From e5c463e5074a940c644d54d8fc29aba06709aab0 Mon Sep 17 00:00:00 2001 From: Mark Korenberg Date: Thu, 5 Nov 2026 00:25:05 +0500 Subject: [PATCH 1/6] ublk: publish a new device by its id only once it is set up ADD_DEV stores the device in ublk_index_idr when it allocates the id, and control commands find it there as soon as ublk_add_chdev() has initialized its struct device, before the char device is added. If dev_set_name() or cdev_device_add() then fails, ublk_add_chdev() drops the initial reference to free the device. A DEL_DEV that found the device meanwhile waits for ublk_ctl_mutex, then calls ublk_remove(): cdev_device_del() of a device that was never added, and a second put of the initial reference, which frees the device under the callers still holding references. Reserve the id with a NULL entry, which idr_find() and the idr iterators skip, and store the device once ublk_add_chdev() succeeded. Fixes: 71f28f3136af ("ublk_drv: add io_uring based userspace block driver") Signed-off-by: Mark Korenberg Reviewed-by: Ming Lei Link: https://patch.msgid.link/20261105-ublk-del-gendisk-hang-v1-1-913a91a4314d@gmail.com Signed-off-by: Jens Axboe --- drivers/block/ublk_drv.c | 23 +++++++++++++++++++++-- 1 file changed, 21 insertions(+), 2 deletions(-) diff --git a/drivers/block/ublk_drv.c b/drivers/block/ublk_drv.c index 3d322fa1bc83..1f7b11ee1bd9 100644 --- a/drivers/block/ublk_drv.c +++ b/drivers/block/ublk_drv.c @@ -4405,14 +4405,18 @@ static int ublk_alloc_dev_number(struct ublk_device *ub, int idx) int i = idx; int err; + /* + * Only reserve the id: the device is published by + * ublk_publish_dev_number() once it is fully set up. + */ spin_lock(&ublk_idr_lock); /* allocate id, if @id >= 0, we're requesting that specific id */ if (i >= 0) { - err = idr_alloc(&ublk_index_idr, ub, i, i + 1, GFP_NOWAIT); + err = idr_alloc(&ublk_index_idr, NULL, i, i + 1, GFP_NOWAIT); if (err == -ENOSPC) err = -EEXIST; } else { - err = idr_alloc(&ublk_index_idr, ub, 0, UBLK_MAX_UBLKS, + err = idr_alloc(&ublk_index_idr, NULL, 0, UBLK_MAX_UBLKS, GFP_NOWAIT); } spin_unlock(&ublk_idr_lock); @@ -4423,6 +4427,19 @@ static int ublk_alloc_dev_number(struct ublk_device *ub, int idx) return err; } +/* + * Make the device visible to control commands by its id. Before its char + * device is added, DEL_DEV could find it and remove a char device that was + * never added, and drop the reference that ublk_add_chdev() drops on + * failure. + */ +static void ublk_publish_dev_number(struct ublk_device *ub) +{ + spin_lock(&ublk_idr_lock); + WARN_ON_ONCE(idr_replace(&ublk_index_idr, ub, ub->ub_number) != NULL); + spin_unlock(&ublk_idr_lock); +} + static void ublk_free_dev_number(struct ublk_device *ub) { spin_lock(&ublk_idr_lock); @@ -4991,6 +5008,8 @@ static int ublk_ctrl_add_dev(const struct ublksrv_ctrl_cmd *header) * ublk_add_chdev() will cleanup everything if it fails. */ ret = ublk_add_chdev(ub); + if (!ret) + ublk_publish_dev_number(ub); goto out_unlock; out_deinit_queues: From 8067050e7226844fb32062b5cea947e5897d3d7e Mon Sep 17 00:00:00 2001 From: Mark Korenberg Date: Thu, 5 Nov 2026 00:25:06 +0500 Subject: [PATCH 2/6] ublk: don't take ublk_ctl_mutex in DEL_DEV ublk_ctrl_del_dev() holds the global ublk_ctl_mutex across ublk_remove(). Stopping the device calls del_gendisk(), which