- Linux-stable-mirror - lists.linaro.org

FAILED: patch "[PATCH] vmxnet3: update MTU after device quiesce" failed to apply to 5.15-stable tree

by gregkh＠linuxfoundation.org

The patch below does not apply to the 5.15-stable tree. If someone wants it applied there, or to any other stable or longterm tree, then please email the backport, including the original git commit id to <stable(a)vger.kernel.org>. To reproduce the conflict and resubmit, you may use the following commands: git fetch https://git.kernel.org/pub/scm/linux/kernel/git/stable/linux.git/ linux-5.15.y git checkout FETCH_HEAD git cherry-pick -x 43f0999af011fba646e015f0bb08b6c3002a0170 # <resolve conflicts, build, test, etc.> git commit -s git send-email --to '<stable(a)vger.kernel.org>' --in-reply-to '2025052425-posture-finlike-af9b@gregkh' --subject-prefix 'PATCH 5.15.y' HEAD^.. Possible dependencies: thanks, greg k-h ------------------ original commit in Linus's tree ------------------ From 43f0999af011fba646e015f0bb08b6c3002a0170 Mon Sep 17 00:00:00 2001 From: Ronak Doshi <ronak.doshi(a)broadcom.com> Date: Thu, 15 May 2025 19:04:56 +0000 Subject: [PATCH] vmxnet3: update MTU after device quiesce Currently, when device mtu is updated, vmxnet3 updates netdev mtu, quiesces the device and then reactivates it for the ESXi to know about the new mtu. So, technically the OS stack can start using the new mtu before ESXi knows about the new mtu. This can lead to issues for TSO packets which use mss as per the new mtu configured. This patch fixes this issue by moving the mtu write after device quiesce. Cc: stable(a)vger.kernel.org Fixes: d1a890fa37f2 ("net: VMware virtual Ethernet NIC driver: vmxnet3") Signed-off-by: Ronak Doshi <ronak.doshi(a)broadcom.com> Acked-by: Guolin Yang <guolin.yang(a)broadcom.com> Changes v1-> v2: Moved MTU write after destroy of rx rings Link: https://patch.msgid.link/20250515190457.8597-1-ronak.doshi@broadcom.com Signed-off-by: Jakub Kicinski <kuba(a)kernel.org> diff --git a/drivers/net/vmxnet3/vmxnet3_drv.c b/drivers/net/vmxnet3/vmxnet3_drv.c index 3df6aabc7e33..c676979c7ab9 100644 --- a/drivers/net/vmxnet3/vmxnet3_drv.c +++ b/drivers/net/vmxnet3/vmxnet3_drv.c @@ -3607,8 +3607,6 @@ vmxnet3_change_mtu(struct net_device *netdev, int new_mtu) struct vmxnet3_adapter *adapter = netdev_priv(netdev); int err = 0; - WRITE_ONCE(netdev->mtu, new_mtu); - /* * Reset_work may be in the middle of resetting the device, wait for its * completion. @@ -3622,6 +3620,7 @@ vmxnet3_change_mtu(struct net_device *netdev, int new_mtu) /* we need to re-create the rx queue based on the new mtu */ vmxnet3_rq_destroy_all(adapter); + WRITE_ONCE(netdev->mtu, new_mtu); vmxnet3_adjust_rx_ring_size(adapter); err = vmxnet3_rq_create_all(adapter); if (err) { @@ -3638,6 +3637,8 @@ vmxnet3_change_mtu(struct net_device *netdev, int new_mtu) "Closing it\n", err); goto out; } + } else { + WRITE_ONCE(netdev->mtu, new_mtu); } out:

2 months, 3 weeks

2
1
0 0

[PATCH] drm/amd/display: Drop dm_prepare_suspend() and dm_complete()

by Mario Limonciello

From: "Mario Limonciello" <mario.limonciello(a)amd.com> [Why] dm_prepare_suspend() was added in commit 50e0bae34fa6b ("drm/amd/display: Add and use new dm_prepare_suspend() callback") to allow display to turn off earlier in the suspend sequence. This caused a regression that HDMI audio sometimes didn't work properly after resume unless audio was playing during suspend. [How] Drop dm_prepare_suspend() callback. All code in it will still run during dm_suspend(). Also drop unnecessary dm_complete() callback. dm_complete() was used for failed prepare and also for any case of successful resume. The code in it already runs in dm_resume(). This change will introduce more time that the display is turned on during suspend sequence. The compositor can turn it off sooner if desired. Cc: Harry Wentland <harry.wentland(a)amd.com> Reported-by: Przemysław Kopa <prz.kopa(a)gmail.com> Closes: https://lore.kernel.org/amd-gfx/1cea0d56-7739-4ad9-bf8e-c9330faea2bb@kernel… Tested-by: Przemysław Kopa <prz.kopa(a)gmail.com> Reported-by: Kalvin <hikaph+oss(a)gmail.com> Closes: https://github.com/alsa-project/alsa-lib/issues/465 Closes: https://gitlab.freedesktop.org/pipewire/pipewire/-/issues/4809 Cc: stable(a)vger.kernel.org Fixes: 50e0bae34fa6b ("drm/amd/display: Add and use new dm_prepare_suspend() callback") Signed-off-by: Mario Limonciello <mario.limonciello(a)amd.com> --- NOTE: The complete pmops callback is still present but does nothing right now. It's left for completeness sake in case another IP needs to do something in prepare() and undo it in a failure with complete(). .../gpu/drm/amd/display/amdgpu_dm/amdgpu_dm.c | 21 ------------------- 1 file changed, 21 deletions(-) diff --git a/drivers/gpu/drm/amd/display/amdgpu_dm/amdgpu_dm.c b/drivers/gpu/drm/amd/display/amdgpu_dm/amdgpu_dm.c index e34d98a945f2..fadc6098eaee 100644 --- a/drivers/gpu/drm/amd/display/amdgpu_dm/amdgpu_dm.c +++ b/drivers/gpu/drm/amd/display/amdgpu_dm/amdgpu_dm.c @@ -3182,25 +3182,6 @@ static void dm_destroy_cached_state(struct amdgpu_device *adev) dm->cached_state = NULL; } -static void dm_complete(struct amdgpu_ip_block *ip_block) -{ - struct amdgpu_device *adev = ip_block->adev; - - dm_destroy_cached_state(adev); -} - -static int dm_prepare_suspend(struct amdgpu_ip_block *ip_block) -{ - struct amdgpu_device *adev = ip_block->adev; - - if (amdgpu_in_reset(adev)) - return 0; - - WARN_ON(adev->dm.cached_state); - - return dm_cache_state(adev); -} - static int dm_suspend(struct amdgpu_ip_block *ip_block) { struct amdgpu_device *adev = ip_block->adev; @@ -3626,10 +3607,8 @@ static const struct amd_ip_funcs amdgpu_dm_funcs = { .early_fini = amdgpu_dm_early_fini, .hw_init = dm_hw_init, .hw_fini = dm_hw_fini, - .prepare_suspend = dm_prepare_suspend, .suspend = dm_suspend, .resume = dm_resume, - .complete = dm_complete, .is_idle = dm_is_idle, .wait_for_idle = dm_wait_for_idle, .check_soft_reset = dm_check_soft_reset, -- 2.49.0

2 months, 3 weeks

2
1
0 0

FAILED: patch "[PATCH] vmxnet3: update MTU after device quiesce" failed to apply to 6.1-stable tree

