Message-ID: <CO1PR11MB50894ACC6EF0DB09FA557188D61CA@CO1PR11MB5089.namprd11.prod.outlook.com>
Date: Wed, 24 Sep 2025 18:14:35 +0000
From: "Keller, Jacob E" <jacob.e.keller@...el.com>
To: Pradyumn Rahar <pradyumn.rahar@...cle.com>, "shayd@...dia.com"
	<shayd@...dia.com>
CC: "anand.a.khoje@...cle.com" <anand.a.khoje@...cle.com>,
	"andrew+netdev@...n.ch" <andrew+netdev@...n.ch>, "davem@...emloft.net"
	<davem@...emloft.net>, "edumazet@...gle.com" <edumazet@...gle.com>,
	"elic@...dia.com" <elic@...dia.com>, "kuba@...nel.org" <kuba@...nel.org>,
	"leon@...nel.org" <leon@...nel.org>, "linux-kernel@...r.kernel.org"
	<linux-kernel@...r.kernel.org>, "linux-rdma@...r.kernel.org"
	<linux-rdma@...r.kernel.org>, "manjunath.b.patil@...cle.com"
	<manjunath.b.patil@...cle.com>, "mbloch@...dia.com" <mbloch@...dia.com>,
	"moshe@...dia.com" <moshe@...dia.com>, "netdev@...r.kernel.org"
	<netdev@...r.kernel.org>, "pabeni@...hat.com" <pabeni@...hat.com>,
	"qing.huang@...cle.com" <qing.huang@...cle.com>,
	"rajesh.sivaramasubramaniom@...cle.com"
	<rajesh.sivaramasubramaniom@...cle.com>, "rama.nichanamatlu@...cle.com"
	<rama.nichanamatlu@...cle.com>, "rohit.sajan.kumar@...cle.com"
	<rohit.sajan.kumar@...cle.com>, "saeedm@...dia.com" <saeedm@...dia.com>,
	"tariqt@...dia.com" <tariqt@...dia.com>
Subject: RE: [PATCH net v2 1/1] net/mlx5: Clean up only new IRQ glue on
 request_irq() failure