syncs the filesystem on the disk. If that filesystem is stuck, DEL_DEV never returns, and ADD_DEV and DEL_DEV of every other ublk device wait for ublk_ctl_mutex behind it, until reboot. This happens with the server alive, and also when it dies: the release work then hangs in the same del_gendisk() with ub->mutex held, and DEL_DEV sleeps on ub->mutex uninterruptibly, holding ublk_ctl_mutex. DEL_DEV doesn't need the mutex. ADD_DEV is done with the device when DEL_DEV finds it: the device is found by its id only once it is set up, since the previous patch. Claiming UB_STATE_DELETED with test_and_set_bit() lets only one DEL_DEV remove the device, and the device number is freed under ublk_idr_lock on release. The mutex also protected unprivileged_ublks_added: make it atomic. ADD_DEV still checks and increments it under the mutex, so a decrement without it can only lower the count, and the limit holds. A second DEL_DEV of the same device used to wait for the first one on ublk_ctl_mutex. Now it sees UB_STATE_DELETED and, unless it is DEL_DEV_ASYNC, waits interruptibly for the device number to be freed. The counter is decremented once the device is removed, so at the limit ADD_DEV of an unprivileged device fails with -EACCES instead of waiting for the removal, and the limit is never exceeded, even if the removal never completes. Reproducer: XFS on a ublk loop device with an external log on dm-linear; "dmsetup suspend --nolockfs --noflush" the log device, write a file and syncfs, SIGKILL the ublk server or not, DEL_DEV the device, ADD_DEV another one. Before, ADD_DEV hangs on ublk_ctl_mutex until the log device is resumed. After, it completes; only the DEL_DEV of the device with the stuck filesystem waits. Signed-off-by: Mark Korenberg Reviewed-by: Ming Lei Link: https://patch.msgid.link/20261105-ublk-del-gendisk-hang-v1-2-913a91a4314d@gmail.com Signed-off-by: Jens Axboe --- drivers/block/ublk_drv.c | 43 ++++++++++++++++++---------------------- 1 file changed, 19 insertions(+), 24 deletions(-) diff --git a/drivers/block/ublk_drv.c b/drivers/block/ublk_drv.c index 1f7b11ee1bd9..d415be37d5ed 100644 --- a/drivers/block/ublk_drv.c +++ b/drivers/block/ublk_drv.c @@ -881,7 +881,8 @@ static ssize_t ublk_batch_copy_io_tags(struct ublk_batch_fetch_cmd *fcmd, * by cgroup. */ static unsigned int unprivileged_ublks_max = 64; -static unsigned int unprivileged_ublks_added; /* protected by ublk_ctl_mutex */ +/* checked and incremented under ublk_ctl_mutex, decremented without it */ +static atomic_t unprivileged_ublks_added = ATOMIC_INIT(0); static struct miscdevice ublk_misc; @@ -4486,7 +4487,7 @@ static int ublk_add_chdev(struct ublk_device *ub) goto fail; if (ub->dev_info.flags & UBLK_F_UNPRIVILEGED_DEV) - unprivileged_ublks_added++; + atomic_inc(&unprivileged_ublks_added); return 0; fail: put_device(dev); @@ -4518,15 +4519,9 @@ static int ublk_add_tag_set(struct ublk_device *ub) static void ublk_remove(struct ublk_device *ub) { - bool unprivileged; - ublk_stop_dev(ub); cdev_device_del(&ub->cdev, &ub->cdev_dev); - unprivileged = ub->dev_info.flags & UBLK_F_UNPRIVILEGED_DEV; ublk_put_device(ub); - - if (unprivileged) - unprivileged_ublks_added--; } static struct ublk_device *ublk_get_device_from_id(int idx) @@ -4916,7 +4911,7 @@ static int ublk_ctrl_add_dev(const