by gregkh＠linuxfoundation.org

The patch below does not apply to the 6.1-stable tree. If someone wants it applied there, or to any other stable or longterm tree, then please email the backport, including the original git commit id to <stable(a)vger.kernel.org>. To reproduce the conflict and resubmit, you may use the following commands: git fetch https://git.kernel.org/pub/scm/linux/kernel/git/stable/linux.git/ linux-6.1.y git checkout FETCH_HEAD git cherry-pick -x 43f0999af011fba646e015f0bb08b6c3002a0170 # <resolve conflicts, build, test, etc.> git commit -s git send-email --to '<stable(a)vger.kernel.org>' --in-reply-to '2025052425-willed-landline-ff5b@gregkh' --subject-prefix 'PATCH 6.1.y' HEAD^.. Possible dependencies: thanks, greg k-h ------------------ original commit in Linus's tree ------------------ From 43f0999af011fba646e015f0bb08b6c3002a0170 Mon Sep 17 00:00:00 2001 From: Ronak Doshi <ronak.doshi(a)broadcom.com> Date: Thu, 15 May 2025 19:04:56 +0000 Subject: [PATCH] vmxnet3: update MTU after device quiesce Currently, when device mtu is updated, vmxnet3 updates netdev mtu, quiesces the device and then reactivates it for the ESXi to know about the new mtu. So, technically the OS stack can start using the new mtu before ESXi knows about the new mtu. This can lead to issues for TSO packets which use mss as per the new mtu configured. This patch fixes this issue by moving the mtu write after device quiesce. Cc: stable(a)vger.kernel.org Fixes: d1a890fa37f2 ("net: VMware virtual Ethernet NIC driver: vmxnet3") Signed-off-by: Ronak Doshi <ronak.doshi(a)broadcom.com> Acked-by: Guolin Yang <guolin.yang(a)broadcom.com> Changes v1-> v2: Moved MTU write after destroy of rx rings Link: https://patch.msgid.link/20250515190457.8597-1-ronak.doshi@broadcom.com Signed-off-by: Jakub Kicinski <kuba(a)kernel.org> diff --git a/drivers/net/vmxnet3/vmxnet3_drv.c b/drivers/net/vmxnet3/vmxnet3_drv.c index 3df6aabc7e33..c676979c7ab9 100644 --- a/drivers/net/vmxnet3/vmxnet3_drv.c +++ b/drivers/net/vmxnet3/vmxnet3_drv.c @@ -3607,8 +3607,6 @@ vmxnet3_change_mtu(struct net_device *netdev, int new_mtu) struct vmxnet3_adapter *adapter = netdev_priv(netdev); int err = 0; - WRITE_ONCE(netdev->mtu, new_mtu); - /* * Reset_work may be in the middle of resetting the device, wait for its * completion. @@ -3622,6 +3620,7 @@ vmxnet3_change_mtu(struct net_device *netdev, int new_mtu) /* we need to re-create the rx queue based on the new mtu */ vmxnet3_rq_destroy_all(adapter); + WRITE_ONCE(netdev->mtu, new_mtu); vmxnet3_adjust_rx_ring_size(adapter); err = vmxnet3_rq_create_all(adapter); if (err) { @@ -3638,6 +3637,8 @@ vmxnet3_change_mtu(struct net_device *netdev, int new_mtu) "Closing it\n", err); goto out; } + } else { + WRITE_ONCE(netdev->mtu, new_mtu); } out:

2 months, 3 weeks

2
1
0 0

FAILED: patch "[PATCH] net: dsa: microchip: linearize skb for tail-tagging switches" failed to apply to 6.6-stable tree

by gregkh＠linuxfoundation.org

The patch below does not apply to the 6.6-stable tree. If someone wants it applied there, or to any other stable or longterm tree, then please email the backport, including the original git commit id to <stable(a)vger.kernel.org>. To reproduce the conflict and resubmit, you may use the following commands: git fetch https://git.kernel.org/pub/scm/linux/kernel/git/stable/linux.git/ linux-6.6.y git checkout FETCH_HEAD git cherry-pick -x ba54bce747fa9e07896c1abd9b48545f7b4b31d2 # <resolve conflicts, build, test, etc.> git commit -s git send-email --to '<stable(a)vger.kernel.org>' --in-reply-to '2025052457-rise-remodeler-f63d@gregkh' --subject-prefix 'PATCH 6.6.y' HEAD^.. Possible dependencies: thanks, greg k-h ------------------ original commit in Linus's tree ------------------ From ba54bce747fa9e07896c1abd9b48545f7b4b31d2 Mon Sep 17 00:00:00 2001 From: Jakob Unterwurzacher <jakobunt(a)gmail.com> Date: Thu, 15 May 2025 09:29:19 +0200 Subject: [PATCH] net: dsa: microchip: linearize skb for tail-tagging switches The pointer arithmentic for accessing the tail tag only works for linear skbs. For nonlinear skbs, it reads uninitialized memory inside the skb headroom, essentially randomizing the tag. I have observed it gets set to 6 most of the time. Example where ksz9477_rcv thinks that the packet from port 1 comes from port 6 (which does not exist for the ksz9896 that's in use), dropping the packet. Debug prints added by me (not included in this patch): [ 256.645337] ksz9477_rcv:323 tag0=6 [ 256.645349] skb len=47 headroom=78 headlen=0 tailroom=0 mac=(64,14) mac_len=14 net=(78,0) trans=78 shinfo(txflags=0 nr_frags=1 gso(size=0 type=0 segs=0)) csum(0x0 start=0 offset=0 ip_summed=0 complete_sw=0 valid=0 level=0) hash(0x0 sw=0 l4=0) proto=0x00f8 pkttype=1 iif=3 priority=0x0 mark=0x0 alloc_cpu=0 vlan_all=0x0 encapsulation=0 inner(proto=0x0000, mac=0, net=0, trans=0) [ 256.645377] dev name=end1 feat=0x0002e10200114bb3 [ 256.645386] skb headroom: 00000000: 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 [ 256.645395] skb headroom: 00000010: 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 [ 256.645403] skb headroom: 00000020: 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 [ 256.645411] skb headroom: 00000030: 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 [ 256.645420] skb headroom: 00000040: ff ff ff ff ff ff 00 1c 19 f2 e2 db 08 06 [ 256.645428] skb frag: 00000000: 00 01 08 00 06 04 00 01 00 1c 19 f2 e2 db 0a 02 [ 256.645436] skb frag: 00000010: 00 83 00 00 00 00 00 00 0a 02 a0 2f 00 00 00 00 [ 256.645444] skb frag: 00000020: 00 00 00 00 00 00 00 00 00 00 00 00 00 00 01 [ 256.645452] ksz_common_rcv:92 dsa_conduit_find_user returned NULL Call skb_linearize before trying to access the tag. This patch fixes ksz9477_rcv which is used by the ksz9896 I have at hand, and also applies the same fix to ksz8795_rcv which seems to have the same problem. Signed-off-by: Jakob Unterwurzacher <jakob.unterwurzacher(a)cherry.de> CC: stable(a)vger.kernel.org Fixes: 016e43a26bab ("net: dsa: ksz: Add KSZ8795 tag code") Fixes: 8b8010fb7876 ("dsa: add support for Microchip KSZ tail tagging") Reviewed-by: Vladimir Oltean <olteanv(a)gmail.com> Link: https://patch.msgid.link/20250515072920.2313014-1-jakob.unterwurzacher@cher… Signed-off-by: Jakub Kicinski <kuba(a)kernel.org> diff --git a/net/dsa/tag_ksz.c b/net/dsa/tag_ksz.c index c33d4bf17929..0b7564b53790 100644 --- a/net/dsa/tag_ksz.c +++ b/net/dsa/tag_ksz.c @@ -140,7 +140,12 @@ static struct sk_buff *ksz8795_xmit(struct sk_buff *skb, struct net_device *dev) static struct sk_buff *ksz8795_rcv(struct sk_buff *skb, struct net_device *dev) { - u8 *tag = skb_tail_pointer(skb) - KSZ_EGRESS_TAG_LEN; + u8 *tag; + + if (skb_linearize(skb)) + return NULL; + + tag = skb_tail_pointer(skb) - KSZ_EGRESS_TAG_LEN; return ksz_common_rcv(skb, dev, tag[0] & KSZ8795_TAIL_TAG_EG_PORT_M, KSZ_EGRESS_TAG_LEN); @@ -311,10 +316,16 @@ static struct sk_buff *ksz9477_xmit(struct sk_buff *skb, static struct sk_buff *ksz9477_rcv(struct sk_buff *skb, struct net_device *dev) { - /* Tag decoding */ - u8 *tag = skb_tail_pointer(skb) - KSZ_EGRESS_TAG_LEN; - unsigned int port = tag[0] & KSZ9477_TAIL_TAG_EG_PORT_M; unsigned int len = KSZ_EGRESS_TAG_LEN; + unsigned int port; + u8 *tag; + + if (skb_linearize(skb)) + return NULL; + + /* Tag decoding */ + tag = skb_tail_pointer(skb) - KSZ_EGRESS_TAG_LEN; + port = tag[0] & KSZ9477_TAIL_TAG_EG_PORT_M; /* Extra 4-bytes PTP timestamp */ if (tag[0] & KSZ9477_PTP_TAG_INDICATION) {

