| Message ID | 20250324014639.4105332-1-dw@davidwei.uk (mailing list archive) |
|---|---|
| State | Changes Requested |
| Delegated to: | Netdev Maintainers |
| Headers | show |
| Series | [net] net: page_pool: replace ASSERT_RTNL() in page_pool_init() | expand |
On 03/23, David Wei wrote: > Replace a stray ASSERT_RTNL() in page_pool_init() with > netdev_assert_locked(). > > Fixes: 1d22d3060b9b ("net: drop rtnl_lock for queue_mgmt operations") > Signed-off-by: David Wei <dw@davidwei.uk> > --- > net/core/page_pool.c | 2 +- > 1 file changed, 1 insertion(+), 1 deletion(-) > > diff --git a/net/core/page_pool.c b/net/core/page_pool.c > index f5e908c9e7ad..2f469b02ea31 100644 > --- a/net/core/page_pool.c > +++ b/net/core/page_pool.c > @@ -281,7 +281,7 @@ static int page_pool_init(struct page_pool *pool, > * configuration doesn't change while we're initializing > * the page_pool. > */ > - ASSERT_RTNL(); > + netdev_assert_locked(params->netdev); > rxq = __netif_get_rx_queue(pool->slow.netdev, > pool->slow.queue_idx); > pool->mp_priv = rxq->mp_params.mp_priv; Not sure we can do this unconditionally. Since we still have plenty of drivers that might not use netdev instance lock (yet). Probably should be something like the following? if (netdev_need_ops_lock(params->netdev)) netdev_assert_locked(params->netdev); else ASSERT_RTNL(); ? We should probably wait for https://lore.kernel.org/netdev/20250312223507.805719-11-kuba@kernel.org/T/#m2053ac617759a9005806e56a7df97c378b76ec77 to land and use netdev_ops_assert_locked instead?
On Sun, Mar 23, 2025 at 06:46:39PM -0700, David Wei wrote: > Replace a stray ASSERT_RTNL() in page_pool_init() with > netdev_assert_locked(). > > Fixes: 1d22d3060b9b ("net: drop rtnl_lock for queue_mgmt operations") > Signed-off-by: David Wei <dw@davidwei.uk> > --- > net/core/page_pool.c | 2 +- > 1 file changed, 1 insertion(+), 1 deletion(-) > > diff --git a/net/core/page_pool.c b/net/core/page_pool.c > index f5e908c9e7ad..2f469b02ea31 100644 > --- a/net/core/page_pool.c > +++ b/net/core/page_pool.c > @@ -281,7 +281,7 @@ static int page_pool_init(struct page_pool *pool, > * configuration doesn't change while we're initializing > * the page_pool. > */ > - ASSERT_RTNL(); > + netdev_assert_locked(params->netdev); Adding a bit more context: if (pool->slow.flags & PP_FLAG_ALLOW_UNREADABLE_NETMEM) { /* We rely on rtnl_lock()ing to make sure netdev_rx_queue * configuration doesn't change while we're initializing * the page_pool. */ ASSERT_RTNL(); If ASSERT_RTNL() is now wrong, you also need to update the comment. Andrew
On 2025-03-23 19:39, Stanislav Fomichev wrote: > On 03/23, David Wei wrote: >> Replace a stray ASSERT_RTNL() in page_pool_init() with >> netdev_assert_locked(). >> >> Fixes: 1d22d3060b9b ("net: drop rtnl_lock for queue_mgmt operations") >> Signed-off-by: David Wei <dw@davidwei.uk> >> --- >> net/core/page_pool.c | 2 +- >> 1 file changed, 1 insertion(+), 1 deletion(-) >> >> diff --git a/net/core/page_pool.c b/net/core/page_pool.c >> index f5e908c9e7ad..2f469b02ea31 100644 >> --- a/net/core/page_pool.c >> +++ b/net/core/page_pool.c >> @@ -281,7 +281,7 @@ static int page_pool_init(struct page_pool *pool, >> * configuration doesn't change while we're initializing >> * the page_pool. >> */ >> - ASSERT_RTNL(); >> + netdev_assert_locked(params->netdev); >> rxq = __netif_get_rx_queue(pool->slow.netdev, >> pool->slow.queue_idx); >> pool->mp_priv = rxq->mp_params.mp_priv; > > Not sure we can do this unconditionally. Since we still have plenty of > drivers that might not use netdev instance lock (yet). Probably should > be something like the following? > > if (netdev_need_ops_lock(params->netdev)) > netdev_assert_locked(params->netdev); > else > ASSERT_RTNL(); > > ? > > We should probably wait for https://lore.kernel.org/netdev/20250312223507.805719-11-kuba@kernel.org/T/#m2053ac617759a9005806e56a7df97c378b76ec77 > to land and use netdev_ops_assert_locked instead? Ah I didn't realise that migration hasn't entirely happened. Yes, I'll wait for netdev_ops_assert_locked() first.
On 2025-03-24 07:18, Andrew Lunn wrote: > On Sun, Mar 23, 2025 at 06:46:39PM -0700, David Wei wrote: >> Replace a stray ASSERT_RTNL() in page_pool_init() with >> netdev_assert_locked(). >> >> Fixes: 1d22d3060b9b ("net: drop rtnl_lock for queue_mgmt operations") >> Signed-off-by: David Wei <dw@davidwei.uk> >> --- >> net/core/page_pool.c | 2 +- >> 1 file changed, 1 insertion(+), 1 deletion(-) >> >> diff --git a/net/core/page_pool.c b/net/core/page_pool.c >> index f5e908c9e7ad..2f469b02ea31 100644 >> --- a/net/core/page_pool.c >> +++ b/net/core/page_pool.c >> @@ -281,7 +281,7 @@ static int page_pool_init(struct page_pool *pool, >> * configuration doesn't change while we're initializing >> * the page_pool. >> */ >> - ASSERT_RTNL(); >> + netdev_assert_locked(params->netdev); > > Adding a bit more context: > > if (pool->slow.flags & PP_FLAG_ALLOW_UNREADABLE_NETMEM) { > /* We rely on rtnl_lock()ing to make sure netdev_rx_queue > * configuration doesn't change while we're initializing > * the page_pool. > */ > ASSERT_RTNL(); > > If ASSERT_RTNL() is now wrong, you also need to update the comment. Thanks, I'll update the docs too after switching to netdev_ops_assert_locked(). > > Andrew
diff --git a/net/core/page_pool.c b/net/core/page_pool.c index f5e908c9e7ad..2f469b02ea31 100644 --- a/net/core/page_pool.c +++ b/net/core/page_pool.c @@ -281,7 +281,7 @@ static int page_pool_init(struct page_pool *pool, * configuration doesn't change while we're initializing * the page_pool. */ - ASSERT_RTNL(); + netdev_assert_locked(params->netdev); rxq = __netif_get_rx_queue(pool->slow.netdev, pool->slow.queue_idx); pool->mp_priv = rxq->mp_params.mp_priv;
Replace a stray ASSERT_RTNL() in page_pool_init() with netdev_assert_locked(). Fixes: 1d22d3060b9b ("net: drop rtnl_lock for queue_mgmt operations") Signed-off-by: David Wei <dw@davidwei.uk> --- net/core/page_pool.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-)