struct ublksrv_ctrl_cmd *header) ret = -EACCES; if ((info.flags & UBLK_F_UNPRIVILEGED_DEV) && - unprivileged_ublks_added >= unprivileged_ublks_max) + atomic_read(&unprivileged_ublks_added) >= unprivileged_ublks_max) goto out_unlock; ret = -ENOMEM; @@ -5042,23 +5037,23 @@ static int ublk_ctrl_del_dev(struct ublk_device **p_ub, bool wait) { struct ublk_device *ub = *p_ub; int idx = ub->ub_number; - int ret; - - ret = mutex_lock_killable(&ublk_ctl_mutex); - if (ret) - return ret; - - if (!test_bit(UB_STATE_DELETED, &ub->state)) { - ublk_remove(ub); - set_bit(UB_STATE_DELETED, &ub->state); - } - - mutex_unlock(&ublk_ctl_mutex); + bool unprivileged = ub->dev_info.flags & UBLK_F_UNPRIVILEGED_DEV; /* - * Drop the reference outside ublk_ctl_mutex: if it is the last one, - * the release frees the tag set, which waits for an SRCU grace period, - * and holding the mutex would serialize concurrent deletions on it. + * Don't take ublk_ctl_mutex: stopping the device may wait for the + * filesystem on its disk indefinitely, and ADD_DEV and DEL_DEV of all + * other ublk devices would wait for that too. ADD_DEV is done with the + * device, it is found by its id only then. + */ + if (!test_and_set_bit(UB_STATE_DELETED, &ub->state)) { + ublk_remove(ub); + if (unprivileged) + atomic_dec(&unprivileged_ublks_added); + } + + /* + * If this is the last reference, the release frees the tag set, which + * waits for an SRCU grace period. */ *p_ub = NULL; ublk_put_device(ub); From ae6f19236b5ae4eef4df2780d51bee19add0eba8 Mon Sep 17 00:00:00 2001 From: Stefan Haberland Date: Thu, 12 Nov 2026 11:26:05 +0100 Subject: [PATCH 3/6] s390/dasd: Fix typo 'immediatly' -> 'immediately' MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit Fix typo in function name dasd_sleep_on_immediatly(). Suggested-by: Heiko Carstens Reviewed-by: Jan Höppner Signed-off-by: Stefan Haberland Link: https://patch.msgid.link/20261112102608.1057852-2-sth@linux.ibm.com Signed-off-by: Jens Axboe --- drivers/s390/block/dasd.c | 4 ++-- drivers/s390/block/dasd_alias.c | 2 +- drivers/s390/block/dasd_eckd.c | 14 +++++++------- drivers/s390/block/dasd_int.h | 2 +- 4 files changed, 11 insertions(+), 11 deletions(-) diff --git a/drivers/s390/block/dasd.c b/drivers/s390/block/dasd.c index 2ffe404b890f..d74ef737ae26 100644 --- a/drivers/s390/block/dasd.c +++ b/drivers/s390/block/dasd.c @@ -2576,7 +2576,7 @@ static inline int _dasd_term_running_cqr(struct dasd_device *device) return rc; } -int dasd_sleep_on_immediatly(struct dasd_ccw_req *cqr) +int dasd_sleep_on_immediately(struct dasd_ccw_req *cqr) { struct dasd_device *device; int rc; @@ -2628,7 +2628,7 @@ int dasd_sleep_on_immediatly(struct dasd_ccw_req *cqr) return rc; } -EXPORT_SYMBOL(dasd_sleep_on_immediatly); +EXPORT_SYMBOL(dasd_sleep_on_immediately); /* * Cancels a request that was started with dasd_sleep_on_req. diff --git a/drivers/s390/block/dasd_alias.c b/drivers/s390/block/dasd_alias.c index 8d6c16d35c22..1ffcfcfe4565 100644 --- a/drivers/s390/block/dasd_alias.c +++ b/drivers/s390/block/dasd_alias.c @@ -750,7 +750,7 @@ static int reset_summary_unit_check(struct alias_lcu *lcu, cqr->buildclk = get_tod_clock(); cqr->status = DASD_CQR_FILLED; - rc = dasd_sleep_on_immediatly(cqr); + rc = dasd_sleep_on_immediately(cqr); return rc; } diff --git