[PATCH 0/4] drivers/net/ethernet: replace __get_free_pages() with kmalloc()

[PATCH 0/4] drivers/net/ethernet: replace __get_free_pages() with kmalloc()

[Mike Rapoport (Microsoft)]

This is a (small) part of larger work of replacing page allocator calls
with kmalloc.

My initial intention a few month ago was to remove ugly casts [1], but then
willy pointed out that Linus objected to something like this [2] and it
looks like more than a decade old technical debt.

Largely, anything that doesn't need struct page (or a memdesc in the
future) should just use kmalloc() or kvmalloc() to allocate memory.
kmalloc() guarantees alignment, physical contiguity and working
virt_to_phys() and beside nicer API that returns void * on alloc and
doesn't require to know the allocation size on free, kmalloc() provides
better debugging capabilities than page allocator.

Another thing is that touching these allocation sites gives the reviewers
opportunity to see if a PAGE_SIZE buffer is actually needed or maybe
another size is appropriate.

For larger allocations that don't need physically contiguous memory
kvmalloc() can be a better option that __get_free_pages() because under
memory pressure it's is easier to allocate several order-0 pages than a
physically contiguous chunk with the same number of pages.

And last, but not least, removing needless calls to page allocator should
help with memdesc (aka project folio) conversion. There will be way less
places to audit to see if the user was actually using struct page.

Also in git:
https://git.kernel.org/pub/scm/linux/kernel/git/rppt/linux.git gfp-to-kmalloc/drivers-net-ethernet

[1] https://lore.kernel.org/all/20251018093002.3660549-1-rppt@kernel.org/
[2] https://lore.kernel.org/all/CA+55aFwp4iy4rtX2gE2WjBGFL=NxMVnoFeHqYa2j1dYOMMGqxg@mail.gmail.com/

---
v2 changes:
- split out ethernet drivers from a larger set 

v1: https://patch.msgid.link/20260630-b4-drivers-net-v1-0-672162a91f37@kernel.org

---
Mike Rapoport (Microsoft) (4):
      bnx2x: use kzalloc() to allocate mac filtering list
      ice: use kzalloc() to allocate staging buffer for reading from GNSS
      sfc/siena: use kmalloc() to allocate logging buffer
      sfc: use kmalloc() to allocate logging buffer

 drivers/net/ethernet/broadcom/bnx2x/bnx2x_sp.c | 6 +++---
 drivers/net/ethernet/intel/ice/ice_gnss.c      | 5 +++--
 drivers/net/ethernet/sfc/mcdi.c                | 7 ++++---
 drivers/net/ethernet/sfc/siena/mcdi.c          | 7 ++++---
 4 files changed, 14 insertions(+), 11 deletions(-)
---
base-commit: dc59e4fea9d83f03bad6bddf3fa2e52491777482
change-id: 20260630-b4-drivers-ethernet-b5e085b98ab1

Best regards,
--  
Sincerely yours,
Mike.

[PATCH 1/4] bnx2x: use kzalloc() to allocate mac filtering list

[Mike Rapoport (Microsoft)]

bnx2x_mcast_enqueue_cmd() allocates memory for mac filtering list using
__get_free_pages().

This memory can be allocated with kzalloc() as there's nothing special
about it to go directly to the page allocator.

kmalloc() provides a better API that does not require ugly casts and
kfree() does not need to know the size of the freed object.

Performance difference between kmalloc() and __get_free_pages() is not
measurable as both allocators take an object/page from a per-CPU list for
fast path allocations.

For the slow path the performance is anyway determined by the amount of
reclaim involved rather than by what allocator is used.

Replace use of __get_free_page() with kzalloc() and free_page() with
kfree().