2 months, 3 weeks

2
2
0 0

FAILED: patch "[PATCH] vmxnet3: update MTU after device quiesce" failed to apply to 6.6-stable tree

by gregkh＠linuxfoundation.org

The patch below does not apply to the 6.6-stable tree. If someone wants it applied there, or to any other stable or longterm tree, then please email the backport, including the original git commit id to <stable(a)vger.kernel.org>. To reproduce the conflict and resubmit, you may use the following commands: git fetch https://git.kernel.org/pub/scm/linux/kernel/git/stable/linux.git/ linux-6.6.y git checkout FETCH_HEAD git cherry-pick -x 43f0999af011fba646e015f0bb08b6c3002a0170 # <resolve conflicts, build, test, etc.> git commit -s git send-email --to '<stable(a)vger.kernel.org>' --in-reply-to '2025052424-banjo-exorcist-8407@gregkh' --subject-prefix 'PATCH 6.6.y' HEAD^.. Possible dependencies: thanks, greg k-h ------------------ original commit in Linus's tree ------------------ From 43f0999af011fba646e015f0bb08b6c3002a0170 Mon Sep 17 00:00:00 2001 From: Ronak Doshi <ronak.doshi(a)broadcom.com> Date: Thu, 15 May 2025 19:04:56 +0000 Subject: [PATCH] vmxnet3: update MTU after device quiesce Currently, when device mtu is updated, vmxnet3 updates netdev mtu, quiesces the device and then reactivates it for the ESXi to know about the new mtu. So, technically the OS stack can start using the new mtu before ESXi knows about the new mtu. This can lead to issues for TSO packets which use mss as per the new mtu configured. This patch fixes this issue by moving the mtu write after device quiesce. Cc: stable(a)vger.kernel.org Fixes: d1a890fa37f2 ("net: VMware virtual Ethernet NIC driver: vmxnet3") Signed-off-by: Ronak Doshi <ronak.doshi(a)broadcom.com> Acked-by: Guolin Yang <guolin.yang(a)broadcom.com> Changes v1-> v2: Moved MTU write after destroy of rx rings Link: https://patch.msgid.link/20250515190457.8597-1-ronak.doshi@broadcom.com Signed-off-by: Jakub Kicinski <kuba(a)kernel.org> diff --git a/drivers/net/vmxnet3/vmxnet3_drv.c b/drivers/net/vmxnet3/vmxnet3_drv.c index 3df6aabc7e33..c676979c7ab9 100644 --- a/drivers/net/vmxnet3/vmxnet3_drv.c +++ b/drivers/net/vmxnet3/vmxnet3_drv.c @@ -3607,8 +3607,6 @@ vmxnet3_change_mtu(struct net_device *netdev, int new_mtu) struct vmxnet3_adapter *adapter = netdev_priv(netdev); int err = 0; - WRITE_ONCE(netdev->mtu, new_mtu); - /* * Reset_work may be in the middle of resetting the device, wait for its * completion. @@ -3622,6 +3620,7 @@ vmxnet3_change_mtu(struct net_device *netdev, int new_mtu) /* we need to re-create the rx queue based on the new mtu */ vmxnet3_rq_destroy_all(adapter); + WRITE_ONCE(netdev->mtu, new_mtu); vmxnet3_adjust_rx_ring_size(adapter); err = vmxnet3_rq_create_all(adapter); if (err) { @@ -3638,6 +3637,8 @@ vmxnet3_change_mtu(struct net_device *netdev, int new_mtu) "Closing it\n", err); goto out; } + } else { + WRITE_ONCE(netdev->mtu, new_mtu); } out:

2 months, 3 weeks

2
1
0 0

New September Order. 89879 Friday, September 5, 2025 at 07:08:40 PM

by Info - Albinayah 684

Hi Stable, Please provide a quote for your products: Include: 1.Pricing (per unit) 2.Delivery cost & timeline 3.Quote expiry date Deadline: September Thanks! Kamal Prasad Albinayah Trading

2 months, 3 weeks

1
0
0 0

[PATCH v3 01/12] ASoC: codecs: wcd937x: set the comp soundwire port correctly

by Srinivas Kandagatla

For some reason we endup with setting soundwire port for HPHL_COMP and HPHR_COMP as zero, this can potentially result in a memory corruption due to accessing and setting -1 th element of port_map array. Fixes: 82be8c62a38c ("ASoC: codecs: wcd937x: add basic controls") Cc: <Stable(a)vger.kernel.org> Signed-off-by: Srinivas Kandagatla <srinivas.kandagatla(a)oss.qualcomm.com> --- sound/soc/codecs/wcd937x.c | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-) diff --git a/sound/soc/codecs/wcd937x.c b/sound/soc/codecs/wcd937x.c index 3b0a8cc314e0..de2dff3c56d3 100644 --- a/sound/soc/codecs/wcd937x.c +++ b/sound/soc/codecs/wcd937x.c @@ -2046,9 +2046,9 @@ static const struct snd_kcontrol_new wcd937x_snd_controls[] = { SOC_ENUM_EXT("RX HPH Mode", rx_hph_mode_mux_enum, wcd937x_rx_hph_mode_get, wcd937x_rx_hph_mode_put), - SOC_SINGLE_EXT("HPHL_COMP Switch", SND_SOC_NOPM, 0, 1, 0, + SOC_SINGLE_EXT("HPHL_COMP Switch", WCD937X_COMP_L, 0, 1, 0, wcd937x_get_compander, wcd937x_set_compander), - SOC_SINGLE_EXT("HPHR_COMP Switch", SND_SOC_NOPM, 1, 1, 0, + SOC_SINGLE_EXT("HPHR_COMP Switch", WCD937X_COMP_R, 1, 1, 0, wcd937x_get_compander, wcd937x_set_compander), SOC_SINGLE_TLV("HPHL Volume", WCD937X_HPH_L_EN, 0, 20, 1, line_gain), -- 2.50.0

2 months, 3 weeks

2
2
0 0

FAILED: patch "[PATCH] randstruct: gcc-plugin: Fix attribute addition" failed to apply to 5.4-stable tree

by gregkh＠linuxfoundation.org

The patch below does not apply to the 5.4-stable tree. If someone wants it applied there, or to any other stable or longterm tree, then please email the backport, including the original git commit id to <stable(a)vger.kernel.org>. To reproduce the conflict and resubmit, you may use the following commands: git fetch https://git.kernel.org/pub/scm/linux/kernel/git/stable/linux.git/ linux-5.4.y git checkout FETCH_HEAD git cherry-pick -x f39f18f3c3531aa802b58a20d39d96e82eb96c14 # <resolve conflicts, build, test, etc.> git commit -s git send-email --to '<stable(a)vger.kernel.org>' --in-reply-to '2025060528-osmosis-arose-1289@gregkh' --subject-prefix 'PATCH 5.4.y' HEAD^.. Possible dependencies: thanks, greg k-h ------------------ original commit in Linus's tree ------------------ From f39f18f3c3531aa802b58a20d39d96e82eb96c14 Mon Sep 17 00:00:00 2001 From: Kees Cook <kees(a)kernel.org> Date: Fri, 30 May 2025 15:18:28 -0700 Subject: [PATCH] randstruct: gcc-plugin: Fix attribute addition Based on changes in the 2021 public version of the randstruct out-of-tree GCC plugin[1], more carefully update the attributes on resulting decls, to avoid tripping checks in GCC 15's comptypes_check_enum_int() when it has been configured with "--enable-checking=misc": arch/arm64/kernel/kexec_image.c:132:14: internal compiler error: in comptypes_check_enum_int, at c/c-typeck.cc:1519 132 | const struct kexec_file_ops kexec_image_ops = { | ^~~~~~~~~~~~~~ internal_error(char const*, ...), at gcc/gcc/diagnostic-global-context.cc:517 fancy_abort(char const*, int, char const*), at gcc/gcc/diagnostic.cc:1803 comptypes_check_enum_int(tree_node*, tree_node*, bool*), at gcc/gcc/c/c-typeck.cc:1519 ... Link: https://archive.org/download/grsecurity/grsecurity-3.1-5.10.41-202105280954… [1] Reported-by: Thiago Jung Bauermann <thiago.bauermann(a)linaro.org> Closes: https://github.com/KSPP/linux/issues/367 Closes: https://lore.kernel.org/lkml/20250530000646.104457-1-thiago.bauermann@linar… Reported-by: Ingo Saitz <ingo(a)hannover.ccc.de> Closes: https://bugs.debian.org/cgi-bin/bugreport.cgi?bug=1104745 Fixes: 313dd1b62921 ("gcc-plugins: Add the randstruct plugin") Tested-by: Thiago Jung Bauermann <thiago.bauermann(a)linaro.org> Link: https://lore.kernel.org/r/20250530221824.work.623-kees@kernel.org Signed-off-by: Kees Cook <kees(a)kernel.org> diff --git a/scripts/gcc-plugins/gcc-common.h b/scripts/gcc-plugins/gcc-common.h index 3222c1070444..ef12c8f929ed 100644 --- a/scripts/gcc-plugins/gcc-common.h +++ b/scripts/gcc-plugins/gcc-common.h @@ -123,6 +123,38 @@ static inline tree build_const_char_string(int len, const char *str) return cstr; } +static inline void __add_type_attr(tree type, const char *attr, tree args) +{ + tree oldattr; + + if (type == NULL_TREE) + return; + oldattr = lookup_attribute(attr, TYPE_ATTRIBUTES(type)); + if (oldattr != NULL_TREE) { + gcc_assert(TREE_VALUE(oldattr) == args || TREE_VALUE(TREE_VALUE(oldattr)) == TREE_VALUE(args)); + return; + } + + TYPE_ATTRIBUTES(type) = copy_list(TYPE_ATTRIBUTES(type)); + TYPE_ATTRIBUTES(type) = tree_cons(get_identifier(attr), args, TYPE_ATTRIBUTES(type)); +} + +static inline void add_type_attr(tree type, const char *attr, tree args) +{ + tree main_variant = TYPE_MAIN_VARIANT(type); + + __add_type_attr(TYPE_CANONICAL(type), attr, args); + __add_type_attr(TYPE_CANONICAL(main_variant), attr, args); + __add_type_attr(main_variant, attr, args); + + for (type = TYPE_NEXT_VARIANT(main_variant); type; type = TYPE_NEXT_VARIANT(type)) { + if (!lookup_attribute(attr, TYPE_ATTRIBUTES(type))) + TYPE_ATTRIBUTES(type) = TYPE_ATTRIBUTES(main_variant); + + __add_type_attr(TYPE_CANONICAL(type), attr, args); + } +} + #define PASS_INFO(NAME, REF, ID, POS) \ struct register_pass_info NAME##_pass_info = { \ .pass = make_##NAME##_pass(), \ diff --git a/scripts/gcc-plugins/randomize_layout_plugin.c b/scripts/gcc-plugins/randomize_layout_plugin.c index 971a1908a8cc..ff65a4f87f24 100644 --- a/scripts/gcc-plugins/randomize_layout_plugin.c +++ b/scripts/gcc-plugins/randomize_layout_plugin.c @@ -73,6 +73,9 @@ static tree handle_randomize_layout_attr(tree *node, tree name, tree args, int f if (TYPE_P(*node)) { type = *node; + } else if (TREE_CODE(*node) == FIELD_DECL) { + *no_add_attrs = false; + return NULL_TREE; } else { gcc_assert(TREE_CODE(*node) == TYPE_DECL); type = TREE_TYPE(*node); @@ -348,15 +351,14 @@ static int relayout_struct(tree type) TREE_CHAIN(newtree[i]) = newtree[i+1]; TREE_CHAIN(newtree[num_fields - 1]) = NULL_TREE; + add_type_attr(type, "randomize_performed", NULL_TREE); + add_type_attr(type, "designated_init", NULL_TREE); + if (has_flexarray) + add_type_attr(type, "has_flexarray", NULL_TREE); + main_variant = TYPE_MAIN_VARIANT(type); - for (variant = main_variant; variant; variant = TYPE_NEXT_VARIANT(variant)) { + for (variant = main_variant; variant; variant = TYPE_NEXT_VARIANT(variant)) TYPE_FIELDS(variant) = newtree[0]; - TYPE_ATTRIBUTES(variant) = copy_list(TYPE_ATTRIBUTES(variant)); - TYPE_ATTRIBUTES(variant) = tree_cons(get_identifier("randomize_performed"), NULL_TREE, TYPE_ATTRIBUTES(variant)); - TYPE_ATTRIBUTES(variant) = tree_cons(get_identifier("designated_init"), NULL_TREE, TYPE_ATTRIBUTES(variant)); - if (has_flexarray) - TYPE_ATTRIBUTES(type) = tree_cons(get_identifier("has_flexarray"), NULL_TREE, TYPE_ATTRIBUTES(type)); - } /* * force a re-layout of the main variant @@ -424,10 +426,8 @@ static void randomize_type(tree type) if (lookup_attribute("randomize_layout", TYPE_ATTRIBUTES(TYPE_MAIN_VARIANT(type))) || is_pure_ops_struct(type)) relayout_struct(type); - for (variant = TYPE_MAIN_VARIANT(type); variant; variant = TYPE_NEXT_VARIANT(variant)) { - TYPE_ATTRIBUTES(type) = copy_list(TYPE_ATTRIBUTES(type)); - TYPE_ATTRIBUTES(type) = tree_cons(get_identifier("randomize_considered"), NULL_TREE, TYPE_ATTRIBUTES(type)); - } + add_type_attr(type, "randomize_considered", NULL_TREE); + #ifdef __DEBUG_PLUGIN fprintf(stderr, "Marking randomize_considered on struct %s\n", ORIG_TYPE_NAME(type)); #ifdef __DEBUG_VERBOSE

2 months, 3 weeks

2
2
0 0

FAILED: patch "[PATCH] arm64: dts: marvell: uDPU: define pinctrl state for alarm" failed to apply to 5.10-stable tree

by gregkh＠linuxfoundation.org

The patch below does not apply to the 5.10-stable tree. If someone wants it applied there, or to any other stable or longterm tree, then please email the backport, including the original git commit id to <stable(a)vger.kernel.org>. To reproduce the conflict and resubmit, you may use the following commands: git fetch https://git.kernel.org/pub/scm/linux/kernel/git/stable/linux.git/ linux-5.10.y git checkout FETCH_HEAD git cherry-pick -x b04f0d89e880bc2cca6a5c73cf287082c91878da # <resolve conflicts, build, test, etc.> git commit -s git send-email --to '<stable(a)vger.kernel.org>' --in-reply-to '2025052402-pantomime-relative-1605@gregkh' --subject-prefix 'PATCH 5.10.y' HEAD^.. Possible dependencies: thanks, greg k-h ------------------ original commit in Linus's tree ------------------ From b04f0d89e880bc2cca6a5c73cf287082c91878da Mon Sep 17 00:00:00 2001 From: Gabor Juhos <j4g8y7(a)gmail.com> Date: Fri, 9 May 2025 15:48:52 +0200 Subject: [PATCH] arm64: dts: marvell: uDPU: define pinctrl state for alarm LEDs The two alarm LEDs of on the uDPU board are stopped working since commit 78efa53e715e ("leds: Init leds class earlier"). The LEDs are driven by the GPIO{15,16} pins of the North Bridge GPIO controller. These pins are part of the 'spi_quad' pin group for which the 'spi' function is selected via the default pinctrl state of the 'spi' node. This is wrong however, since in order to allow controlling the LEDs, the pins should use the 'gpio' function. Before the commit mentined above, the 'spi' function is selected first by the pinctrl core before probing the spi driver, but then it gets overridden to 'gpio' implicitly via the devm_gpiod_get_index_optional() call from the 'leds-gpio' driver. After the commit, the LED subsystem gets initialized before the SPI subsystem, so the function of the pin group remains 'spi' which in turn prevents controlling of the LEDs. Despite the change of the initialization order, the root cause is that the pinctrl state definition is wrong since its initial commit 0d45062cfc89 ("arm64: dts: marvell: Add device tree for uDPU board"), To fix the problem, override the function in the 'spi_quad_pins' node to 'gpio' and move the pinctrl state definition from the 'spi' node into the 'leds' node. Cc: stable(a)vger.kernel.org # needs adjustment for < 6.1 Fixes: 0d45062cfc89 ("arm64: dts: marvell: Add device tree for uDPU board") Signed-off-by: Gabor Juhos <j4g8y7(a)gmail.com> Signed-off-by: Imre Kaloz <kaloz(a)openwrt.org> Signed-off-by: Gregory CLEMENT <gregory.clement(a)bootlin.com> diff --git a/arch/arm64/boot/dts/marvell/armada-3720-uDPU.dtsi b/arch/arm64/boot/dts/marvell/armada-3720-uDPU.dtsi index 3a9b6907185d..242820845707 100644 --- a/arch/arm64/boot/dts/marvell/armada-3720-uDPU.dtsi +++ b/arch/arm64/boot/dts/marvell/armada-3720-uDPU.dtsi @@ -26,6 +26,8 @@ memory@0 { leds { compatible = "gpio-leds"; + pinctrl-names = "default"; + pinctrl-0 = <&spi_quad_pins>; led-power1 { label = "udpu:green:power"; @@ -82,8 +84,6 @@ &sdhci0 { &spi0 { status = "okay"; - pinctrl-names = "default"; - pinctrl-0 = <&spi_quad_pins>; flash@0 { compatible = "jedec,spi-nor"; @@ -108,6 +108,10 @@ partition@180000 { }; }; +&spi_quad_pins { + function = "gpio"; +}; + &pinctrl_nb { i2c2_recovery_pins: i2c2-recovery-pins { groups = "i2c2";

2 months, 3 weeks

2
1
0 0

[PATCH for-rc] IB/rdmavt: Fix lock dependency in rvt_send/recv_cq

by Dennis Dalessandro

From: Nick Child <nchild(a)cornelisnetworks.com> This fixes commit 4d809f69695d4e ("IB/rdmavt: add lock to call to rvt_error_qp to prevent a race condition") and commit 22cbc6c2681a0a ("IB/rdmavt: add missing locks in rvt_ruc_loopback"). The 2 commits introduced ABBA hard lockup windows. The r_lock always must be grabbed before the s_lock. Simply swapping the order of grabbed locks before calling rvt_send_complete is not acceptable since 99% of the time rvt_send_complete will not need the r_lock, the completion queue is likely not full and therefore a call to rvt_error_qp will never happen. Grabbing both r_lock and s_lock before adding something to a completion queue is overkill. On the otherhand, reverting the above commits will allow for a possible lockdep issue. rvt_send/recv_cq CAN invoke rvt_error_qp. In those cases, they MUST have both r and s locks. It turns out that several callers of rvt_error_qp totally ignored this dependency. Therefore, undo the afformentioned commits AND implement a wrapper function around rvt_error_qp that forces the calling functions to specify what locks they hold. This forces all possible paths to be aware of their ownership of the r and s lock. The wrapper can then obtain the correct locks if it needs to. This is a necessary ugliness to continue to manage layered spinlocks in a function that is called from several possible codepaths. Note there are cases when callers only held the s lock. In order to prevent further ABBA spinlocks we must not attempt to grab the r lock while someone else holds it. If someone holds it we must drop the s lock and grab r then s. Fixes: 4d809f69695d4e ("IB/rdmavt: add lock to call to rvt_error_qp to prevent a race condition") Fixes: 22cbc6c2681a0a ("IB/rdmavt: add missing locks in rvt_ruc_loopback"). Cc: stable(a)vger.kernel.org Signed-off-by: Nick Child <nchild(a)cornelisnetworks.com> Signed-off-by: Dennis Dalessandro <dennis.dalessandro(a)cornelisnetworks.com> --- drivers/infiniband/hw/hfi1/qp.c | 3 + drivers/infiniband/hw/hfi1/rc.c | 47 ++++++++------ drivers/infiniband/hw/hfi1/tid_rdma.c | 16 +++-- drivers/infiniband/hw/hfi1/uc.c | 9 ++- drivers/infiniband/hw/hfi1/ud.c | 11 ++- drivers/infiniband/hw/hfi1/verbs.c | 9 ++- drivers/infiniband/hw/hfi1/verbs.h | 5 +- drivers/infiniband/sw/rdmavt/qp.c | 110 +++++++++++++++++++++++---------- include/rdma/rdmavt_qp.h | 35 ++++++++--- 9 files changed, 165 insertions(+), 80 deletions(-) diff --git a/drivers/infiniband/hw/hfi1/qp.c b/drivers/infiniband/hw/hfi1/qp.c index f3d8c0c193ac..d4e4fa1ce7a7 100644 --- a/drivers/infiniband/hw/hfi1/qp.c +++ b/drivers/infiniband/hw/hfi1/qp.c @@ -895,7 +895,8 @@ static void hfi1_qp_iter_cb(struct rvt_qp *qp, u64 v) spin_lock_irq(&qp->r_lock); spin_lock(&qp->s_hlock); spin_lock(&qp->s_lock); - lastwqe = rvt_error_qp(qp, IB_WC_WR_FLUSH_ERR); + lastwqe = rvt_error_qp(qp, IB_WC_WR_FLUSH_ERR, + RVT_QP_LOCK_STATE_RS); spin_unlock(&qp->s_lock); spin_unlock(&qp->s_hlock); spin_unlock_irq(&qp->r_lock); diff --git a/drivers/infiniband/hw/hfi1/rc.c b/drivers/infiniband/hw/hfi1/rc.c index b36242c9d42c..b3edd0030c7e 100644 --- a/drivers/infiniband/hw/hfi1/rc.c +++ b/drivers/infiniband/hw/hfi1/rc.c @@ -357,11 +357,7 @@ static int make_rc_ack(struct hfi1_ibdev *dev, struct rvt_qp *qp, return 1; error_qp: spin_unlock_irqrestore(&qp->s_lock, ps->flags); - spin_lock_irqsave(&qp->r_lock, ps->flags); - spin_lock(&qp->s_lock); - rvt_error_qp(qp, IB_WC_WR_FLUSH_ERR); - spin_unlock(&qp->s_lock); - spin_unlock_irqrestore(&qp->r_lock, ps->flags); + rvt_error_qp(qp, IB_WC_WR_FLUSH_ERR, RVT_QP_LOCK_STATE_NONE); spin_lock_irqsave(&qp->s_lock, ps->flags); bail: qp->s_ack_state = OP(ACKNOWLEDGE); @@ -448,7 +444,8 @@ int hfi1_make_rc_req(struct rvt_qp *qp, struct hfi1_pkt_state *ps) clear_ahg(qp); wqe = rvt_get_swqe_ptr(qp, qp->s_last); hfi1_trdma_send_complete(qp, wqe, qp->s_last != qp->s_acked ? - IB_WC_SUCCESS : IB_WC_WR_FLUSH_ERR); + IB_WC_SUCCESS : IB_WC_WR_FLUSH_ERR, + RVT_QP_LOCK_STATE_S); /* will get called again */ goto done_free_tx; } @@ -523,7 +520,8 @@ int hfi1_make_rc_req(struct rvt_qp *qp, struct hfi1_pkt_state *ps) } rvt_send_complete(qp, wqe, err ? IB_WC_LOC_PROT_ERR - : IB_WC_SUCCESS); + : IB_WC_SUCCESS, + RVT_QP_LOCK_STATE_S); if (local_ops) atomic_dec(&qp->local_ops_pending); goto done_free_tx; @@ -1063,7 +1061,8 @@ int hfi1_make_rc_req(struct rvt_qp *qp, struct hfi1_pkt_state *ps) hfi1_kern_exp_rcv_clear_all(req); hfi1_kern_clear_hw_flow(priv->rcd, qp); - hfi1_trdma_send_complete(qp, wqe, IB_WC_LOC_QP_OP_ERR); + hfi1_trdma_send_complete(qp, wqe, IB_WC_LOC_QP_OP_ERR, + RVT_QP_LOCK_STATE_S); goto bail; } req->state = TID_REQUEST_RESEND; @@ -1601,8 +1600,10 @@ void hfi1_restart_rc(struct rvt_qp *qp, u32 psn, int wait) } hfi1_trdma_send_complete(qp, wqe, - IB_WC_RETRY_EXC_ERR); - rvt_error_qp(qp, IB_WC_WR_FLUSH_ERR); + IB_WC_RETRY_EXC_ERR, + RVT_QP_LOCK_STATE_RS); + rvt_error_qp(qp, IB_WC_WR_FLUSH_ERR, + RVT_QP_LOCK_STATE_RS); } return; } else { /* need to handle delayed completion */ @@ -1795,7 +1796,7 @@ void hfi1_rc_send_complete(struct rvt_qp *qp, struct hfi1_opa_header *opah) rvt_qp_complete_swqe(qp, wqe, ib_hfi1_wc_opcode[wqe->wr.opcode], - IB_WC_SUCCESS); + IB_WC_SUCCESS, RVT_QP_LOCK_STATE_S); } /* * If we were waiting for sends to complete before re-sending, @@ -1827,6 +1828,7 @@ struct rvt_swqe *do_rc_completion(struct rvt_qp *qp, { struct hfi1_qp_priv *priv = qp->priv; + lockdep_assert_held(&qp->r_lock); lockdep_assert_held(&qp->s_lock); /* * Don't decrement refcount and don't generate a @@ -1838,10 +1840,10 @@ struct rvt_swqe *do_rc_completion(struct rvt_qp *qp, cmp_psn(qp->s_sending_psn, qp->s_sending_hpsn) > 0) { trdma_clean_swqe(qp, wqe); trace_hfi1_qp_send_completion(qp, wqe, qp->s_last); - rvt_qp_complete_swqe(qp, - wqe, + rvt_qp_complete_swqe(qp, wqe, ib_hfi1_wc_opcode[wqe->wr.opcode], - IB_WC_SUCCESS); + IB_WC_SUCCESS, + RVT_QP_LOCK_STATE_RS); } else { struct hfi1_pportdata *ppd = ppd_from_ibp(ibp); @@ -1958,7 +1960,7 @@ static void update_qp_retry_state(struct rvt_qp *qp, u32 psn, u32 spsn, * * This is called from rc_rcv_resp() to process an incoming RC ACK * for the given QP. - * May be called at interrupt level, with the QP s_lock held. + * May be called at interrupt level. Both r and s lock should be held. * Returns 1 if OK, 0 if current operation should be aborted (NAK). */ int do_rc_ack(struct rvt_qp *qp, u32 aeth, u32 psn, int opcode, @@ -1973,6 +1975,7 @@ int do_rc_ack(struct rvt_qp *qp, u32 aeth, u32 psn, int opcode, int diff; struct rvt_dev_info *rdi; + lockdep_assert_held(&qp->r_lock); lockdep_assert_held(&qp->s_lock); /* * Note that NAKs implicitly ACK outstanding SEND and RDMA write @@ -2232,8 +2235,10 @@ int do_rc_ack(struct rvt_qp *qp, u32 aeth, u32 psn, int opcode, if (wqe->wr.opcode == IB_WR_TID_RDMA_READ) hfi1_kern_read_tid_flow_free(qp); - hfi1_trdma_send_complete(qp, wqe, status); - rvt_error_qp(qp, IB_WC_WR_FLUSH_ERR); + hfi1_trdma_send_complete(qp, wqe, status, + RVT_QP_LOCK_STATE_RS); + rvt_error_qp(qp, IB_WC_WR_FLUSH_ERR, + RVT_QP_LOCK_STATE_RS); } break; @@ -2265,6 +2270,7 @@ static void rdma_seq_err(struct rvt_qp *qp, struct hfi1_ibport *ibp, u32 psn, { struct rvt_swqe *wqe; + lockdep_assert_held(&qp->r_lock); lockdep_assert_held(&qp->s_lock); /* Remove QP from retry timer */ rvt_stop_rc_timers(qp); @@ -2472,8 +2478,8 @@ static void rc_rcv_resp(struct hfi1_packet *packet) status = IB_WC_LOC_LEN_ERR; ack_err: if (qp->s_last == qp->s_acked) { - rvt_send_complete(qp, wqe, status); - rvt_error_qp(qp, IB_WC_WR_FLUSH_ERR); + rvt_send_complete(qp, wqe, status, RVT_QP_LOCK_STATE_RS); + rvt_error_qp(qp, IB_WC_WR_FLUSH_ERR, RVT_QP_LOCK_STATE_RS); } ack_done: spin_unlock_irqrestore(&qp->s_lock, flags); @@ -2963,7 +2969,8 @@ void hfi1_rc_rcv(struct hfi1_packet *packet) wc.dlid_path_bits = 0; wc.port_num = 0; /* Signal completion event if the solicited bit is set. */ - rvt_recv_cq(qp, &wc, ib_bth_is_solicited(ohdr)); + rvt_recv_cq(qp, &wc, ib_bth_is_solicited(ohdr), + RVT_QP_LOCK_STATE_R); break; case OP(RDMA_WRITE_ONLY): diff --git a/drivers/infiniband/hw/hfi1/tid_rdma.c b/drivers/infiniband/hw/hfi1/tid_rdma.c index eafd2f157e32..e7b58e7ad233 100644 --- a/drivers/infiniband/hw/hfi1/tid_rdma.c +++ b/drivers/infiniband/hw/hfi1/tid_rdma.c @@ -2569,7 +2569,7 @@ void hfi1_rc_rcv_tid_rdma_read_resp(struct hfi1_packet *packet) * all remaining requests. */ if (qp->s_last == qp->s_acked) - rvt_error_qp(qp, IB_WC_WR_FLUSH_ERR); + rvt_error_qp(qp, IB_WC_WR_FLUSH_ERR, RVT_QP_LOCK_STATE_RS); ack_done: spin_unlock_irqrestore(&qp->s_lock, flags); @@ -4195,13 +4195,14 @@ void hfi1_rc_rcv_tid_rdma_write_resp(struct hfi1_packet *packet) ack_op_err: status = IB_WC_LOC_QP_OP_ERR; ack_err: - rvt_error_qp(qp, status); + rvt_error_qp(qp, status, RVT_QP_LOCK_STATE_RS); ack_done: if (fecn) qp->s_flags |= RVT_S_ECN; spin_unlock_irqrestore(&qp->s_lock, flags); } +/* called with s_lock held */ bool hfi1_build_tid_rdma_packet(struct rvt_swqe *wqe, struct ib_other_headers *ohdr, u32 *bth1, u32 *bth2, u32 *len) @@ -4218,8 +4219,9 @@ bool hfi1_build_tid_rdma_packet(struct rvt_swqe *wqe, bool last_pkt; if (!tidlen) { - hfi1_trdma_send_complete(qp, wqe, IB_WC_REM_INV_RD_REQ_ERR); - rvt_error_qp(qp, IB_WC_REM_INV_RD_REQ_ERR); + hfi1_trdma_send_complete(qp, wqe, IB_WC_REM_INV_RD_REQ_ERR, + RVT_QP_LOCK_STATE_S); + rvt_error_qp(qp, IB_WC_REM_INV_RD_REQ_ERR, RVT_QP_LOCK_STATE_S); } *len = min_t(u32, qp->pmtu, tidlen - flow->tid_offset); @@ -4816,8 +4818,10 @@ static void hfi1_tid_retry_timeout(struct timer_list *t) (u64)priv->tid_retry_timeout_jiffies); wqe = rvt_get_swqe_ptr(qp, qp->s_acked); - hfi1_trdma_send_complete(qp, wqe, IB_WC_RETRY_EXC_ERR); - rvt_error_qp(qp, IB_WC_WR_FLUSH_ERR); + hfi1_trdma_send_complete(qp, wqe, IB_WC_RETRY_EXC_ERR, + RVT_QP_LOCK_STATE_RS); + rvt_error_qp(qp, IB_WC_WR_FLUSH_ERR, + RVT_QP_LOCK_STATE_RS); } else { wqe = rvt_get_swqe_ptr(qp, qp->s_acked); req = wqe_to_tid_req(wqe); diff --git a/drivers/infiniband/hw/hfi1/uc.c b/drivers/infiniband/hw/hfi1/uc.c index 33d2c2a218e2..e758aa55421b 100644 --- a/drivers/infiniband/hw/hfi1/uc.c +++ b/drivers/infiniband/hw/hfi1/uc.c @@ -47,7 +47,8 @@ int hfi1_make_uc_req(struct rvt_qp *qp, struct hfi1_pkt_state *ps) } clear_ahg(qp); wqe = rvt_get_swqe_ptr(qp, qp->s_last); - rvt_send_complete(qp, wqe, IB_WC_WR_FLUSH_ERR); + rvt_send_complete(qp, wqe, IB_WC_WR_FLUSH_ERR, + RVT_QP_LOCK_STATE_S); goto done_free_tx; } @@ -100,7 +101,8 @@ int hfi1_make_uc_req(struct rvt_qp *qp, struct hfi1_pkt_state *ps) local_ops = 1; } rvt_send_complete(qp, wqe, err ? IB_WC_LOC_PROT_ERR - : IB_WC_SUCCESS); + : IB_WC_SUCCESS, + RVT_QP_LOCK_STATE_S); if (local_ops) atomic_dec(&qp->local_ops_pending); goto done_free_tx; @@ -430,7 +432,8 @@ void hfi1_uc_rcv(struct hfi1_packet *packet) wc.dlid_path_bits = 0; wc.port_num = 0; /* Signal completion event if the solicited bit is set. */ - rvt_recv_cq(qp, &wc, ib_bth_is_solicited(ohdr)); + rvt_recv_cq(qp, &wc, ib_bth_is_solicited(ohdr), + RVT_QP_LOCK_STATE_R); break; case OP(RDMA_WRITE_FIRST): diff --git a/drivers/infiniband/hw/hfi1/ud.c b/drivers/infiniband/hw/hfi1/ud.c index 89d1bae8f824..1e54c9a2087b 100644 --- a/drivers/infiniband/hw/hfi1/ud.c +++ b/drivers/infiniband/hw/hfi1/ud.c @@ -213,7 +213,8 @@ static void ud_loopback(struct rvt_qp *sqp, struct rvt_swqe *swqe) wc.dlid_path_bits = rdma_ah_get_dlid(ah_attr) & ((1 << ppd->lmc) - 1); wc.port_num = qp->port_num; /* Signal completion event if the solicited bit is set. */ - rvt_recv_cq(qp, &wc, swqe->wr.send_flags & IB_SEND_SOLICITED); + rvt_recv_cq(qp, &wc, swqe->wr.send_flags & IB_SEND_SOLICITED, + RVT_QP_LOCK_STATE_R); ibp->rvp.n_loop_pkts++; bail_unlock: spin_unlock_irqrestore(&qp->r_lock, flags); @@ -458,7 +459,8 @@ int hfi1_make_ud_req(struct rvt_qp *qp, struct hfi1_pkt_state *ps) goto bail; } wqe = rvt_get_swqe_ptr(qp, qp->s_last); - rvt_send_complete(qp, wqe, IB_WC_WR_FLUSH_ERR); + rvt_send_complete(qp, wqe, IB_WC_WR_FLUSH_ERR, + RVT_QP_LOCK_STATE_S); goto done_free_tx; } @@ -500,7 +502,8 @@ int hfi1_make_ud_req(struct rvt_qp *qp, struct hfi1_pkt_state *ps) ud_loopback(qp, wqe); spin_lock_irqsave(&qp->s_lock, tflags); ps->flags = tflags; - rvt_send_complete(qp, wqe, IB_WC_SUCCESS); + rvt_send_complete(qp, wqe, IB_WC_SUCCESS, + RVT_QP_LOCK_STATE_S); goto done_free_tx; } } @@ -1015,7 +1018,7 @@ void hfi1_ud_rcv(struct hfi1_packet *packet) dlid & ((1 << ppd_from_ibp(ibp)->lmc) - 1); wc.port_num = qp->port_num; /* Signal completion event if the solicited bit is set. */ - rvt_recv_cq(qp, &wc, solicited); + rvt_recv_cq(qp, &wc, solicited, RVT_QP_LOCK_STATE_R); return; drop: diff --git a/drivers/infiniband/hw/hfi1/verbs.c b/drivers/infiniband/hw/hfi1/verbs.c index 3cbbfccdd8cd..c7743e17cb33 100644 --- a/drivers/infiniband/hw/hfi1/verbs.c +++ b/drivers/infiniband/hw/hfi1/verbs.c @@ -592,7 +592,8 @@ static void verbs_sdma_complete( spin_lock(&qp->s_lock); if (tx->wqe) { - rvt_send_complete(qp, tx->wqe, IB_WC_SUCCESS); + rvt_send_complete(qp, tx->wqe, IB_WC_SUCCESS, + RVT_QP_LOCK_STATE_S); } else if (qp->ibqp.qp_type == IB_QPT_RC) { struct hfi1_opa_header *hdr; @@ -1060,7 +1061,8 @@ int hfi1_verbs_send_pio(struct rvt_qp *qp, struct hfi1_pkt_state *ps, pio_bail: spin_lock_irqsave(&qp->s_lock, flags); if (qp->s_wqe) { - rvt_send_complete(qp, qp->s_wqe, wc_status); + rvt_send_complete(qp, qp->s_wqe, wc_status, + RVT_QP_LOCK_STATE_S); } else if (qp->ibqp.qp_type == IB_QPT_RC) { if (unlikely(wc_status == IB_WC_GENERAL_ERR)) hfi1_rc_verbs_aborted(qp, &ps->s_txreq->phdr.hdr); @@ -1268,7 +1270,8 @@ int hfi1_verbs_send(struct rvt_qp *qp, struct hfi1_pkt_state *ps) hfi1_cdbg(PIO, "%s() Failed. Completing with err", __func__); spin_lock_irqsave(&qp->s_lock, flags); - rvt_send_complete(qp, qp->s_wqe, IB_WC_GENERAL_ERR); + rvt_send_complete(qp, qp->s_wqe, IB_WC_GENERAL_ERR, + RVT_QP_LOCK_STATE_S); spin_unlock_irqrestore(&qp->s_lock, flags); } return -EINVAL; diff --git a/drivers/infiniband/hw/hfi1/verbs.h b/drivers/infiniband/hw/hfi1/verbs.h index 070e4f0babe8..a6d3023adcc8 100644 --- a/drivers/infiniband/hw/hfi1/verbs.h +++ b/drivers/infiniband/hw/hfi1/verbs.h @@ -446,10 +446,11 @@ void hfi1_wait_kmem(struct rvt_qp *qp); static inline void hfi1_trdma_send_complete(struct rvt_qp *qp, struct rvt_swqe *wqe, - enum ib_wc_status status) + enum ib_wc_status status, + enum rvt_qp_lock_state lock_state) { trdma_clean_swqe(qp, wqe); - rvt_send_complete(qp, wqe, status); + rvt_send_complete(qp, wqe, status, lock_state); } extern const enum ib_wc_opcode ib_hfi1_wc_opcode[]; diff --git a/drivers/infiniband/sw/rdmavt/qp.c b/drivers/infiniband/sw/rdmavt/qp.c index e825e2ef7966..d6d314de2930 100644 --- a/drivers/infiniband/sw/rdmavt/qp.c +++ b/drivers/infiniband/sw/rdmavt/qp.c @@ -703,7 +703,8 @@ void rvt_qp_mr_clean(struct rvt_qp *qp, u32 lkey) if (rvt_ss_has_lkey(&qp->r_sge, lkey) || rvt_qp_sends_has_lkey(qp, lkey) || rvt_qp_acks_has_lkey(qp, lkey)) - lastwqe = rvt_error_qp(qp, IB_WC_LOC_PROT_ERR); + lastwqe = rvt_error_qp(qp, IB_WC_LOC_PROT_ERR, + RVT_QP_LOCK_STATE_RS); check_lwqe: spin_unlock(&qp->s_lock); spin_unlock(&qp->s_hlock); @@ -1271,18 +1272,7 @@ int rvt_create_qp(struct ib_qp *ibqp, struct ib_qp_init_attr *init_attr, return ret; } -/** - * rvt_error_qp - put a QP into the error state - * @qp: the QP to put into the error state - * @err: the receive completion error to signal if a RWQE is active - * - * Flushes both send and receive work queues. - * - * Return: true if last WQE event should be generated. - * The QP r_lock and s_lock should be held and interrupts disabled. - * If we are already in error state, just return. - */ -int rvt_error_qp(struct rvt_qp *qp, enum ib_wc_status err) +static int __rvt_error_qp_locked(struct rvt_qp *qp, enum ib_wc_status err) { struct ib_wc wc; int ret = 0; @@ -1362,6 +1352,64 @@ int rvt_error_qp(struct rvt_qp *qp, enum ib_wc_status err) bail: return ret; } + +/** + * rvt_error_qp - put a QP into the error state + * @qp: the QP to put into the error state + * @err: the receive completion error to signal if a RWQE is active + * @lock_state: caller ownership representation of r and s lock + * + * Flushes both send and receive work queues. + * + * Return: true if last WQE event should be generated. + * The QP r_lock and s_lock should be held and interrupts disabled. + * If we are already in error state, just return. + */ +int rvt_error_qp(struct rvt_qp *qp, enum ib_wc_status err, + enum rvt_qp_lock_state lock_state) +{ + int ret; + + switch (lock_state) { + case RVT_QP_LOCK_STATE_NONE: + unsigned long flags; + + /* only case where caller may not have handled irqs */ + spin_lock_irqsave(&qp->r_lock, flags); + spin_lock(&qp->s_lock); + ret = __rvt_error_qp_locked(qp, err); + spin_unlock(&qp->s_lock); + spin_unlock_irqrestore(&qp->r_lock, flags); + break; + + case RVT_QP_LOCK_STATE_S: + /* r_lock -> s_lock ordering can be broken here iff + * no one else is using the r_lock at the moment + */ + if (!spin_trylock(&qp->r_lock)) { + /* otherwise we must respect ordering */ + spin_unlock(&qp->s_lock); + spin_lock(&qp->r_lock); + spin_lock(&qp->s_lock); + } + ret = __rvt_error_qp_locked(qp, err); + spin_unlock(&qp->r_lock); + break; + + case RVT_QP_LOCK_STATE_R: + spin_lock(&qp->s_lock); + ret = __rvt_error_qp_locked(qp, err); + spin_unlock(&qp->s_lock); + break; + + case RVT_QP_LOCK_STATE_RS: + fallthrough; + default: + ret = __rvt_error_qp_locked(qp, err); + } + + return ret; +} EXPORT_SYMBOL(rvt_error_qp); /* @@ -1549,7 +1597,8 @@ int rvt_modify_qp(struct ib_qp *ibqp, struct ib_qp_attr *attr, break; case IB_QPS_ERR: - lastwqe = rvt_error_qp(qp, IB_WC_WR_FLUSH_ERR); + lastwqe = rvt_error_qp(qp, IB_WC_WR_FLUSH_ERR, + RVT_QP_LOCK_STATE_RS); break; default: @@ -2460,15 +2509,13 @@ void rvt_comm_est(struct rvt_qp *qp) } EXPORT_SYMBOL(rvt_comm_est); +/* assumes