a/drivers/s390/block/dasd_eckd.c b/drivers/s390/block/dasd_eckd.c index 9d1b11020b4e..53f823a9310e 100644 --- a/drivers/s390/block/dasd_eckd.c +++ b/drivers/s390/block/dasd_eckd.c @@ -922,7 +922,7 @@ static int dasd_eckd_read_conf_immediately(struct dasd_device *device, set_bit(DASD_CQR_ALLOW_SLOCK, &cqr->flags); cqr->retries = 5; cqr->callback = read_conf_cb; - rc = dasd_sleep_on_immediatly(cqr); + rc = dasd_sleep_on_immediately(cqr); return rc; } @@ -6008,7 +6008,7 @@ __context_unsafe(/* Conditional locking */) cqr->buildclk = get_tod_clock(); cqr->status = DASD_CQR_FILLED; - rc = dasd_sleep_on_immediatly(cqr); + rc = dasd_sleep_on_immediately(cqr); if (!rc) clear_bit(DASD_FLAG_IS_RESERVED, &device->flags); @@ -6064,7 +6064,7 @@ __context_unsafe(/* Conditional locking */) cqr->buildclk = get_tod_clock(); cqr->status = DASD_CQR_FILLED; - rc = dasd_sleep_on_immediatly(cqr); + rc = dasd_sleep_on_immediately(cqr); if (!rc) set_bit(DASD_FLAG_IS_RESERVED, &device->flags); @@ -6119,7 +6119,7 @@ __context_unsafe(/* Conditional locking */) cqr->buildclk = get_tod_clock(); cqr->status = DASD_CQR_FILLED; - rc = dasd_sleep_on_immediatly(cqr); + rc = dasd_sleep_on_immediately(cqr); if (!rc) set_bit(DASD_FLAG_IS_RESERVED, &device->flags); @@ -6183,7 +6183,7 @@ __context_unsafe(/* Conditional locking */) cqr->status = DASD_CQR_FILLED; cqr->lpm = usrparm.path_mask; - rc = dasd_sleep_on_immediatly(cqr); + rc = dasd_sleep_on_immediately(cqr); /* verify that I/O processing didn't modify the path mask */ if (!rc && usrparm.path_mask && (cqr->lpm != usrparm.path_mask)) rc = -EIO; @@ -6815,7 +6815,7 @@ retry: cqr->block = NULL; cqr->expires = 10 * HZ; set_bit(DASD_CQR_VERIFY_PATH, &cqr->flags); - /* dasd_sleep_on_immediatly does not do complex error + /* dasd_sleep_on_immediately does not do complex error * recovery so clear erp flag and set retry counter to * do basic erp */ clear_bit(DASD_CQR_FLAGS_USE_ERP, &cqr->flags); @@ -6847,7 +6847,7 @@ retry: cqr->buildclk = get_tod_clock(); cqr->status = DASD_CQR_FILLED; - rc = dasd_sleep_on_immediatly(cqr); + rc = dasd_sleep_on_immediately(cqr); if (rc == 0) { prssdp = (struct dasd_psf_prssd_data *) cqr->data; message_buf = (struct dasd_rssd_messages *) diff --git a/drivers/s390/block/dasd_int.h b/drivers/s390/block/dasd_int.h index ef4930432c09..26ecf9f47d2e 100644 --- a/drivers/s390/block/dasd_int.h +++ b/drivers/s390/block/dasd_int.h @@ -998,7 +998,7 @@ void dasd_schedule_device_bh(struct dasd_device *); void dasd_schedule_block_bh(struct dasd_block *); int dasd_sleep_on(struct dasd_ccw_req *); int dasd_sleep_on_queue(struct list_head *); -int dasd_sleep_on_immediatly(struct dasd_ccw_req *); +int dasd_sleep_on_immediately(struct dasd_ccw_req *); int dasd_sleep_on_queue_interruptible(struct list_head *); int dasd_sleep_on_interruptible(struct dasd_ccw_req *); void dasd_device_set_timer(struct dasd_device *, int); From 25cdc1f63227a1ea69a1ec823a8a860e33299169 Mon Sep 17 00:00:00 2001 From: Stefan Haberland Date: Thu, 12 Nov 2026 11:26:06 +0100 Subject: [PATCH 4/6] s390/dasd: Re-initialize RCD buffer with "V1.0" before every start MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit Under z/VM the Read Configuration Data (RCD) buffer has to start with EBCDIC "V1.0" to request the virtual device SNEQ that carries the minidisk