Link: https://lore.kernel.org/all/635405e4-9423-4a25-a6e7-e03c8ea0bcbe@redhat.com
Signed-off-by: Mike Rapoport (Microsoft) <rppt@kernel.org>
---
 drivers/net/ethernet/broadcom/bnx2x/bnx2x_sp.c | 6 +++---
 1 file changed, 3 insertions(+), 3 deletions(-)

diff --git a/drivers/net/ethernet/broadcom/bnx2x/bnx2x_sp.c b/drivers/net/ethernet/broadcom/bnx2x/bnx2x_sp.c
index 07a908a2c72f..d560524d317d 100644
--- a/drivers/net/ethernet/broadcom/bnx2x/bnx2x_sp.c
+++ b/drivers/net/ethernet/broadcom/bnx2x/bnx2x_sp.c
@@ -26,6 +26,7 @@
 #include <linux/netdevice.h>
 #include <linux/etherdevice.h>
 #include <linux/crc32c.h>
+#include <linux/slab.h>
 #include "bnx2x.h"
 #include "bnx2x_cmn.h"
 #include "bnx2x_sp.h"
@@ -2664,7 +2665,7 @@ static void bnx2x_free_groups(struct list_head *mcast_group_list)
 				      struct bnx2x_mcast_elem_group,
 				      mcast_group_link);
 		list_del(&current_mcast_group->mcast_group_link);
-		free_page((unsigned long)current_mcast_group);
+		kfree(current_mcast_group);
 	}
 }
 
@@ -2713,8 +2714,7 @@ static int bnx2x_mcast_enqueue_cmd(struct bnx2x *bp,
 				total_elems = BNX2X_MCAST_BINS_NUM;
 		}
 		while (total_elems > 0) {
-			elem_group = (struct bnx2x_mcast_elem_group *)
-				     __get_free_page(GFP_ATOMIC | __GFP_ZERO);
+			elem_group = kzalloc(PAGE_SIZE, GFP_ATOMIC);
 			if (!elem_group) {
 				bnx2x_free_groups(&new_cmd->group_head);
 				kfree(new_cmd);

-- 
2.53.0

Re: [PATCH 1/4] bnx2x: use kzalloc() to allocate mac filtering list

[Przemek Kitszel]

> @@ -2713,8 +2714,7 @@ static int bnx2x_mcast_enqueue_cmd(struct bnx2x *bp,
>   				total_elems = BNX2X_MCAST_BINS_NUM;
>   		}
>   		while (total_elems > 0) {
> -			elem_group = (struct bnx2x_mcast_elem_group *)
> -				     __get_free_page(GFP_ATOMIC | __GFP_ZERO);
> +			elem_group = kzalloc(PAGE_SIZE, GFP_ATOMIC);

what is the current rule of thumb for kzalloc vs kvzalloc size under
GFP_ATOMIC?

>   			if (!elem_group) {
>   				bnx2x_free_groups(&new_cmd->group_head);
>   				kfree(new_cmd);
> 

Re: [PATCH 1/4] bnx2x: use kzalloc() to allocate mac filtering list

[Mike Rapoport]

On Thu, Jul 02, 2026 at 03:52:13PM +0200, Przemek Kitszel wrote:
> 
> > @@ -2713,8 +2714,7 @@ static int bnx2x_mcast_enqueue_cmd(struct bnx2x *bp,
> >   				total_elems = BNX2X_MCAST_BINS_NUM;
> >   		}
> >   		while (total_elems > 0) {
> > -			elem_group = (struct bnx2x_mcast_elem_group *)
> > -				     __get_free_page(GFP_ATOMIC | __GFP_ZERO);
> > +			elem_group = kzalloc(PAGE_SIZE, GFP_ATOMIC);
> 
> what is the current rule of thumb for kzalloc vs kvzalloc size under
> GFP_ATOMIC?

kvzalloc() with GFP_ATOMIC is kzalloc():

	/*
	 * vmalloc uses GFP_KERNEL for some internal allocations (e.g page tables)
	 * so the given set of flags has to be compatible.
	 */
	if ((flags & GFP_KERNEL) != GFP_KERNEL)
		return kmalloc_node(size, flags, node);

so it's only physically contiguous allocations.

 
> >   			if (!elem_group) {
> >   				bnx2x_free_groups(&new_cmd->group_head);
> >   				kfree(new_cmd);
> > 
> 

-- 
Sincerely yours,
Mike.

[PATCH 2/4] ice: use kzalloc() to allocate staging buffer for reading from GNSS

[Mike Rapoport (Microsoft)]

ice_gnss_read() uses get_zeroed_page() to  allocate a staging buffer for
reading GNSS module data via I2C bus.

This buffer can be allocated with kmalloc() as there's nothing special
about it to go directly to the page allocator.

kmalloc() provides a better API that does not require ugly casts and
kfree() does not need to know the size of the freed object.

