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

[PATCH 6.6] drm/amd/display: Don't refer to dc_sink in is_dsc_need_re_compute

by Bin Lan

From: Wayne Lin <wayne.lin(a)amd.com> [ Upstream commit fcf6a49d79923a234844b8efe830a61f3f0584e4 ] [Why] When unplug one of monitors connected after mst hub, encounter null pointer dereference. It's due to dc_sink get released immediately in early_unregister() or detect_ctx(). When commit new state which directly referring to info stored in dc_sink will cause null pointer dereference. [how] Remove redundant checking condition. Relevant condition should already be covered by checking if dsc_aux is null or not. Also reset dsc_aux to NULL when the connector is disconnected. Reviewed-by: Jerry Zuo <jerry.zuo(a)amd.com> Acked-by: Zaeem Mohamed <zaeem.mohamed(a)amd.com> Signed-off-by: Wayne Lin <wayne.lin(a)amd.com> Tested-by: Daniel Wheeler <daniel.wheeler(a)amd.com> Signed-off-by: Alex Deucher <alexander.deucher(a)amd.com> [ Resolve minor conflicts ] Signed-off-by: Bin Lan <bin.lan.cn(a)windriver.com> --- drivers/gpu/drm/amd/display/amdgpu_dm/amdgpu_dm_mst_types.c | 4 ++++ 1 file changed, 4 insertions(+) diff --git a/drivers/gpu/drm/amd/display/amdgpu_dm/amdgpu_dm_mst_types.c b/drivers/gpu/drm/amd/display/amdgpu_dm/amdgpu_dm_mst_types.c index d390e3d62e56..9ec9792f115a 100644 --- a/drivers/gpu/drm/amd/display/amdgpu_dm/amdgpu_dm_mst_types.c +++ b/drivers/gpu/drm/amd/display/amdgpu_dm/amdgpu_dm_mst_types.c @@ -179,6 +179,8 @@ amdgpu_dm_mst_connector_early_unregister(struct drm_connector *connector) dc_sink_release(dc_sink); aconnector->dc_sink = NULL; aconnector->edid = NULL; + aconnector->dsc_aux = NULL; + port->passthrough_aux = NULL; } aconnector->mst_status = MST_STATUS_DEFAULT; @@ -487,6 +489,8 @@ dm_dp_mst_detect(struct drm_connector *connector, dc_sink_release(aconnector->dc_sink); aconnector->dc_sink = NULL; aconnector->edid = NULL; + aconnector->dsc_aux = NULL; + port->passthrough_aux = NULL; amdgpu_dm_set_mst_status(&aconnector->mst_status, MST_REMOTE_EDID | MST_ALLOCATE_NEW_PAYLOAD | MST_CLEAR_ALLOCATED_PAYLOAD, -- 2.43.0

2 months, 2 weeks

2
1
0 0

[PATCH 3/3] drm/xe/guc_submit: fix race around suspend_pending

by Matthew Auld

Currently in some testcases we can trigger: xe 0000:03:00.0: [drm] Assertion `exec_queue_destroyed(q)` failed! .... WARNING: CPU: 18 PID: 2640 at drivers/gpu/drm/xe/xe_guc_submit.c:1826 xe_guc_sched_done_handler+0xa54/0xef0 [xe] xe 0000:03:00.0: [drm] *ERROR* GT1: DEREGISTER_DONE: Unexpected engine state 0x00a1, guc_id=57 Looking at a snippet of corresponding ftrace for this GuC id we can see: 162.673311: xe_sched_msg_add: dev=0000:03:00.0, gt=1 guc_id=57, opcode=3 162.673317: xe_sched_msg_recv: dev=0000:03:00.0, gt=1 guc_id=57, opcode=3 162.673319: xe_exec_queue_scheduling_disable: dev=0000:03:00.0, 1:0x2, gt=1, width=1, guc_id=57, guc_state=0x29, flags=0x0 162.674089: xe_exec_queue_kill: dev=0000:03:00.0, 1:0x2, gt=1, width=1, guc_id=57, guc_state=0x29, flags=0x0 162.674108: xe_exec_queue_close: dev=0000:03:00.0, 1:0x2, gt=1, width=1, guc_id=57, guc_state=0xa9, flags=0x0 162.674488: xe_exec_queue_scheduling_done: dev=0000:03:00.0, 1:0x2, gt=1, width=1, guc_id=57, guc_state=0xa9, flags=0x0 162.678452: xe_exec_queue_deregister: dev=0000:03:00.0, 1:0x2, gt=1, width=1, guc_id=57, guc_state=0xa1, flags=0x0 It looks like we try to suspend the queue (opcode=3), setting suspend_pending and triggering a disable_scheduling. The user then closes the queue. However closing the queue seems to forcefully signal the fence after killing the queue, however when the G2H response for disable_scheduling comes back we have now cleared suspend_pending when signalling the suspend fence, so the disable_scheduling now incorrectly tries to also deregister the queue, leading to warnings since the queue has yet to even be marked for destruction. We also seem to trigger errors later with trying to double unregister the same queue. To fix this tweak the ordering when handling the response to ensure we don't race with a disable_scheduling that doesn't actually intend to actually unregister. The destruction path should now also correctly wait for any pending_disable before marking as destroyed. Fixes: dd08ebf6c352 ("drm/xe: Introduce a new DRM driver for Intel GPUs") Closes: https://gitlab.freedesktop.org/drm/xe/kernel/-/issues/3371 Signed-off-by: Matthew Auld <matthew.auld(a)intel.com> Cc: Matthew Brost <matthew.brost(a)intel.com> Cc: <stable(a)vger.kernel.org> # v6.8+ --- drivers/gpu/drm/xe/xe_guc_submit.c | 18 ++++++++++++++++-- 1 file changed, 16 insertions(+), 2 deletions(-) diff --git a/drivers/gpu/drm/xe/xe_guc_submit.c b/drivers/gpu/drm/xe/xe_guc_submit.c index f3c22b101916..f82f286fd431 100644 --- a/drivers/gpu/drm/xe/xe_guc_submit.c +++ b/drivers/gpu/drm/xe/xe_guc_submit.c @@ -1867,16 +1867,30 @@ static void handle_sched_done(struct xe_guc *guc, struct xe_exec_queue *q, xe_gt_assert(guc_to_gt(guc), runnable_state == 0); xe_gt_assert(guc_to_gt(guc), exec_queue_pending_disable(q)); - clear_exec_queue_pending_disable(q); if (q->guc->suspend_pending) { suspend_fence_signal(q); + clear_exec_queue_pending_disable(q); } else { if (exec_queue_banned(q) || check_timeout) { smp_wmb(); wake_up_all(&guc->ct.wq); } - if (!check_timeout) + if (!check_timeout && exec_queue_destroyed(q)) { + /* + * Make sure we clear the pending_disable only + * after the sampling the destroyed state. We + * want to ensure we don't trigger the + * unregister too early with something only + * intending to only disable scheduling. The + * caller doing the destroy must wait for an + * ongoing pending_destroy before marking as + * destroyed. + */ + clear_exec_queue_pending_disable(q); deregister_exec_queue(guc, q); + } else { + clear_exec_queue_pending_disable(q); + } } } } -- 2.47.0

2 months, 2 weeks

2
1
0 0

[PATCH] scsi: ufs: core: sysfs: Prevent div by zero

by Gwendal Grignou

Prevent a division by 0 when monitoring is not enabled. Fixes: 1d8613a23f3c ("scsi: ufs: core: Introduce HBA performance monitor sysfs nodes") Cc: stable(a)vger.kernel.org Signed-off-by: Gwendal Grignou <gwendal(a)chromium.org> --- drivers/ufs/core/ufs-sysfs.c | 6 ++++++ 1 file changed, 6 insertions(+) diff --git a/drivers/ufs/core/ufs-sysfs.c b/drivers/ufs/core/ufs-sysfs.c index c95906443d5f9..3692b39b35e78 100644 --- a/drivers/ufs/core/ufs-sysfs.c +++ b/drivers/ufs/core/ufs-sysfs.c @@ -485,6 +485,9 @@ static ssize_t read_req_latency_avg_show(struct device *dev, struct ufs_hba *hba = dev_get_drvdata(dev); struct ufs_hba_monitor *m = &hba->monitor; + if (!m->nr_req[READ]) + return sysfs_emit(buf, "0\n"); + return sysfs_emit(buf, "%llu\n", div_u64(ktime_to_us(m->lat_sum[READ]), m->nr_req[READ])); } @@ -552,6 +555,9 @@ static ssize_t write_req_latency_avg_show(struct device *dev, struct ufs_hba *hba = dev_get_drvdata(dev); struct ufs_hba_monitor *m = &hba->monitor; + if (!m->nr_req[WRITE]) + return sysfs_emit(buf, "0\n"); + return sysfs_emit(buf, "%llu\n", div_u64(ktime_to_us(m->lat_sum[WRITE]), m->nr_req[WRITE])); } -- 2.47.0.338.g60cca15819-goog

2 months, 2 weeks

3
4
0 0

[PATCH] drm/dp_mst: Fix MST sideband message body length check

by Imre Deak

Fix the MST sideband message body length check, which must be at least 1 byte accounting for the message body CRC (aka message data CRC) at the end of the message. This fixes a case where an MST branch device returns a header with a correct header CRC (indicating a correctly received body length), with the body length being incorrectly set to 0. This will later lead to a memory corruption in drm_dp_sideband_append_payload() and the following errors in dmesg: UBSAN: array-index-out-of-bounds in drivers/gpu/drm/display/drm_dp_mst_topology.c:786:25 index -1 is out of range for type 'u8 [48]' Call Trace: drm_dp_sideband_append_payload+0x33d/0x350 [drm_display_helper] drm_dp_get_one_sb_msg+0x3ce/0x5f0 [drm_display_helper] drm_dp_mst_hpd_irq_handle_event+0xc8/0x1580 [drm_display_helper] memcpy: detected field-spanning write (size 18446744073709551615) of single field "&msg->msg[msg->curlen]" at drivers/gpu/drm/display/drm_dp_mst_topology.c:791 (size 256) Call Trace: drm_dp_sideband_append_payload+0x324/0x350 [drm_display_helper] drm_dp_get_one_sb_msg+0x3ce/0x5f0 [drm_display_helper] drm_dp_mst_hpd_irq_handle_event+0xc8/0x1580 [drm_display_helper] Cc: <stable(a)vger.kernel.org> Cc: Lyude Paul <lyude(a)redhat.com> Signed-off-by: Imre Deak <imre.deak(a)intel.com> --- drivers/gpu/drm/display/drm_dp_mst_topology.c | 3 +++ 1 file changed, 3 insertions(+) diff --git a/drivers/gpu/drm/display/drm_dp_mst_topology.c b/drivers/gpu/drm/display/drm_dp_mst_topology.c index ac90118b9e7a8..e6ee180815b20 100644 --- a/drivers/gpu/drm/display/drm_dp_mst_topology.c +++ b/drivers/gpu/drm/display/drm_dp_mst_topology.c @@ -320,6 +320,9 @@ static bool drm_dp_decode_sideband_msg_hdr(const struct drm_dp_mst_topology_mgr hdr->broadcast = (buf[idx] >> 7) & 0x1; hdr->path_msg = (buf[idx] >> 6) & 0x1; hdr->msg_len = buf[idx] & 0x3f; + if (hdr->msg_len < 1) /* min space for body CRC */ + return false; + idx++; hdr->somt = (buf[idx] >> 7) & 0x1; hdr->eomt = (buf[idx] >> 6) & 0x1; -- 2.44.2

2 months, 2 weeks

2
1
0 0

[PATCH 1/2] mm: Open-code PageTail in folio_flags() and const_folio_flags()

by Matthew Wilcox (Oracle)

It is unsafe to call PageTail() in dump_page() as page_is_fake_head() will almost certainly return true when called on a head page that is copied to the stack. That will cause the VM_BUG_ON_PGFLAGS() in const_folio_flags() to trigger when it shouldn't. Fortunately, we don't need to call PageTail() here; it's fine to have a pointer to a virtual alias of the page's flag word rather than the real page's flag word. Fixes: fae7d834c43c (mm: add __dump_folio()) Signed-off-by: Matthew Wilcox (Oracle) <willy(a)infradead.org> Cc: stable(a)vger.kernel.org --- include/linux/page-flags.h | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-) diff --git a/include/linux/page-flags.h b/include/linux/page-flags.h index 2220bfec278e..cf46ac720802 100644 --- a/include/linux/page-flags.h +++ b/include/linux/page-flags.h @@ -306,7 +306,7 @@ static const unsigned long *const_folio_flags(const struct folio *folio, { const struct page *page = &folio->page; - VM_BUG_ON_PGFLAGS(PageTail(page), page); + VM_BUG_ON_PGFLAGS(page->compound_head & 1, page); VM_BUG_ON_PGFLAGS(n > 0 && !test_bit(PG_head, &page->flags), page); return &page[n].flags; } @@ -315,7 +315,7 @@ static unsigned long *folio_flags(struct folio *folio, unsigned n) { struct page *page = &folio->page; - VM_BUG_ON_PGFLAGS(PageTail(page), page); + VM_BUG_ON_PGFLAGS(page->compound_head & 1, page); VM_BUG_ON_PGFLAGS(n > 0 && !test_bit(PG_head, &page->flags), page); return &page[n].flags; } -- 2.45.2

2 months, 2 weeks

1
1
0 0

[PATCH v4 0/7] media: uvcvideo: Implement the Privacy GPIO as a evdev

by Ricardo Ribalda

Some notebooks have a button to disable the camera (not to be mistaken with the mechanical cover). This is a standard GPIO linked to the camera via the ACPI table. 4 years ago we added support for this button in UVC via the Privacy control. This has three issues: - If the camera has its own privacy control, it will be masked. - We need to power-up the camera to read the privacy control gpio. - Other drivers have not followed this approach and have used evdev. We tried to fix the power-up issues implementing "granular power saving" but it has been more complicated than anticipated... This patchset implements the Privacy GPIO as a evdev. The first patch of this set is already in Laurent's tree... but I include it to get some CI coverage. Signed-off-by: Ricardo Ribalda <ribalda(a)chromium.org> --- Changes in v4: - Remove gpio entity, it is not needed. - Use unit->gpio.irq in free_irq to make smatch happy. - Link to v3: https://lore.kernel.org/r/20241112-uvc-subdev-v3-0-0ea573d41a18@chromium.org Changes in v3: - CodeStyle (Thanks Sakari) - Re-implement as input device - Make the code depend on UVC_INPUT_EVDEV - Link to v2: https://lore.kernel.org/r/20241108-uvc-subdev-v2-0-85d8a051a3d3@chromium.org Changes in v2: - Rebase on top of https://patchwork.linuxtv.org/project/linux-media/patch/20241106-uvc-crashr… - Create uvc_gpio_cleanup and uvc_gpio_deinit - Refactor quirk: do not disable irq - Change define number for MEDIA_ENT_F_GPIO - Link to v1: https://lore.kernel.org/r/20241031-uvc-subdev-v1-0-a68331cedd72@chromium.org --- Ricardo Ribalda (7): media: uvcvideo: Fix crash during unbind if gpio unit is in use media: uvcvideo: Factor out gpio functions to its own file media: uvcvideo: Re-implement privacy GPIO as an input device Revert "media: uvcvideo: Allow entity-defined get_info and get_cur" media: uvcvideo: Introduce UVC_QUIRK_PRIVACY_DURING_STREAM media: uvcvideo: Make gpio_unit entity-less media: uvcvideo: Remove UVC_EXT_GPIO entity drivers/media/usb/uvc/Kconfig | 2 +- drivers/media/usb/uvc/Makefile | 3 + drivers/media/usb/uvc/uvc_ctrl.c | 40 ++--------- drivers/media/usb/uvc/uvc_driver.c | 123 ++-------------------------------- drivers/media/usb/uvc/uvc_entity.c | 7 +- drivers/media/usb/uvc/uvc_gpio.c | 134 +++++++++++++++++++++++++++++++++++++ drivers/media/usb/uvc/uvc_status.c | 13 +++- drivers/media/usb/uvc/uvc_video.c | 4 ++ drivers/media/usb/uvc/uvcvideo.h | 43 +++++++----- 9 files changed, 195 insertions(+), 174 deletions(-) --- base-commit: 72ad4ff638047bbbdf3232178fea4bec1f429319 change-id: 20241030-uvc-subdev-89f4467a00b5 Best regards, -- Ricardo Ribalda <ribalda(a)chromium.org>

2 months, 2 weeks

1
1
0 0

[PATCH 5.15] Revert "drivers: clk: zynqmp: update divider round rate logic"

by jguittet.opensource＠witekio.com

From: Joel Guittet <jguittet(a)witekio.com> This reverts commit 9117fc44fd3a9538261e530ba5a022dfc9519620 which is commit 1fe15be1fb613534ecbac5f8c3f8744f757d237d upstream. It is reported to cause regressions in the 5.15.y tree, so revert it for now. Link: https://www.spinics.net/lists/kernel/msg5397998.html Signed-off-by: Joel Guittet <jguittet.opensource(a)witekio.com> --- drivers/clk/zynqmp/divider.c | 66 +++++++++++++++++++++++++++++++++--- 1 file changed, 61 insertions(+), 5 deletions(-) diff --git a/drivers/clk/zynqmp/divider.c b/drivers/clk/zynqmp/divider.c index e25c76ff2739..47a199346ddf 100644 --- a/drivers/clk/zynqmp/divider.c +++ b/drivers/clk/zynqmp/divider.c @@ -110,6 +110,52 @@ static unsigned long zynqmp_clk_divider_recalc_rate(struct clk_hw *hw, return DIV_ROUND_UP_ULL(parent_rate, value); } +static void zynqmp_get_divider2_val(struct clk_hw *hw, + unsigned long rate, + struct zynqmp_clk_divider *divider, + u32 *bestdiv) +{ + int div1; + int div2; + long error = LONG_MAX; + unsigned long div1_prate; + struct clk_hw *div1_parent_hw; + struct zynqmp_clk_divider *pdivider; + struct clk_hw *div2_parent_hw = clk_hw_get_parent(hw); + + if (!div2_parent_hw) + return; + + pdivider = to_zynqmp_clk_divider(div2_parent_hw); + if (!pdivider) + return; + + div1_parent_hw = clk_hw_get_parent(div2_parent_hw); + if (!div1_parent_hw) + return; + + div1_prate = clk_hw_get_rate(div1_parent_hw); + *bestdiv = 1; + for (div1 = 1; div1 <= pdivider->max_div;) { + for (div2 = 1; div2 <= divider->max_div;) { + long new_error = ((div1_prate / div1) / div2) - rate; + + if (abs(new_error) < abs(error)) { + *bestdiv = div2; + error = new_error; + } + if (divider->flags & CLK_DIVIDER_POWER_OF_TWO) + div2 = div2 << 1; + else + div2++; + } + if (pdivider->flags & CLK_DIVIDER_POWER_OF_TWO) + div1 = div1 << 1; + else + div1++; + } +} + /** * zynqmp_clk_divider_round_rate() - Round rate of divider clock * @hw: handle between common and hardware-specific interfaces @@ -128,7 +174,6 @@ static long zynqmp_clk_divider_round_rate(struct clk_hw *hw, u32 div_type = divider->div_type; u32 bestdiv; int ret; - u8 width; /* if read only, just return current value */ if (divider->flags & CLK_DIVIDER_READ_ONLY) { @@ -148,12 +193,23 @@ static long zynqmp_clk_divider_round_rate(struct clk_hw *hw, return DIV_ROUND_UP_ULL((u64)*prate, bestdiv); } - width = fls(divider->max_div); + bestdiv = zynqmp_divider_get_val(*prate, rate, divider->flags); + + /* + * In case of two divisors, compute best divider values and return + * divider2 value based on compute value. div1 will be automatically + * set to optimum based on required total divider value. + */ + if (div_type == TYPE_DIV2 && + (clk_hw_get_flags(hw) & CLK_SET_RATE_PARENT)) { + zynqmp_get_divider2_val(hw, rate, divider, &bestdiv); + } - rate = divider_round_rate(hw, rate, prate, NULL, width, divider->flags); + if ((clk_hw_get_flags(hw) & CLK_SET_RATE_PARENT) && divider->is_frac) + bestdiv = rate % *prate ? 1 : bestdiv; - if (divider->is_frac && (clk_hw_get_flags(hw) & CLK_SET_RATE_PARENT) && (rate % *prate)) - *prate = rate; + bestdiv = min_t(u32, bestdiv, divider->max_div); + *prate = rate * bestdiv; return rate; } -- 2.25.1

2 months, 2 weeks

2
1
0 0

[PATCH 6.1.y] mptcp: fix possible integer overflow in mptcp_reset_tout_timer

by Matthieu Baerts (NGI0)

From: Dmitry Kandybka <d.kandybka(a)gmail.com> commit b169e76ebad22cbd055101ee5aa1a7bed0e66606 upstream. In 'mptcp_reset_tout_timer', promote 'probe_timestamp' to unsigned long to avoid possible integer overflow. Compile tested only. Found by Linux Verification Center (linuxtesting.org) with SVACE. Signed-off-by: Dmitry Kandybka <d.kandybka(a)gmail.com> Link: https://patch.msgid.link/20241107103657.1560536-1-d.kandybka@gmail.com Signed-off-by: Jakub Kicinski <kuba(a)kernel.org> [ Conflict in this version because commit d866ae9aaa43 ("mptcp: add a new sysctl for make after break timeout") is not in this version, and replaced TCP_TIMEWAIT_LEN in the expression. The fix can still be applied the same way: by forcing a cast to unsigned long for the first item. ] Signed-off-by: Matthieu Baerts (NGI0) <matttbe(a)kernel.org> --- net/mptcp/protocol.c | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-) diff --git a/net/mptcp/protocol.c b/net/mptcp/protocol.c index 1acd4e37a0ea..370afcac2623 100644 --- a/net/mptcp/protocol.c +++ b/net/mptcp/protocol.c @@ -2708,8 +2708,8 @@ void mptcp_reset_tout_timer(struct mptcp_sock *msk, unsigned long fail_tout) if (!fail_tout && !inet_csk(sk)->icsk_mtup.probe_timestamp) return; - close_timeout = inet_csk(sk)->icsk_mtup.probe_timestamp - tcp_jiffies32 + jiffies + - TCP_TIMEWAIT_LEN; + close_timeout = (unsigned long)inet_csk(sk)->icsk_mtup.probe_timestamp - + tcp_jiffies32 + jiffies + TCP_TIMEWAIT_LEN; /* the close timeout takes precedence on the fail one, and here at least one of * them is active -- 2.45.2

2 months, 2 weeks

2
1
0 0

[PATCH 5.4] nvme: fix metadata handling in nvme-passthrough

by Hagar Hemdan

From: Puranjay Mohan <pjy(a)amazon.com> [ Upstream commit 7c2fd76048e95dd267055b5f5e0a48e6e7c81fd9 ] On an NVMe namespace that does not support metadata, it is possible to send an IO command with metadata through io-passthru. This allows issues like [1] to trigger in the completion code path. nvme_map_user_request() doesn't check if the namespace supports metadata before sending it forward. It also allows admin commands with metadata to be processed as it ignores metadata when bdev == NULL and may report success. Reject an IO command with metadata when the NVMe namespace doesn't support it and reject an admin command if it has metadata. [1] https://lore.kernel.org/all/mb61pcylvnym8.fsf@amazon.com/ Suggested-by: Christoph Hellwig <hch(a)lst.de> Signed-off-by: Puranjay Mohan <pjy(a)amazon.com> Reviewed-by: Christoph Hellwig <hch(a)lst.de> Reviewed-by: Sagi Grimberg <sagi(a)grimberg.me> Reviewed-by: Anuj Gupta <anuj20.g(a)samsung.com> Signed-off-by: Keith Busch <kbusch(a)kernel.org> [ Move the changes from nvme_map_user_request() to nvme_submit_user_cmd() to make it work on 5.4 ] Signed-off-by: Hagar Hemdan <hagarhem(a)amazon.com> --- drivers/nvme/host/core.c | 7 ++++++- 1 file changed, 6 insertions(+), 1 deletion(-) diff --git a/drivers/nvme/host/core.c b/drivers/nvme/host/core.c index 0676637e1eab..a841fd4929ad 100644 --- a/drivers/nvme/host/core.c +++ b/drivers/nvme/host/core.c @@ -921,11 +921,16 @@ static int nvme_submit_user_cmd(struct request_queue *q, bool write = nvme_is_write(cmd); struct nvme_ns *ns = q->queuedata; struct gendisk *disk = ns ? ns->disk : NULL; + bool supports_metadata = disk && blk_get_integrity(disk); + bool has_metadata = meta_buffer && meta_len; struct request *req; struct bio *bio = NULL; void *meta = NULL; int ret; + if (has_metadata && !supports_metadata) + return -EINVAL; + req = nvme_alloc_request(q, cmd, 0, NVME_QID_ANY); if (IS_ERR(req)) return PTR_ERR(req); @@ -940,7 +945,7 @@ static int nvme_submit_user_cmd(struct request_queue *q, goto out; bio = req->bio; bio->bi_disk = disk; - if (disk && meta_buffer && meta_len) { + if (has_metadata) { meta = nvme_add_user_metadata(bio, meta_buffer, meta_len, meta_seed, write); if (IS_ERR(meta)) { -- 2.40.1

2 months, 2 weeks

2
1
0 0

[PATCH 4.19] nvme: fix metadata handling in nvme-passthrough

by Hagar Hemdan

From: Puranjay Mohan <pjy(a)amazon.com> [ Upstream commit 7c2fd76048e95dd267055b5f5e0a48e6e7c81fd9 ] On an NVMe namespace that does not support metadata, it is possible to send an IO command with metadata through io-passthru. This allows issues like [1] to trigger in the completion code path. nvme_map_user_request() doesn't check if the namespace supports metadata before sending it forward. It also allows admin commands with metadata to be processed as it ignores metadata when bdev == NULL and may report success. Reject an IO command with metadata when the NVMe namespace doesn't support it and reject an admin command if it has metadata. [1] https://lore.kernel.org/all/mb61pcylvnym8.fsf@amazon.com/ Suggested-by: Christoph Hellwig <hch(a)lst.de> Signed-off-by: Puranjay Mohan <pjy(a)amazon.com> Reviewed-by: Christoph Hellwig <hch(a)lst.de> Reviewed-by: Sagi Grimberg <sagi(a)grimberg.me> Reviewed-by: Anuj Gupta <anuj20.g(a)samsung.com> Signed-off-by: Keith Busch <kbusch(a)kernel.org> [ Move the changes from nvme_map_user_request() to nvme_submit_user_cmd() to make it work on 4.19 ] Signed-off-by: Hagar Hemdan <hagarhem(a)amazon.com> --- drivers/nvme/host/core.c | 7 ++++++- 1 file changed, 6 insertions(+), 1 deletion(-) diff --git a/drivers/nvme/host/core.c b/drivers/nvme/host/core.c index 6adff541282b..fcf062f3b507 100644 --- a/drivers/nvme/host/core.c +++ b/drivers/nvme/host/core.c @@ -802,11 +802,16 @@ static int nvme_submit_user_cmd(struct request_queue *q, bool write = nvme_is_write(cmd); struct nvme_ns *ns = q->queuedata; struct gendisk *disk = ns ? ns->disk : NULL; + bool supports_metadata = disk && blk_get_integrity(disk); + bool has_metadata = meta_buffer && meta_len; struct request *req; struct bio *bio = NULL; void *meta = NULL; int ret; + if (has_metadata && !supports_metadata) + return -EINVAL; + req = nvme_alloc_request(q, cmd, 0, NVME_QID_ANY); if (IS_ERR(req)) return PTR_ERR(req); @@ -821,7 +826,7 @@ static int nvme_submit_user_cmd(struct request_queue *q, goto out; bio = req->bio; bio->bi_disk = disk; - if (disk && meta_buffer && meta_len) { + if (has_metadata) { meta = nvme_add_user_metadata(bio, meta_buffer, meta_len, meta_seed, write); if (IS_ERR(meta)) { -- 2.40.1

2 months, 2 weeks

2
1
0 0

2025

2024

2023

2022

2021

2020

2019

2018

2017

Linux-stable-mirror