Re: [PATCH RESEND 0/3] ice: use better bitmap API

[Yury Norov <ynorov@nvidia.com>]

On Wed, Feb 25, 2026 at 01:02:35PM +0100, Przemek Kitszel wrote:
> On 2/25/26 01:00, Yury Norov wrote:
> > Use better bitmap API where appropriate.
> > 
> > Original series:
> > 
> > https://lore.kernel.org/all/20251223162303.434659-1-yury.norov@gmail.com/
> > 
> > RESEND: rebase on top of v7.0-rc1
> > 
> > Yury Norov (3):
> >    bitmap: introduce bitmap_weighted_xor()
> >    ice: use bitmap_weighted_xor() in ice_find_free_recp_res_idx()
> 
> Thank you for working on better API.
> Do you want this to go through intel, then netdev, tree?
> Likely that would slow any future user, if you have already one in mind.
>
> >    ice: use bitmap_empty() in ice_vf_has_no_qs_ena
> 
> For unrelated reasons I have one series that will possibly conflict
> with your patch. Would prefer to not wait/block each other, so will
> be best if we (Tony) just take this one (in case you will proceed
> with the first two patches via your tree)

Another trivial user (recently added) for bitmap_weighted_xx() is
xe_pagefault_queue_init() in drivers/gpu/drm/xe/xe_pagefault.c, but
it's again the Intel's driver, so it will go with your tree anyways.
 
I'm not aware of others, so I believe Tony's tree is the best route.

> >   drivers/net/ethernet/intel/ice/ice_switch.c |  4 +---
> >   drivers/net/ethernet/intel/ice/ice_vf_lib.c |  4 ++--
> >   include/linux/bitmap.h                      | 14 ++++++++++++++
> >   lib/bitmap.c                                |  7 +++++++
> >   4 files changed, 24 insertions(+), 5 deletions(-)

 + Matthew Brost

Attaching a patch for xe_pagefault_queue_init(). Feel free to append it
to this series, or let me know if it's better to send it separately.

Thanks,
Yury

From 21804f4ae1674aa166e3566fa898996806ebd3e3 Mon Sep 17 00:00:00 2001
From: Yury Norov <ynorov@nvidia.com>
Date: Wed, 25 Feb 2026 10:02:22 -0500
Subject: [PATCH] drm/xe: switch xe_pagefault_queue_init() to using
 bitmap_weighted_or()

The function calls bitmap_or() immediately followed by bitmap_weight().
Switch to using the dedicated bitmap_weighted_or() and save one bitmap
traverse.

Signed-off-by: Yury Norov <ynorov@nvidia.com>
---
 drivers/gpu/drm/xe/xe_pagefault.c | 3 +--
 1 file changed, 1 insertion(+), 2 deletions(-)

diff --git a/drivers/gpu/drm/xe/xe_pagefault.c b/drivers/gpu/drm/xe/xe_pagefault.c
index 6bee53d6ffc3..c4ce3cfe2164 100644
--- a/drivers/gpu/drm/xe/xe_pagefault.c
+++ b/drivers/gpu/drm/xe/xe_pagefault.c
@@ -285,10 +285,9 @@ static int xe_pagefault_queue_init(struct xe_device *xe,
 		xe_dss_mask_t all_dss;
 		int num_dss, num_eus;
 
-		bitmap_or(all_dss, gt->fuse_topo.g_dss_mask,
+		num_dss = bitmap_weighted_or(all_dss, gt->fuse_topo.g_dss_mask,
 			  gt->fuse_topo.c_dss_mask, XE_MAX_DSS_FUSE_BITS);
 
-		num_dss = bitmap_weight(all_dss, XE_MAX_DSS_FUSE_BITS);
 		num_eus = bitmap_weight(gt->fuse_topo.eu_mask_per_dss,
 					XE_MAX_EU_FUSE_BITS) * num_dss;
 
-- 
2.43.0