Performance difference between kmalloc() and __get_free_pages() is not
measurable as both allocators take an object/page from a per-CPU list for
fast path allocations.

For the slow path the performance is anyway determined by the amount of
reclaim involved rather than by what allocator is used.

Replace use of get_zeroed_page() with kzalloc() and free_page() with
kfree().

Link: https://lore.kernel.org/all/635405e4-9423-4a25-a6e7-e03c8ea0bcbe@redhat.com
Signed-off-by: Mike Rapoport (Microsoft) <rppt@kernel.org>
---
 drivers/net/ethernet/intel/ice/ice_gnss.c | 5 +++--
 1 file changed, 3 insertions(+), 2 deletions(-)

diff --git a/drivers/net/ethernet/intel/ice/ice_gnss.c b/drivers/net/ethernet/intel/ice/ice_gnss.c
index 8fd954f1ebd6..7d21c3417b0b 100644
--- a/drivers/net/ethernet/intel/ice/ice_gnss.c
+++ b/drivers/net/ethernet/intel/ice/ice_gnss.c
@@ -2,6 +2,7 @@
 /* Copyright (C) 2021-2022, Intel Corporation. */
 
 #include "ice.h"
+#include <linux/slab.h>
 #include "ice_lib.h"
 
 /**
@@ -124,7 +125,7 @@ static void ice_gnss_read(struct kthread_work *work)
 
 	data_len = min_t(typeof(data_len), data_len, PAGE_SIZE);
 
-	buf = (char *)get_zeroed_page(GFP_KERNEL);
+	buf = kzalloc(PAGE_SIZE, GFP_KERNEL);
 	if (!buf) {
 		err = -ENOMEM;
 		goto requeue;
@@ -151,7 +152,7 @@ static void ice_gnss_read(struct kthread_work *work)
 			 count, i);
 	delay = ICE_GNSS_TIMER_DELAY_TIME;
 free_buf:
-	free_page((unsigned long)buf);
+	kfree(buf);
 requeue:
 	kthread_queue_delayed_work(gnss->kworker, &gnss->read_work, delay);
 	if (err)

-- 
2.53.0

Re: [PATCH 2/4] ice: use kzalloc() to allocate staging buffer for reading from GNSS

[Przemek Kitszel]

On 7/1/26 15:57, Mike Rapoport (Microsoft) wrote:
> ice_gnss_read() uses get_zeroed_page() to  allocate a staging buffer for
> reading GNSS module data via I2C bus.
> 
> This buffer can be allocated with kmalloc() as there's nothing special
> about it to go directly to the page allocator.
> 
> kmalloc() provides a better API that does not require ugly casts and
> kfree() does not need to know the size of the freed object.
> 
> Performance difference between kmalloc() and __get_free_pages() is not
> measurable as both allocators take an object/page from a per-CPU list for
> fast path allocations.
> 
> For the slow path the performance is anyway determined by the amount of
> reclaim involved rather than by what allocator is used.
> 
> Replace use of get_zeroed_page() with kzalloc() and free_page() with
> kfree().
> 
> Link: https://lore.kernel.org/all/635405e4-9423-4a25-a6e7-e03c8ea0bcbe@redhat.com
> Signed-off-by: Mike Rapoport (Microsoft) <rppt@kernel.org>
> ---
>   drivers/net/ethernet/intel/ice/ice_gnss.c | 5 +++--
>   1 file changed, 3 insertions(+), 2 deletions(-)
> 
> diff --git a/drivers/net/ethernet/intel/ice/ice_gnss.c b/drivers/net/ethernet/intel/ice/ice_gnss.c
> index 8fd954f1ebd6..7d21c3417b0b 100644
> --- a/drivers/net/ethernet/intel/ice/ice_gnss.c
> +++ b/drivers/net/ethernet/intel/ice/ice_gnss.c
> @@ -2,6 +2,7 @@
>   /* Copyright (C) 2021-2022, Intel Corporation. */
>   
>   #include "ice.h"
> +#include <linux/slab.h>
>   #include "ice_lib.h"
>   
>   /**
> @@ -124,7 +125,7 @@ static void ice_gnss_read(struct kthread_work *work)
>   
>   	data_len = min_t(typeof(data_len), data_len, PAGE_SIZE);
>   
> -	buf = (char *)get_zeroed_page(GFP_KERNEL);
> +	buf = kzalloc(PAGE_SIZE, GFP_KERNEL);

nit:
from the code it is clear that we read at most a page, and @data_len
stores the actual amount needed

comment:
I don't know why we limit to a page, it's outside of the scope of this
series, but likely you have removed the limit (which will go into the
loop - single AQ call is likely limited by a PAGE too).

