Linux kernel -stable discussions
 help / color / mirror / Atom feed
From: Ronja Meyer <rnj@google.com>
To: Jack Wang <jinpu.wang@cloud.ionos.com>,
	 "James E.J. Bottomley" <James.Bottomley@HansenPartnership.com>,
	 "Martin K. Petersen" <martin.petersen@oracle.com>,
	Tom Peng <tom_peng@usish.com>,  Kevin Ao <aoqingyun@usish.com>,
	Lindar Liu <lindar_liu@usish.com>,
	 James Bottomley <James.Bottomley@suse.de>
Cc: jack wang <jack_wang@usish.com>,
	linux-scsi@vger.kernel.org,  linux-kernel@vger.kernel.org,
	Ronja Meyer <rnj@google.com>,
	stable@vger.kernel.org,  Igor Pylypiv <ipylypiv@google.com>
Subject: [PATCH 2/2] scsi: pm8001: Match hw_event_resp to HBA data layout
Date: Fri, 15 May 2026 12:15:40 +0000	[thread overview]
Message-ID: <20260515-fortify_pm80-v1-2-2863187f6d4b@google.com> (raw)
In-Reply-To: <20260515-fortify_pm80-v1-0-2863187f6d4b@google.com>

Correct the hw_event_resp struct definition to match the layout of data
sent by the HBA. Remove pointer arithmetics previously required to work
around incorrect struct definition.

Looking at the struct definition before this patch:
  struct hw_event_resp {
           [...]
           struct	sas_identify_frame sas_identify;
           struct dev_to_host_fis	sata_fis;
   } __attribute__((packed, aligned(4)));

Previously the memcpy() in hw_event_sata_phy_up() crossed reading
from the sas_identify struct over into the sata_fis struct. This was
necessary, because the hw_event_resp struct definition didn't align
properly with what the HBA actually sent. The member sas_identify right
before the member sata_fis was 4 bytes too long, causing the first
4 bytes of the sata_fis to be shifted into the last 4 bytes of
sas_identify. The code worked around this by subtracting 4 bytes from
both the sata_fis pointer, as well as sizeof(sas_identify), when they
were used.

FORTIFY_SOURCE detected this deliberate choice to cross struct member
boundaries as an out-of-bounds read, even though in this case it didn't
lead to a vulnerability. Hence the following fortify-panic was
triggered:

  kernel BUG at lib/string_helpers.c:1044!
  RIP: 0010:__fortify_panic+0x9/0x10
  hw_event_sata_phy_up+0xea/0x120 [pm80xx]
  process_one_iomb+0x634e/0x6360 [pm80xx]
  process_oq+0x391/0x430 [pm80xx]
  pm80xx_chip_isr+0x78/0x100 [pm80xx]
  tasklet_action_common+0x16a/0x2b0
  handle_softirqs+0xcd/0x2a0
  __irq_exit_rcu+0x50/0x100
  common_interrupt+0x89/0xa0

Furthermore hw_event_resp was 64 bytes before this patch, which is
4 bytes too long. Messages exchanged between the pm8001 and the host
kernel can be a maximum of 64 bytes, as defined in iomb_size. The
message structs defined in pm8001_hwi.h must have a size of 60 bytes,
in order to leave space for a 4 byte header that implicitly precedes
each message.

Luckily the code interacting with hw_event_resp doesn't ever seem to
read or write the last 4 bytes of the struct and doesn't seem to use
the incorrect size of the struct in a copy operation. Hence it doesn't
overflow in practice. Further the pm80xx driver was unaffected by this
bug. While the pm80xx struct was also 64 bytes, the message size on
pm80xx is 128 bytes. Hence it is able to fit the 68 byte header and
message without overflowing.

This is not security critical AFAICT.