> -----Original Message-----
> From: Pradyumn Rahar <pradyumn.rahar@...cle.com>
> Sent: Monday, September 22, 2025 11:28 PM
> To: shayd@...dia.com
> Cc: anand.a.khoje@...cle.com; andrew+netdev@...n.ch;
> davem@...emloft.net; edumazet@...gle.com; elic@...dia.com; Keller, Jacob E
> <jacob.e.keller@...el.com>; kuba@...nel.org; leon@...nel.org; linux-
> kernel@...r.kernel.org; linux-rdma@...r.kernel.org;
> manjunath.b.patil@...cle.com; mbloch@...dia.com;
> pradyumn.rahar@...cle.com; moshe@...dia.com; netdev@...r.kernel.org;
> pabeni@...hat.com; qing.huang@...cle.com;
> rajesh.sivaramasubramaniom@...cle.com; rama.nichanamatlu@...cle.com;
> rohit.sajan.kumar@...cle.com; saeedm@...dia.com; tariqt@...dia.com
> Subject: [PATCH net v2 1/1] net/mlx5: Clean up only new IRQ glue on request_irq()
> failure
> 
> The mlx5_irq_alloc() function can inadvertently free the entire rmap
> and end up in a crash[1] when the other threads tries to access this,
> when request_irq() fails due to exhausted IRQ vectors. This commit
> modifies the cleanup to remove only the specific IRQ mapping that was
> just added.
> 
> This prevents removal of other valid mappings and ensures precise
> cleanup of the failed IRQ allocation's associated glue object.
> 
> Note: This error is observed when both fwctl and rds configs are enabled.
> 
> [1]
> mlx5_core 0000:05:00.0: Successfully registered panic handler for port 1
> mlx5_core 0000:05:00.0: mlx5_irq_alloc:293:(pid 66740): Failed to
> request irq. err = -28
> infiniband mlx5_0: mlx5_ib_test_wc:290:(pid 66740): Error -28 while
> trying to test write-combining support
> mlx5_core 0000:05:00.0: Successfully unregistered panic handler for port 1
> mlx5_core 0000:06:00.0: Successfully registered panic handler for port 1
> mlx5_core 0000:06:00.0: mlx5_irq_alloc:293:(pid 66740): Failed to
> request irq. err = -28
> infiniband mlx5_0: mlx5_ib_test_wc:290:(pid 66740): Error -28 while
> trying to test write-combining support
> mlx5_core 0000:06:00.0: Successfully unregistered panic handler for port 1
> mlx5_core 0000:03:00.0: mlx5_irq_alloc:293:(pid 28895): Failed to
> request irq. err = -28
> mlx5_core 0000:05:00.0: mlx5_irq_alloc:293:(pid 28895): Failed to
> request irq. err = -28
> general protection fault, probably for non-canonical address
> 0xe277a58fde16f291: 0000 [#1] SMP NOPTI
> 
> RIP: 0010:free_irq_cpu_rmap+0x23/0x7d
> Call Trace:
>    <TASK>
>    ? show_trace_log_lvl+0x1d6/0x2f9
>    ? show_trace_log_lvl+0x1d6/0x2f9
>    ? mlx5_irq_alloc.cold+0x5d/0xf3 [mlx5_core]
>    ? __die_body.cold+0x8/0xa
>    ? die_addr+0x39/0x53
>    ? exc_general_protection+0x1c4/0x3e9
>    ? dev_vprintk_emit+0x5f/0x90
>    ? asm_exc_general_protection+0x22/0x27
>    ? free_irq_cpu_rmap+0x23/0x7d
>    mlx5_irq_alloc.cold+0x5d/0xf3 [mlx5_core]
>    irq_pool_request_vector+0x7d/0x90 [mlx5_core]
>    mlx5_irq_request+0x2e/0xe0 [mlx5_core]
>    mlx5_irq_request_vector+0xad/0xf7 [mlx5_core]
>    comp_irq_request_pci+0x64/0xf0 [mlx5_core]
>    create_comp_eq+0x71/0x385 [mlx5_core]
>    ? mlx5e_open_xdpsq+0x11c/0x230 [mlx5_core]
>    mlx5_comp_eqn_get+0x72/0x90 [mlx5_core]
>    ? xas_load+0x8/0x91
>    mlx5_comp_irqn_get+0x40/0x90 [mlx5_core]
>    mlx5e_open_channel+0x7d/0x3c7 [mlx5_core]
>    mlx5e_open_channels+0xad/0x250 [mlx5_core]
>    mlx5e_open_locked+0x3e/0x110 [mlx5_core]
>    mlx5e_open+0x23/0x70 [mlx5_core]
>    __dev_open+0xf1/0x1a5
>    __dev_change_flags+0x1e1/0x249
>    dev_change_flags+0x21/0x5c
>    do_setlink+0x28b/0xcc4
>    ? __nla_parse+0x22/0x3d
>    ? inet6_validate_link_af+0x6b/0x108
>    ? cpumask_next+0x1f/0x35
>    ? __snmp6_fill_stats64.constprop.0+0x66/0x107
>    ? __nla_validate_parse+0x48/0x1e6
>    __rtnl_newlink+0x5ff/0xa57
>    ? kmem_cache_alloc_trace+0x164/0x2ce
>    rtnl_newlink+0x44/0x6e
>    rtnetlink_rcv_msg+0x2bb/0x362
>    ? __netlink_sendskb+0x4c/0x6c