UIT. The request and the response share the same buffer: CP evaluates the magic on the way out and the RCD data is written into the same buffer on the way back. The magic is only written when the request is built and again in read_conf_cb() when the original request fails. Several restart paths do not run that callback: - ERP use the same buffer and if the same error occurs again, the ERP restarts the existing ERP request; if a different error occurs, it adds another ERP request. Neither re-initializes the magic. - Terminated (timed out) ERP requests, killed requests and requests requeued after the last path was gone are restarted without it. - dasd_sleep_on_immediately() overwrites cqr->callback, so read_conf_cb() never ran for dasd_eckd_read_conf_immediately(), and its fastpath retries in dasd_int_handler() do not call a callback either. If such a restart succeeds, z/VM omits the virtual device SNEQ, the UID of that path is built without the UIT and the paths are reported as "Not all channel paths lead to the same device" although they lead to the same minidisk. Every start of a request, whatever path led to it, goes through dasd_start_IO(). Add an optional prepare_io function to the request that dasd_start_IO() calls before every start of the request and of its ERP requests, and use it to write the magic for RCD requests. The ERP handling stays unchanged. read_conf_cb() is no longer needed and is removed, which also gets rid of its memset() that only cleared a single byte. Fixes: 5915a873fcb1 ("[S390] dasd: re-initialize read_conf buffer for retries") Cc: stable@vger.kernel.org Reviewed-by: Jan Höppner Signed-off-by: Stefan Haberland Link: https://patch.msgid.link/20261112102608.1057852-3-sth@linux.ibm.com Signed-off-by: Jens Axboe --- drivers/s390/block/dasd.c | 1 + drivers/s390/block/dasd_eckd.c | 52 ++++++++++++---------------------- drivers/s390/block/dasd_int.h | 15 ++++++++++ 3 files changed, 34 insertions(+), 34 deletions(-) diff --git a/drivers/s390/block/dasd.c b/drivers/s390/block/dasd.c index d74ef737ae26..3a55c937d0d6 100644 --- a/drivers/s390/block/dasd.c +++ b/drivers/s390/block/dasd.c @@ -1396,6 +1396,7 @@ int dasd_start_IO(struct dasd_ccw_req *cqr) cqr->status = DASD_CQR_ERROR; return -EIO; } + dasd_start_IO_prepare(cqr); cqr->startclk = get_tod_clock(); cqr->starttime = jiffies; cqr->retries--; diff --git a/drivers/s390/block/dasd_eckd.c b/drivers/s390/block/dasd_eckd.c index 53f823a9310e..efaf117e39f4 100644 --- a/drivers/s390/block/dasd_eckd.c +++ b/drivers/s390/block/dasd_eckd.c @@ -845,20 +845,29 @@ static int dasd_eckd_compare_path_uid(struct dasd_device *device, return memcmp(&device_uid, &path_uid, sizeof(struct dasd_uid)); } +/* + * The RCD buffer has to start with EBCDIC "V1.0" to show support for + * virtual device SNEQ. The RCD data is written into the same buffer, so + * a failed attempt may already have overwritten the magic. Re-initialize + * it before every start, including restarts by ERP requests. + */ +static void dasd_eckd_rcd_prepare_io(struct dasd_ccw_req *cqr) +{ + struct ccw1 *ccw = cqr->cpaddr; + __u8 *rcd_buffer = dma32_to_virt(ccw->cda); + + rcd_buffer[0] = 0xE5; + rcd_buffer[1] = 0xF1; + rcd_buffer[2] = 0x4B; + rcd_buffer[3] = 0xF0; +} + static void dasd_eckd_fill_rcd_cqr(struct dasd_device *device, struct dasd_ccw_req *cqr, __u8 *rcd_buffer, __u8 lpm) { struct ccw1 *ccw; - /* - * buffer has to start with EBCDIC "V1.0" to show - * support for virtual device SNEQ - */ - rcd_buffer[0] = 0xE5; - rcd_buffer[1] = 0xF1; - rcd_buffer[2] = 0x4B; - rcd_buffer[3] = 0xF0; ccw = cqr->cpaddr; ccw->cmd_code = DASD_ECKD_CCW_RCD; @@ -875,33 +884,10 @@ static void dasd_eckd_fill_rcd_cqr(struct dasd_device *device, cqr->retries = 256; cqr->buildclk = get_tod_clock(); cqr->status = DASD_CQR_FILLED; + cqr->prepare_io = dasd_eckd_rcd_prepare_io; set_bit(DASD_CQR_VERIFY_PATH, &cqr->flags); } -/* - * Wakeup helper for read_conf - * if the cqr is not done and needs some error recovery - * the buffer has to be re-initialized with the EBCDIC "V1.0" - * to show support for virtual device SNEQ - */ -static void read_conf_cb(struct dasd_ccw_req *cqr, void *data) -{ - struct ccw1 *ccw; - __u8 *rcd_buffer; - - if (cqr->status != DASD_CQR_DONE) { - ccw = cqr->cpaddr; - rcd_buffer = dma32_to_virt(ccw->cda); - memset(rcd_buffer, 0, sizeof(*rcd_buffer)); - - rcd_buffer[0] = 0xE5; - rcd_buffer[1] = 0xF1; - rcd_buffer[2] = 0x4B; - rcd_buffer[3] = 0xF0; - } - dasd_wakeup_cb(cqr, data); -} - static int dasd_eckd_read_conf_immediately(struct dasd_device *device, struct dasd_ccw_req *cqr, __u8 *rcd_buffer, @@ -921,7 +907,6 @@ static int dasd_eckd_read_conf_immediately(struct dasd_device *device, clear_bit(DASD_CQR_FLAGS_USE_ERP, &cqr->flags); set_bit(DASD_CQR_ALLOW_SLOCK, &cqr->flags); cqr->retries = 5; - cqr->callback = read_conf_cb; rc = dasd_sleep_on_immediately(cqr); return rc; } @@ -959,7 +944,6 @@ static int dasd_eckd_read_conf_lpm(struct dasd_device *device, goto out_error; } dasd_eckd_fill_rcd_cqr(device, cqr, rcd_buf, lpm); - cqr->callback = read_conf_cb; ret = dasd_sleep_on(cqr); /* * on success we update the user input parms diff --git a/drivers/s390/block/dasd_int.h b/drivers/s390/block/dasd_int.h index 26ecf9f47d2e..53c450036fe1 100644 --- a/drivers/s390/block/dasd_int.h +++ b/drivers/s390/block/dasd_int.h @@ -147,6 +147,8 @@ struct dasd_ccw_req { void *data; /* pointer to data area */ struct irb irb; /* device status in case of an error */ struct dasd_ccw_req *refers; /* ERP-chain queueing. */ + /* optional prepare function called before every start_io, must not sleep */ + void (*prepare_io)(struct dasd_ccw_req *cqr); void *function; /* originating ERP action */ void *mem_chunk; @@ -892,6 +894,19 @@ static inline void *dasd_get_callback_data(struct dasd_ccw_req *cqr) return cqr->callback_data; } +/* + * Walk down the ERP chain to the original CQR and call its + * prepare_io function if set. + */ +static inline void dasd_start_IO_prepare(struct dasd_ccw_req *cqr) +{ + /* ERP requests use the prepare function of the original request */ + while (cqr->refers) + cqr = cqr->refers; + if (cqr->prepare_io) + cqr->prepare_io(cqr); +} + static inline bool dasd_req_conflict(struct dasd_ccw_req *cqr1, struct dasd_ccw_req *cqr2) { From df56f7a6c20a5eb2794c0a7391bde50ba7c8c54b Mon Sep 17 00:00:00 2001 From: Stefan Haberland Date: Thu, 12 Nov 2026 11:26:07 +0100 Subject: [PATCH 5/6] s390/dasd: Fix wrong free and refcount leak in build_check_tcw error path MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit dasd_eckd_build_check_tcw() allocates its request with