Cc: stable@vger.kernel.org
Fixes: dbf9bfe61571 ("[SCSI] pm8001: add SAS/SATA HBA driver")
Co-developed-by: Igor Pylypiv <ipylypiv@google.com>
Signed-off-by: Igor Pylypiv <ipylypiv@google.com>
Signed-off-by: Ronja Meyer <rnj@google.com>
---
 drivers/scsi/pm8001/pm8001_hwi.c | 6 +++---
 drivers/scsi/pm8001/pm8001_hwi.h | 2 +-
 drivers/scsi/pm8001/pm80xx_hwi.c | 6 +++---
 drivers/scsi/pm8001/pm80xx_hwi.h | 4 ++--
 4 files changed, 9 insertions(+), 9 deletions(-)

diff --git a/drivers/scsi/pm8001/pm8001_hwi.c b/drivers/scsi/pm8001/pm8001_hwi.c
index fff8d877abb9..e90f2d98d8ed 100644
--- a/drivers/scsi/pm8001/pm8001_hwi.c
+++ b/drivers/scsi/pm8001/pm8001_hwi.c
@@ -3164,8 +3164,8 @@ hw_event_sas_phy_up(struct pm8001_hba_info *pm8001_ha, void *piomb)
 	sas_notify_phy_event(&phy->sas_phy, PHYE_OOB_DONE, GFP_ATOMIC);
 	spin_lock_irqsave(&phy->sas_phy.frame_rcvd_lock, flags);
 	memcpy(phy->frame_rcvd, &pPayload->sas_identify,
-		sizeof(struct sas_identify_frame)-4);
-	phy->frame_rcvd_size = sizeof(struct sas_identify_frame) - 4;
+		sizeof(struct sas_identify_frame_local));
+	phy->frame_rcvd_size = sizeof(struct sas_identify_frame_local);
 	pm8001_get_attached_sas_addr(phy, phy->sas_phy.attached_sas_addr);
 	spin_unlock_irqrestore(&phy->sas_phy.frame_rcvd_lock, flags);
 	if (pm8001_ha->flags == PM8001F_RUN_TIME)
@@ -3208,7 +3208,7 @@ hw_event_sata_phy_up(struct pm8001_hba_info *pm8001_ha, void *piomb)
 	phy->sas_phy.oob_mode = SATA_OOB_MODE;
 	sas_notify_phy_event(&phy->sas_phy, PHYE_OOB_DONE, GFP_ATOMIC);
 	spin_lock_irqsave(&phy->sas_phy.frame_rcvd_lock, flags);
-	memcpy(phy->frame_rcvd, ((u8 *)&pPayload->sata_fis - 4),
+	memcpy(phy->frame_rcvd, &pPayload->sata_fis,
 		sizeof(struct dev_to_host_fis));
 	phy->frame_rcvd_size = sizeof(struct dev_to_host_fis);
 	phy->identify.target_port_protocols = SAS_PROTOCOL_SATA;
diff --git a/drivers/scsi/pm8001/pm8001_hwi.h b/drivers/scsi/pm8001/pm8001_hwi.h
index 14b162f93eb8..2b5483989886 100644
--- a/drivers/scsi/pm8001/pm8001_hwi.h
+++ b/drivers/scsi/pm8001/pm8001_hwi.h
@@ -326,7 +326,7 @@ struct hw_event_resp {
 	__le32	lr_evt_status_phyid_portid;
 	__le32	evt_param;
 	__le32	npip_portstate;
-	struct sas_identify_frame	sas_identify;
+	struct	sas_identify_frame_local sas_identify;	/* _local to omit CRC field */
 	struct dev_to_host_fis	sata_fis;
 } __attribute__((packed, aligned(4)));
 