r_lock is held */ void rvt_rc_error(struct rvt_qp *qp, enum ib_wc_status err) { - unsigned long flags; int lastwqe; - spin_lock_irqsave(&qp->s_lock, flags); - lastwqe = rvt_error_qp(qp, err); - spin_unlock_irqrestore(&qp->s_lock, flags); - + lockdep_assert_held(&qp->r_lock); + lastwqe = rvt_error_qp(qp, err, RVT_QP_LOCK_STATE_R); if (lastwqe) { struct ib_event ev; @@ -2772,10 +2819,11 @@ void rvt_qp_iter(struct rvt_dev_info *rdi, EXPORT_SYMBOL(rvt_qp_iter); /* - * This should be called with s_lock and r_lock held. + * This should be called with s_lock held. */ void rvt_send_complete(struct rvt_qp *qp, struct rvt_swqe *wqe, - enum ib_wc_status status) + enum ib_wc_status status, + enum rvt_qp_lock_state lock_state) { u32 old_last, last; struct rvt_dev_info *rdi; @@ -2787,7 +2835,7 @@ void rvt_send_complete(struct rvt_qp *qp, struct rvt_swqe *wqe, old_last = qp->s_last; trace_rvt_qp_send_completion(qp, wqe, old_last); last = rvt_qp_complete_swqe(qp, wqe, rdi->wc_opcode[wqe->wr.opcode], - status); + status, lock_state); if (qp->s_acked == old_last) qp->s_acked = last; if (qp->s_cur == old_last) @@ -3123,7 +3171,8 @@ void rvt_ruc_loopback(struct rvt_qp *sqp) wc.sl = rdma_ah_get_sl(&qp->remote_ah_attr); wc.port_num = 1; /* Signal completion event if the solicited bit is set. */ - rvt_recv_cq(qp, &wc, wqe->wr.send_flags & IB_SEND_SOLICITED); + rvt_recv_cq(qp, &wc, wqe->wr.send_flags & IB_SEND_SOLICITED, + RVT_QP_LOCK_STATE_R); send_comp: spin_unlock_irqrestore(&qp->r_lock, flags); @@ -3131,9 +3180,7 @@ void rvt_ruc_loopback(struct rvt_qp *sqp) rvp->n_loop_pkts++; flush_send: sqp->s_rnr_retry = sqp->s_rnr_retry_cnt; - spin_lock(&sqp->r_lock); - rvt_send_complete(sqp, wqe, send_status); - spin_unlock(&sqp->r_lock); + rvt_send_complete(sqp, wqe, send_status, RVT_QP_LOCK_STATE_S); if (local_ops) { atomic_dec(&sqp->local_ops_pending); local_ops = 0; @@ -3187,18 +3234,15 @@ void rvt_ruc_loopback(struct rvt_qp *sqp) spin_unlock_irqrestore(&qp->r_lock, flags); serr_no_r_lock: spin_lock_irqsave(&sqp->s_lock, flags); - spin_lock(&sqp->r_lock); - rvt_send_complete(sqp, wqe, send_status); - spin_unlock(&sqp->r_lock); + rvt_send_complete(sqp, wqe, send_status, RVT_QP_LOCK_STATE_S); if (sqp->ibqp.qp_type == IB_QPT_RC) { int lastwqe; - spin_lock(&sqp->r_lock); - lastwqe = rvt_error_qp(sqp, IB_WC_WR_FLUSH_ERR); - spin_unlock(&sqp->r_lock); - + lastwqe = rvt_error_qp(sqp, IB_WC_WR_FLUSH_ERR, + RVT_QP_LOCK_STATE_S); sqp->s_flags &= ~RVT_S_BUSY; spin_unlock_irqrestore(&sqp->s_lock, flags); + if (lastwqe) { struct ib_event ev; diff --git a/include/rdma/rdmavt_qp.h b/include/rdma/rdmavt_qp.h index d67892944193..35339bbe39cc 100644 --- a/include/rdma/rdmavt_qp.h +++ b/include/rdma/rdmavt_qp.h @@ -137,6 +137,16 @@ #define RVT_SEND_OR_FLUSH_OR_RECV_OK \ (RVT_PROCESS_SEND_OK | RVT_FLUSH_SEND | RVT_PROCESS_RECV_OK) +/* + * Internal relay of held Queue Pair lock state + */ +enum rvt_qp_lock_state { + RVT_QP_LOCK_STATE_NONE = 0, + RVT_QP_LOCK_STATE_S, + RVT_QP_LOCK_STATE_R, + RVT_QP_LOCK_STATE_RS +}; + /* * Internal send flags */ @@ -767,7 +777,8 @@ rvt_qp_swqe_incr(struct rvt_qp *qp, u32 val) return val; } -int rvt_error_qp(struct rvt_qp *qp, enum ib_wc_status err); +int rvt_error_qp(struct rvt_qp *qp, enum ib_wc_status err, + enum rvt_qp_lock_state lock_state); /** * rvt_recv_cq - add a new entry to completion queue @@ -775,18 +786,20 @@ int rvt_error_qp(struct rvt_qp *qp, enum ib_wc_status err); * @qp: receive queue * @wc: work completion entry to add * @solicited: true if @entry is solicited + * @lock_state: caller ownership representation of r and s lock * * This is wrapper function for rvt_enter_cq function call by * receive queue. If rvt_cq_enter return false, it means cq is * full and the qp is put into error state. */ static inline void rvt_recv_cq(struct rvt_qp *qp, struct ib_wc *wc, - bool solicited) + bool solicited, + enum rvt_qp_lock_state lock_state) { struct rvt_cq *cq = ibcq_to_rvtcq(qp->ibqp.recv_cq); if (unlikely(!rvt_cq_enter(cq, wc, solicited))) - rvt_error_qp(qp, IB_WC_LOC_QP_OP_ERR); + rvt_error_qp(qp, IB_WC_LOC_QP_OP_ERR, lock_state); } /** @@ -795,18 +808,20 @@ static inline void rvt_recv_cq(struct rvt_qp *qp, struct ib_wc *wc, * @qp: send queue * @wc: work completion entry to add * @solicited: true if @entry is solicited + * @lock_state: caller ownership representation of r and s lock * * This is wrapper function for rvt_enter_cq function call by * send queue. If rvt_cq_enter return false, it means cq is * full and the qp is put into error state. */ static inline void rvt_send_cq(struct rvt_qp *qp, struct ib_wc *wc, - bool solicited) + bool solicited, + enum rvt_qp_lock_state lock_state) { struct rvt_cq *cq = ibcq_to_rvtcq(qp->ibqp.send_cq); if (unlikely(!rvt_cq_enter(cq, wc, solicited))) - rvt_error_qp(qp, IB_WC_LOC_QP_OP_ERR); + rvt_error_qp(qp, IB_WC_LOC_QP_OP_ERR, lock_state); } /** @@ -829,13 +844,15 @@ static inline u32 rvt_qp_complete_swqe(struct rvt_qp *qp, struct rvt_swqe *wqe, enum ib_wc_opcode opcode, - enum ib_wc_status status) + enum ib_wc_status status, + enum rvt_qp_lock_state lock_state) { bool need_completion; u64 wr_id; u32 byte_len, last; int flags = wqe->wr.send_flags; + lockdep_assert_held(&qp->s_lock); rvt_qp_wqe_unreserve(qp, flags); rvt_put_qp_swqe(qp, wqe); @@ -860,7 +877,8 @@ rvt_qp_complete_swqe(struct rvt_qp *qp, .qp = &qp->ibqp, .byte_len = byte_len, }; - rvt_send_cq(qp, &w, status != IB_WC_SUCCESS); + rvt_send_cq(qp, &w, status != IB_WC_SUCCESS, + lock_state); } return last; } @@ -886,7 +904,8 @@ void rvt_copy_sge(struct rvt_qp *qp, struct rvt_sge_state *ss, void *data, u32 length, bool release, bool copy_last); void rvt_send_complete(struct rvt_qp *qp, struct rvt_swqe *wqe, - enum ib_wc_status status); + enum ib_wc_status status, + enum rvt_qp_lock_state lock_state); void rvt_ruc_loopback(struct rvt_qp *qp); /**

2 months, 3 weeks

2
1
0 0

2025

2024

2023

2022

2021

2020

2019

2018

2017

Linux-stable-mirror