dasd_fmalloc_request(), but its out_err path frees it with dasd_sfree_request(). The latter uses a different pool and therefor, leading to invalid pointer arithmetic and a crash. Use the matching dasd_ffree_request() instead. The same error path also leaves start_priv->count incremented without ever decrementing it again. This leaks a permanent reference on every build failure. Decrement it before freeing the request. Reported-by: Sashiko Closes: https://sashiko.dev/#/patchset/20260722103704.1709668-1-sth@linux.ibm.com Fixes: 8fd575200db5 ("s390/dasd: Add new ioctl BIODASDCHECKFMT") Cc: stable@vger.kernel.org #4.7 Reviewed-by: Jan Höppner Signed-off-by: Stefan Haberland Fixes: 5e2b17e712cf ("s390/dasd: Add dynamic formatting support for ESE volumes") Link: https://patch.msgid.link/20261112102608.1057852-4-sth@linux.ibm.com Signed-off-by: Jens Axboe --- drivers/s390/block/dasd_eckd.c | 3 ++- 1 file changed, 2 insertions(+), 1 deletion(-) diff --git a/drivers/s390/block/dasd_eckd.c b/drivers/s390/block/dasd_eckd.c index efaf117e39f4..7e60d5b3b343 100644 --- a/drivers/s390/block/dasd_eckd.c +++ b/drivers/s390/block/dasd_eckd.c @@ -2764,7 +2764,8 @@ dasd_eckd_build_check_tcw(struct dasd_device *base, struct format_data_t *fdata, return cqr; out_err: - dasd_sfree_request(cqr, startdev); + start_priv->count--; + dasd_ffree_request(cqr, startdev); return ERR_PTR(rc); } From 7cc2dda4f7aaf596c1796184649007151ff98f8d Mon Sep 17 00:00:00 2001 From: Stefan Haberland Date: Thu, 12 Nov 2026 11:26:08 +0100 Subject: [PATCH 6/6] s390/dasd: Don't trust userspace-supplied result field in format check MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit BIODASDCHECKFMT used to copy the entire format_check_t struct from userspace, including the result/unit/rec/num_records/blksize/ key_length fields that are output fields only. dasd_eckd_format_evaluate_tracks() relies on cdata->result being zero to decide whether to apply an index correction (i--/pos--) after a successful check. A caller that pre-sets result to a non-zero value can suppress that correction, leaving pos advanced past the last valid entry; the (partially out-of-bounds) fmt_buffer[pos] fields are then copied back to userspace, leaking heap contents. Only expect is an Input field; copy just that member in and zero the rest of cdata. Fixes: 8fd575200db5 ("s390/dasd: Add new ioctl BIODASDCHECKFMT") Reported-by: Sashiko Closes: https://sashiko.dev/#/patchset/20260722103704.1709668-1-sth@linux.ibm.com Cc: stable@vger.kernel.org #4.7 Reviewed-by: Jan Höppner Signed-off-by: Stefan Haberland Link: https://patch.msgid.link/20261112102608.1057852-5-sth@linux.ibm.com Signed-off-by: Jens Axboe --- drivers/s390/block/dasd_ioctl.c | 4 +++- 1 file changed, 3 insertions(+), 1 deletion(-) diff --git a/drivers/s390/block/dasd_ioctl.c b/drivers/s390/block/dasd_ioctl.c index e5b8b413f5ab..2f2ec2bea333 100644 --- a/drivers/s390/block/dasd_ioctl.c +++ b/drivers/s390/block/dasd_ioctl.c @@ -303,7 +303,9 @@ static int dasd_ioctl_check_format(struct block_device *bdev, void __user *argp) goto out_err; } - if (copy_from_user(&cdata, argp, sizeof(cdata))) { + /* copy only input fields, clear the rest */ + memset(&cdata, 0, sizeof(cdata)); + if (copy_from_user(&cdata.expect, argp, sizeof(cdata.expect))) { rc = -EFAULT; goto out_err; }