diff --git a/drivers/scsi/pm8001/pm80xx_hwi.c b/drivers/scsi/pm8001/pm80xx_hwi.c
index 954f307352e6..03293e9b84e6 100644
--- a/drivers/scsi/pm8001/pm80xx_hwi.c
+++ b/drivers/scsi/pm8001/pm80xx_hwi.c
@@ -3241,8 +3241,8 @@ hw_event_sas_phy_up(struct pm8001_hba_info *pm8001_ha, void *piomb)
 	sas_notify_phy_event(&phy->sas_phy, PHYE_OOB_DONE, GFP_ATOMIC);
 	spin_lock_irqsave(&phy->sas_phy.frame_rcvd_lock, flags);
 	memcpy(phy->frame_rcvd, &pPayload->sas_identify,
-		sizeof(struct sas_identify_frame)-4);
-	phy->frame_rcvd_size = sizeof(struct sas_identify_frame) - 4;
+		sizeof(struct sas_identify_frame_local));
+	phy->frame_rcvd_size = sizeof(struct sas_identify_frame_local);
 	pm8001_get_attached_sas_addr(phy, phy->sas_phy.attached_sas_addr);
 	spin_unlock_irqrestore(&phy->sas_phy.frame_rcvd_lock, flags);
 	if (pm8001_ha->flags == PM8001F_RUN_TIME)
@@ -3289,7 +3289,7 @@ hw_event_sata_phy_up(struct pm8001_hba_info *pm8001_ha, void *piomb)
 	phy->sas_phy.oob_mode = SATA_OOB_MODE;
 	sas_notify_phy_event(&phy->sas_phy, PHYE_OOB_DONE, GFP_ATOMIC);
 	spin_lock_irqsave(&phy->sas_phy.frame_rcvd_lock, flags);
-	memcpy(phy->frame_rcvd, ((u8 *)&pPayload->sata_fis - 4),
+	memcpy(phy->frame_rcvd, &pPayload->sata_fis,
 		sizeof(struct dev_to_host_fis));
 	phy->frame_rcvd_size = sizeof(struct dev_to_host_fis);
 	phy->identify.target_port_protocols = SAS_PROTOCOL_SATA;
diff --git a/drivers/scsi/pm8001/pm80xx_hwi.h b/drivers/scsi/pm8001/pm80xx_hwi.h
index d8a63b7fed6a..5ccc010a2368 100644
--- a/drivers/scsi/pm8001/pm80xx_hwi.h
+++ b/drivers/scsi/pm8001/pm80xx_hwi.h
@@ -351,7 +351,7 @@ struct mpi_msg_hdr {
 struct phy_start_req {
 	__le32	tag;
 	__le32	ase_sh_lm_slr_phyid;
-	struct sas_identify_frame_local sas_identify; /* 28 Bytes */
+	struct	sas_identify_frame_local sas_identify;	/* _local to omit CRC field */
 	__le32 spasti;
 	u32	reserved[21];
 } __attribute__((packed, aligned(4)));
@@ -427,7 +427,7 @@ struct hw_event_resp {
 	__le32	lr_status_evt_portid;
 	__le32	evt_param;
 	__le32	phyid_npip_portstate;
-	struct sas_identify_frame	sas_identify;
+	struct	sas_identify_frame_local	sas_identify;	/* _local to omit CRC field */
 	struct dev_to_host_fis	sata_fis;
 } __attribute__((packed, aligned(4)));
 

-- 
2.54.0.563.g4f69b47b94-goog


      parent reply	other threads:[~2026-05-15 12:17 UTC|newest]

Thread overview: 4+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2026-05-15 12:15 [PATCH 0/2] scsi: pm8001: Fix struct layout and FORTIFY_SOURCE crash Ronja Meyer
2026-05-15 12:15 ` [PATCH 1/2] scsi: pm8001: Redefine sas_identify_frame structure Ronja Meyer
2026-05-15 14:48   ` James Bottomley
2026-05-15 12:15 ` Ronja Meyer [this message]

Reply instructions:

You may reply publicly to this message via plain-text email
using any one of the following methods:

* Save the following mbox file, import it into your mail client,
  and reply-to-all from there: mbox

  Avoid top-posting and favor interleaved quoting:
  https://en.wikipedia.org/wiki/Posting_style#Interleaved_style

* Reply using the --to, --cc, and --in-reply-to
  switches of git-send-email(1):

  git send-email \
    --in-reply-to=20260515-fortify_pm80-v1-2-2863187f6d4b@google.com \
    --to=rnj@google.com \
    --cc=James.Bottomley@HansenPartnership.com \
    --cc=James.Bottomley@suse.de \
    --cc=aoqingyun@usish.com \
    --cc=ipylypiv@google.com \
    --cc=jack_wang@usish.com \
    --cc=jinpu.wang@cloud.ionos.com \
    --cc=lindar_liu@usish.com \
    --cc=linux-kernel@vger.kernel.org \
    --cc=linux-scsi@vger.kernel.org \
    --cc=martin.petersen@oracle.com \
    --cc=stable@vger.kernel.org \
    --cc=tom_peng@usish.com \
    /path/to/YOUR_REPLY

  https://kernel.org/pub/software/scm/git/docs/git-send-email.html

* If your mail client supports setting the In-Reply-To header
  via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox