nintrs = 0;
- }
-
- if (pdsc->kern_dbpage) {
- iounmap(pdsc->kern_dbpage);
- pdsc->kern_dbpage = NULL;
- }
+ pdsc_dev_uninit(pdsc);
- pci_free_irq_vectors(pdsc->pdev);
set_bit(PDSC_S_FW_DEAD, &pdsc->state);
}
diff --git a/drivers/net/ethernet/amd/pds_core/core.h b/drivers/net/ethernet/amd/pds_core/core.h
index 110c4b826b22..3468a63f5ae9 100644
--- a/drivers/net/ethernet/amd/pds_core/core.h
+++ b/drivers/net/ethernet/amd/pds_core/core.h
@@ -282,6 +282,7 @@ int pdsc_devcmd_locked(struct pdsc *pdsc, union pds_core_dev_cmd *cmd,
int pdsc_devcmd_init(struct pdsc *pdsc);
int pdsc_devcmd_reset(struct pdsc *pdsc);
int pdsc_dev_init(struct pdsc *pdsc);
+void pdsc_dev_uninit(struct pdsc *pdsc);
void pdsc_reset_prepare(struct pci_dev *pdev);
void pdsc_reset_done(struct pci_dev *pdev);
diff --git a/drivers/net/ethernet/amd/pds_core/dev.c b/drivers/net/ethernet/amd/pds_core/dev.c
index 7dc102a31185..e494e1298dc9 100644
--- a/drivers/net/ethernet/amd/pds_core/dev.c
+++ b/drivers/net/ethernet/amd/pds_core/dev.c
@@ -316,6 +316,22 @@ static int pdsc_identify(struct pdsc *pdsc)
return 0;
}
+void pdsc_dev_uninit(struct pdsc *pdsc)
+{
+ if (pdsc->intr_info) {
+ int i;
+
+ for (i = 0; i < pdsc->nintrs; i++)
+ pdsc_intr_free(pdsc, i);
+
+ kfree(pdsc->intr_info);
+ pdsc->intr_info = NULL;
+ pdsc->nintrs = 0;
+ }
+
+ pci_free_irq_vectors(pdsc->pdev);
+}
+
int pdsc_dev_init(struct pdsc *pdsc)
{
unsigned int nintrs;
--
2.17.1
^ permalink raw reply related [flat|nested] 6+ messages in thread* Re: [PATCH net-next 0/4] pds_core: Various improvements/cleanups
2024-02-02 19:59 [PATCH net-next 0/4] pds_core: Various improvements/cleanups Brett Creeley
` (3 preceding siblings ...)
2024-02-02 19:59 ` [PATCH net-next 4/4] pds_core: Clean up init/uninit flows to be more readable Brett Creeley
@ 2024-02-06 12:30 ` patchwork-bot+netdevbpf
4 siblings, 0 replies; 6+ messages in thread
From: patchwork-bot+netdevbpf @ 2024-02-06 12:30 UTC (permalink / raw)
To: Brett Creeley
Cc: davem, edumazet, kuba, pabeni, netdev, linux-kernel,
shannon.nelson
Hello:
This series was applied to netdev/net-next.git (main)
by Paolo Abeni <pabeni@redhat.com>:
On Fri, 2 Feb 2024 11:59:07 -0800 you wrote:
> This series contains various improvements and cleanups for the
> pds_core driver. These patches were originally part of the following
> net-next series:
>
> https://lore.kernel.org/netdev/20240126174255.17052-1-brett.creeley@amd.com/
>
> However, some of the patches from the above series were actually fixes,
> so they were pushed and accepted to net. That series can be found here:
>
> [...]
Here is the summary with links:
- [net-next,1/4] pds_core: Don't assign interrupt index/bound_intr to notifyq
https://git.kernel.org/netdev/net-next/c/02daffa903e6
- [net-next,2/4] pds_core: Unmask adminq interrupt in work thread
https://git.kernel.org/netdev/net-next/c/bca10f2c2518
- [net-next,3/4] pds_core: Fix up some minor issues
https://git.kernel.org/netdev/net-next/c/247c4ed03321
- [net-next,4/4] pds_core: Clean up init/uninit flows to be more readable
https://git.kernel.org/netdev/net-next/c/792d36ccc163
You are awesome, thank you!
--
Deet-doot-dot, I am a bot.
https://korg.docs.kernel.org/patchwork/pwbot.html
^ permalink raw reply [flat|nested] 6+ messages in thread