Reviewed-by: Przemek Kitszel <przemyslaw.kitszel@intel.com>

>   	if (!buf) {
>   		err = -ENOMEM;
>   		goto requeue;
> @@ -151,7 +152,7 @@ static void ice_gnss_read(struct kthread_work *work)
>   			 count, i);
>   	delay = ICE_GNSS_TIMER_DELAY_TIME;
>   free_buf:
> -	free_page((unsigned long)buf);
> +	kfree(buf);
>   requeue:
>   	kthread_queue_delayed_work(gnss->kworker, &gnss->read_work, delay);
>   	if (err)
> 

Re: [PATCH 2/4] ice: use kzalloc() to allocate staging buffer for reading from GNSS

[Mike Rapoport]

On Thu, Jul 02, 2026 at 03:49:36PM +0200, Przemek Kitszel wrote:
> On 7/1/26 15:57, Mike Rapoport (Microsoft) wrote:
> > ice_gnss_read() uses get_zeroed_page() to  allocate a staging buffer for
> > reading GNSS module data via I2C bus.
> > 
> > This buffer can be allocated with kmalloc() as there's nothing special
> > about it to go directly to the page allocator.
> > 
> > kmalloc() provides a better API that does not require ugly casts and
> > kfree() does not need to know the size of the freed object.
> > 
> > Performance difference between kmalloc() and __get_free_pages() is not
> > measurable as both allocators take an object/page from a per-CPU list for
> > fast path allocations.
> > 
> > For the slow path the performance is anyway determined by the amount of
> > reclaim involved rather than by what allocator is used.
> > 
> > Replace use of get_zeroed_page() with kzalloc() and free_page() with
> > kfree().
> > 
> > Link: https://lore.kernel.org/all/635405e4-9423-4a25-a6e7-e03c8ea0bcbe@redhat.com
> > Signed-off-by: Mike Rapoport (Microsoft) <rppt@kernel.org>
> > ---
> >   drivers/net/ethernet/intel/ice/ice_gnss.c | 5 +++--
> >   1 file changed, 3 insertions(+), 2 deletions(-)
> > 
> > diff --git a/drivers/net/ethernet/intel/ice/ice_gnss.c b/drivers/net/ethernet/intel/ice/ice_gnss.c
> > index 8fd954f1ebd6..7d21c3417b0b 100644
> > --- a/drivers/net/ethernet/intel/ice/ice_gnss.c
> > +++ b/drivers/net/ethernet/intel/ice/ice_gnss.c
> > @@ -2,6 +2,7 @@
> >   /* Copyright (C) 2021-2022, Intel Corporation. */
> >   #include "ice.h"
> > +#include <linux/slab.h>
> >   #include "ice_lib.h"
> >   /**
> > @@ -124,7 +125,7 @@ static void ice_gnss_read(struct kthread_work *work)
> >   	data_len = min_t(typeof(data_len), data_len, PAGE_SIZE);
> > -	buf = (char *)get_zeroed_page(GFP_KERNEL);
> > +	buf = kzalloc(PAGE_SIZE, GFP_KERNEL);
> 
> nit:
> from the code it is clear that we read at most a page, and @data_len
> stores the actual amount needed
> 
> comment:
> I don't know why we limit to a page, it's outside of the scope of this
> series, but likely you have removed the limit (which will go into the
> loop - single AQ call is likely limited by a PAGE too).

Not sure I follow how the limit changed. buf remains PAGE_SIZE regardless
of allocation API.
 
> Reviewed-by: Przemek Kitszel <przemyslaw.kitszel@intel.com>

Thanks!

-- 
Sincerely yours,
Mike.

[PATCH 3/4] sfc/siena: use kmalloc() to allocate logging buffer

[Mike Rapoport (Microsoft)]

efx_siena_mcdi_init() allocates a logging buffer for MCDI firmware
communication diagnostics.

This buffer can be allocated with kmalloc() as there's nothing special
about it to go directly to the page allocator.

kmalloc() provides a better API that does not require ugly casts and
kfree() does not need to know the size of the freed object.

Performance difference between kmalloc() and __get_free_pages() is not
measurable as both allocators take an object/page from a per-CPU list for
fast path allocations.

For the slow path the performance is anyway determined by the amount of
reclaim involved rather than by what allocator is used.

Replace use of __get_free_page() with kmalloc() and free_page() with
kfree().

Link: https://lore.kernel.org/all/635405e4-9423-4a25-a6e7-e03c8ea0bcbe@redhat.com
Signed-off-by: Mike Rapoport (Microsoft) <rppt@kernel.org>
---
 drivers/net/ethernet/sfc/siena/mcdi.c | 7 ++++---
 1 file changed, 4 insertions(+), 3 deletions(-)

diff --git a/drivers/net/ethernet/sfc/siena/mcdi.c b/drivers/net/ethernet/sfc/siena/mcdi.c
index 4d0d6bd5d3d1..048c1e6017c0 100644
--- a/drivers/net/ethernet/sfc/siena/mcdi.c
+++ b/drivers/net/ethernet/sfc/siena/mcdi.c
@@ -7,6 +7,7 @@
 #include <linux/delay.h>
 #include <linux/moduleparam.h>
 #include <linux/atomic.h>
+#include <linux/slab.h>
 #include "net_driver.h"
 #include "nic.h"
 #include "io.h"
@@ -73,7 +74,7 @@ int efx_siena_mcdi_init(struct efx_nic *efx)
 	mcdi->efx = efx;
 #ifdef CONFIG_SFC_SIENA_MCDI_LOGGING
 	/* consuming code assumes buffer is page-sized */
-	mcdi->logging_buffer = (char *)__get_free_page(GFP_KERNEL);
+	mcdi->logging_buffer = kmalloc(PAGE_SIZE, GFP_KERNEL);
 	if (!mcdi->logging_buffer)
 		goto fail1;
 	mcdi->logging_enabled = efx_siena_mcdi_logging_default;
@@ -116,7 +117,7 @@ int efx_siena_mcdi_init(struct efx_nic *efx)
 	return 0;
 fail2:
 #ifdef CONFIG_SFC_SIENA_MCDI_LOGGING
-	free_page((unsigned long)mcdi->logging_buffer);
+	kfree(mcdi->logging_buffer);
 fail1:
 #endif
 	kfree(efx->mcdi);
@@ -142,7 +143,7 @@ void efx_siena_mcdi_fini(struct efx_nic *efx)
 		return;
 
 #ifdef CONFIG_SFC_SIENA_MCDI_LOGGING
-	free_page((unsigned long)efx->mcdi->iface.logging_buffer);
+	kfree(efx->mcdi->iface.logging_buffer);
 #endif
 
 	kfree(efx->mcdi);

-- 
2.53.0

Re: [PATCH 3/4] sfc/siena: use kmalloc() to allocate logging buffer

[Edward Cree]

On 01/07/2026 14:57, Mike Rapoport (Microsoft) wrote:
> efx_siena_mcdi_init() allocates a logging buffer for MCDI firmware
> communication diagnostics.
> 
> This buffer can be allocated with kmalloc() as there's nothing special
> about it to go directly to the page allocator.
> 
> kmalloc() provides a better API that does not require ugly casts and
> kfree() does not need to know the size of the freed object.
> 
> Performance difference between kmalloc() and __get_free_pages() is not
> measurable as both allocators take an object/page from a per-CPU list for
> fast path allocations.
> 
> For the slow path the performance is anyway determined by the amount of
> reclaim involved rather than by what allocator is used.
> 
> Replace use of __get_free_page() with kmalloc() and free_page() with
> kfree().
> 
> Link: https://lore.kernel.org/all/635405e4-9423-4a25-a6e7-e03c8ea0bcbe@redhat.com
> Signed-off-by: Mike Rapoport (Microsoft) <rppt@kernel.org>

Reviewed-by: Edward Cree <ecree.xilinx@gmail.com>

(resending since I hit 'reply' instead of 'reply all' the first time)

> ---
>  drivers/net/ethernet/sfc/siena/mcdi.c | 7 ++++---
>  1 file changed, 4 insertions(+), 3 deletions(-)
> 
> diff --git a/drivers/net/ethernet/sfc/siena/mcdi.c b/drivers/net/ethernet/sfc/siena/mcdi.c
> index 4d0d6bd5d3d1..048c1e6017c0 100644
> --- a/drivers/net/ethernet/sfc/siena/mcdi.c
> +++ b/drivers/net/ethernet/sfc/siena/mcdi.c
> @@ -7,6 +7,7 @@
>  #include <linux/delay.h>
>  #include <linux/moduleparam.h>
>  #include <linux/atomic.h>
> +#include <linux/slab.h>
>  #include "net_driver.h"
>  #include "nic.h"
>  #include "io.h"
> @@ -73,7 +74,7 @@ int efx_siena_mcdi_init(struct efx_nic *efx)
>  	mcdi->efx = efx;
>  #ifdef CONFIG_SFC_SIENA_MCDI_LOGGING
>  	/* consuming code assumes buffer is page-sized */
> -	mcdi->logging_buffer = (char *)__get_free_page(GFP_KERNEL);
> +	mcdi->logging_buffer = kmalloc(PAGE_SIZE, GFP_KERNEL);
>  	if (!mcdi->logging_buffer)
>  		goto fail1;
>  	mcdi->logging_enabled = efx_siena_mcdi_logging_default;
> @@ -116,7 +117,7 @@ int efx_siena_mcdi_init(struct efx_nic *efx)
>  	return 0;
>  fail2:
>  #ifdef CONFIG_SFC_SIENA_MCDI_LOGGING
> -	free_page((unsigned long)mcdi->logging_buffer);
> +	kfree(mcdi->logging_buffer);
>  fail1:
>  #endif
>  	kfree(efx->mcdi);
> @@ -142,7 +143,7 @@ void efx_siena_mcdi_fini(struct efx_nic *efx)
>  		return;
>  
>  #ifdef CONFIG_SFC_SIENA_MCDI_LOGGING
> -	free_page((unsigned long)efx->mcdi->iface.logging_buffer);
> +	kfree(efx->mcdi->iface.logging_buffer);
>  #endif
>  
>  	kfree(efx->mcdi);
> 

[PATCH 4/4] sfc: use kmalloc() to allocate logging buffer

[Mike Rapoport (Microsoft)]

efx_mcdi_init() allocates a logging buffer for MCDI firmware
communication diagnostics.

This buffer can be allocated with kmalloc() as there's nothing special
about it to go directly to the page allocator.

kmalloc() provides a better API that does not require ugly casts and
kfree() does not need to know the size of the freed object.

Performance difference between kmalloc() and __get_free_pages() is not
measurable as both allocators take an object/page from a per-CPU list for
fast path allocations.

For the slow path the performance is anyway determined by the amount of
reclaim involved rather than by what allocator is used.

Replace use of __get_free_page() with kmalloc() and free_page() with
kfree().

Link: https://lore.kernel.org/all/635405e4-9423-4a25-a6e7-e03c8ea0bcbe@redhat.com
Signed-off-by: Mike Rapoport (Microsoft) <rppt@kernel.org>
---
 drivers/net/ethernet/sfc/mcdi.c | 7 ++++---
 1 file changed, 4 insertions(+), 3 deletions(-)

diff --git a/drivers/net/ethernet/sfc/mcdi.c b/drivers/net/ethernet/sfc/mcdi.c
index e65db9b70724..b806d3d90c42 100644
--- a/drivers/net/ethernet/sfc/mcdi.c
+++ b/drivers/net/ethernet/sfc/mcdi.c
@@ -7,6 +7,7 @@
 #include <linux/delay.h>
 #include <linux/moduleparam.h>
 #include <linux/atomic.h>
+#include <linux/slab.h>
 #include "net_driver.h"
 #include "nic.h"
 #include "io.h"
@@ -71,7 +72,7 @@ int efx_mcdi_init(struct efx_nic *efx)
 	mcdi->efx = efx;
 #ifdef CONFIG_SFC_MCDI_LOGGING
 	/* consuming code assumes buffer is page-sized */
-	mcdi->logging_buffer = (char *)__get_free_page(GFP_KERNEL);
+	mcdi->logging_buffer = kmalloc(PAGE_SIZE, GFP_KERNEL);
 	if (!mcdi->logging_buffer)
 		goto fail1;
 	mcdi->logging_enabled = mcdi_logging_default;
@@ -112,7 +113,7 @@ int efx_mcdi_init(struct efx_nic *efx)
 	return 0;
 fail2:
 #ifdef CONFIG_SFC_MCDI_LOGGING
-	free_page((unsigned long)mcdi->logging_buffer);
+	kfree(mcdi->logging_buffer);
 fail1:
 #endif
 	kfree(efx->mcdi);
@@ -138,7 +139,7 @@ void efx_mcdi_fini(struct efx_nic *efx)
 		return;
 
 #ifdef CONFIG_SFC_MCDI_LOGGING
-	free_page((unsigned long)efx->mcdi->iface.logging_buffer);
+	kfree(efx->mcdi->iface.logging_buffer);
 #endif
 
 	kfree(efx->mcdi);

-- 
2.53.0

Re: [PATCH 4/4] sfc: use kmalloc() to allocate logging buffer

[Edward Cree]

On 01/07/2026 14:57, Mike Rapoport (Microsoft) wrote:
> efx_mcdi_init() allocates a logging buffer for MCDI firmware
> communication diagnostics.
> 
> This buffer can be allocated with kmalloc() as there's nothing special
> about it to go directly to the page allocator.
> 
> kmalloc() provides a better API that does not require ugly casts and
> kfree() does not need to know the size of the freed object.
> 
> Performance difference between kmalloc() and __get_free_pages() is not
> measurable as both allocators take an object/page from a per-CPU list for
> fast path allocations.
> 
> For the slow path the performance is anyway determined by the amount of
> reclaim involved rather than by what allocator is used.
> 
> Replace use of __get_free_page() with kmalloc() and free_page() with
> kfree().
> 
> Link: https://lore.kernel.org/all/635405e4-9423-4a25-a6e7-e03c8ea0bcbe@redhat.com
> Signed-off-by: Mike Rapoport (Microsoft) <rppt@kernel.org>

Reviewed-by: Edward Cree <ecree.xilinx@gmail.com>

> ---
>  drivers/net/ethernet/sfc/mcdi.c | 7 ++++---
>  1 file changed, 4 insertions(+), 3 deletions(-)
> 
> diff --git a/drivers/net/ethernet/sfc/mcdi.c b/drivers/net/ethernet/sfc/mcdi.c
> index e65db9b70724..b806d3d90c42 100644
> --- a/drivers/net/ethernet/sfc/mcdi.c
> +++ b/drivers/net/ethernet/sfc/mcdi.c
> @@ -7,6 +7,7 @@
>  #include <linux/delay.h>
>  #include <linux/moduleparam.h>
>  #include <linux/atomic.h>
> +#include <linux/slab.h>
>  #include "net_driver.h"
>  #include "nic.h"
>  #include "io.h"
> @@ -71,7 +72,7 @@ int efx_mcdi_init(struct efx_nic *efx)
>  	mcdi->efx = efx;
>  #ifdef CONFIG_SFC_MCDI_LOGGING
>  	/* consuming code assumes buffer is page-sized */
> -	mcdi->logging_buffer = (char *)__get_free_page(GFP_KERNEL);
> +	mcdi->logging_buffer = kmalloc(PAGE_SIZE, GFP_KERNEL);
>  	if (!mcdi->logging_buffer)
>  		goto fail1;
>  	mcdi->logging_enabled = mcdi_logging_default;
> @@ -112,7 +113,7 @@ int efx_mcdi_init(struct efx_nic *efx)
>  	return 0;
>  fail2:
>  #ifdef CONFIG_SFC_MCDI_LOGGING
> -	free_page((unsigned long)mcdi->logging_buffer);
> +	kfree(mcdi->logging_buffer);
>  fail1:
>  #endif
>  	kfree(efx->mcdi);
> @@ -138,7 +139,7 @@ void efx_mcdi_fini(struct efx_nic *efx)
>  		return;
>  
>  #ifdef CONFIG_SFC_MCDI_LOGGING
> -	free_page((unsigned long)efx->mcdi->iface.logging_buffer);
> +	kfree(efx->mcdi->iface.logging_buffer);
>  #endif
>  
>  	kfree(efx->mcdi);
>