>    ? netlink_unicast+0x28f/0x2ce
>    ? rtnl_calcit.isra.0+0x150/0x146
>    netlink_rcv_skb+0x5f/0x112
>    netlink_unicast+0x213/0x2ce
>    netlink_sendmsg+0x24f/0x4d9
>    __sock_sendmsg+0x65/0x6a
>    ____sys_sendmsg+0x28f/0x2c9
>    ? import_iovec+0x17/0x2b
>    ___sys_sendmsg+0x97/0xe0
>    __sys_sendmsg+0x81/0xd8
>    do_syscall_64+0x35/0x87
>    entry_SYSCALL_64_after_hwframe+0x6e/0x0
> RIP: 0033:0x7fc328603727
> Code: c3 66 90 41 54 41 89 d4 55 48 89 f5 53 89 fb 48 83 ec 10 e8 0b ed
> ff ff 44 89 e2 48 89 ee 89 df 41 89 c0 b8 2e 00 00 00 0f 05 <48> 3d 00
> f0 ff ff 77 35 44 89 c7 48 89 44 24 08 e8 44 ed ff ff 48
> RSP: 002b:00007ffe8eb3f1a0 EFLAGS: 00000293 ORIG_RAX:
> 000000000000002e
> RAX: ffffffffffffffda RBX: 000000000000000d RCX: 00007fc328603727
> RDX: 0000000000000000 RSI: 00007ffe8eb3f1f0 RDI: 000000000000000d
> RBP: 00007ffe8eb3f1f0 R08: 0000000000000000 R09: 0000000000000000
> R10: 0000000000000000 R11: 0000000000000293 R12: 0000000000000000
> R13: 0000000000000000 R14: 00007ffe8eb3f3c8 R15: 00007ffe8eb3f3bc
>    </TASK>
> ---[ end trace f43ce73c3c2b13a2 ]---
> RIP: 0010:free_irq_cpu_rmap+0x23/0x7d
> Code: 0f 1f 80 00 00 00 00 48 85 ff 74 6b 55 48 89 fd 53 66 83 7f 06 00
> 74 24 31 db 48 8b 55 08 0f b7 c3 48 8b 04 c2 48 85 c0 74 09 <8b> 38 31
> f6 e8 c4 0a b8 ff 83 c3 01 66 3b 5d 06 72 de b8 ff ff ff
> RSP: 0018:ff384881640eaca0 EFLAGS: 00010282
> RAX: e277a58fde16f291 RBX: 0000000000000000 RCX: 0000000000000000
> RDX: ff2335e2e20b3600 RSI: 0000000000000000 RDI: ff2335e2e20b3400
> RBP: ff2335e2e20b3400 R08: 0000000000000000 R09: 0000000000000000
> R10: 0000000000000000 R11: 00000000ffffffe4 R12: ff384881640ead88
> R13: ff2335c3760751e0 R14: ff2335e2e1672200 R15: ff2335c3760751f8
> FS:  00007fc32ac22480(0000) GS:ff2335e2d6e00000(0000)
> knlGS:0000000000000000
> CS:  0010 DS: 0000 ES: 0000 CR0: 0000000080050033
> CR2: 00007f651ab54000 CR3: 00000029f1206003 CR4: 0000000000771ef0
> DR0: 0000000000000000 DR1: 0000000000000000 DR2: 0000000000000000
> DR3: 0000000000000000 DR6: 00000000fffe0ff0 DR7: 0000000000000400
> PKRU: 55555554
> Kernel panic - not syncing: Fatal exception
> Kernel Offset: 0x1dc00000 from 0xffffffff81000000 (relocation range:
> 0xffffffff80000000-0xffffffffbfffffff)
> kvm-guest: disable async PF for cpu 0
> 
> Fixes: 3354822cde5a ("net/mlx5: Use dynamic msix vectors allocation")
> Signed-off-by: Mohith Kumar
> Thummaluru<mohith.k.kumar.thummaluru@...cle.com>
> Tested-by: Mohith Kumar
> Thummaluru<mohith.k.kumar.thummaluru@...cle.com>
> Reviewed-by: Moshe Shemesh<moshe@...dia.com>
> Signed-off-by: Pradyumn Rahar <pradyumn.rahar@...cle.com>
> ---
> v1->v2: removed unnecessary braces from if conditon.
> ---
>  drivers/net/ethernet/mellanox/mlx5/core/pci_irq.c | 6 ++----
>  1 file changed, 2 insertions(+), 4 deletions(-)
> 
> diff --git a/drivers/net/ethernet/mellanox/mlx5/core/pci_irq.c
> b/drivers/net/ethernet/mellanox/mlx5/core/pci_irq.c
> index 692ef9c2f729..82ada674f8e2 100644
> --- a/drivers/net/ethernet/mellanox/mlx5/core/pci_irq.c
> +++ b/drivers/net/ethernet/mellanox/mlx5/core/pci_irq.c
> @@ -324,10 +324,8 @@ struct mlx5_irq *mlx5_irq_alloc(struct mlx5_irq_pool
> *pool, int i,
>  	free_irq(irq->map.virq, &irq->nh);
>  err_req_irq:
>  #ifdef CONFIG_RFS_ACCEL
> -	if (i && rmap && *rmap) {
> -		free_irq_cpu_rmap(*rmap);
> -		*rmap = NULL;
> -	}
> +	if (i && rmap && *rmap)
> +		irq_cpu_rmap_remove(*rmap, irq->map.virq);

Presumably if this fails during initialization, the caller of mlx5_irq_alloc which allocates multiple IRQs would be responsible for cleaning up anything it allocated before failing. Makes sense. Cleaning up only what this function did makes more sense.

Reviewed-by: Jacob Keller <jacob.e.keller@...el.com>

>  err_irq_rmap:
>  #endif
>  	if (i && pci_msix_can_alloc_dyn(dev->pdev))
> --
> 2.43.7

Powered by blists - more mailing lists