Re: [RFC v2 3/7] powerpc: atomic: Implement atomic{,64}_{add,sub}_return_* variants
On Fri, Sep 25, 2015 at 02:29:04PM -0700, Paul E. McKenney wrote: > On Wed, Sep 23, 2015 at 08:07:55AM +0800, Boqun Feng wrote: > > On Tue, Sep 22, 2015 at 08:25:40AM -0700, Paul E. McKenney wrote: > > > On Tue, Sep 22, 2015 at 07:37:04AM +0800, Boqun Feng wrote: > > > > On Tue, Sep 22, 2015 at 07:26:56AM +0800, Boqun Feng wrote: > > > > > On Mon, Sep 21, 2015 at 11:24:27PM +0100, Will Deacon wrote: > > > > > > Hi Boqun, > > > > > > > > > > > > On Sun, Sep 20, 2015 at 09:23:03AM +0100, Boqun Feng wrote: > > > > > > > On Sat, Sep 19, 2015 at 11:33:10PM +0800, Boqun Feng wrote: > > > > > > > > On Fri, Sep 18, 2015 at 05:59:02PM +0100, Will Deacon wrote: > > > > > > > > > On Wed, Sep 16, 2015 at 04:49:31PM +0100, Boqun Feng wrote: > > > > > > > > > > On powerpc, we don't need a general memory barrier to > > > > > > > > > > achieve acquire and > > > > > > > > > > release semantics, so __atomic_op_{acquire,release} can be > > > > > > > > > > implemented > > > > > > > > > > using "lwsync" and "isync". > > > > > > > > > > > > > > > > > > I'm assuming isync+ctrl isn't transitive, so we need to get > > > > > > > > > to the bottom > > > > > > > > > > > > > > > > Actually the transitivity is still guaranteed here, I think ;-) > > > > > > > > > > > > The litmus test I'm thinking of is: > > > > > > > > > > > > > > > > > > { > > > > > > 0:r2=x; > > > > > > 1:r2=x; 1:r5=z; > > > > > > 2:r2=z; 2:r4=x; > > > > > > } > > > > > > P0 | P1| P2 ; > > > > > > li r1,1 | lwz r1,0(r2) | lwz r1,0(r2) ; > > > > > > stw r1,0(r2) | cmpw r1,r1| cmpw r1,r1 ; > > > > > > | beq LC00 | beq LC01; > > > > > > | LC00: | LC01:; > > > > > > | isync | isync; > > > > > > | li r4,1 | lwz r3,0(r4) ; > > > > > > | stw r4,0(r5) | ; > > > > > > exists > > > > > > (1:r1=1 /\ 2:r1=1 /\ 2:r3=0) > > > > > > > > > > > > > > > > > > Which appears to be allowed. I don't think you need to worry about > > > > > > backwards > > > > > > branches for the ctrl+isync construction (none of the current > > > > > > example do, > > > > > > afaict). > > > > > > > > > > > > > > > > Yes.. my care of backwards branches is not quite related to the > > > > > topic, I > > > > > concerned that mostly because my test is using atomic operation, and I > > > > > just want to test the exact asm code. > > > > > > > > > > > Anyway, all the problematic cases seem to arise when we start mixing > > > > > > ACQUIRE/RELEASE accesses with relaxed accesses (i.e. where an > > > > > > access from > > > > > > one group reads from an access in the other group). It would be > > > > > > simplest > > > > > > to say that this doesn't provide any transitivity guarantees, and > > > > > > that > > > > > > an ACQUIRE must always read from a RELEASE if transitivity is > > > > > > required. > > > > > > > > > > > > > > > > Agreed. RELEASE alone doesn't provide transitivity and transitivity is > > > > ^^^ > > > > This should be ACQUIRE... > > > > > > > > > guaranteed only if an ACQUIRE read from a RELEASE. That's exactly the > > > > > direction which the link (https://lkml.org/lkml/2015/9/15/836) is > > > > > heading to. So I think we are fine here to use ctrl+isync here, right? > > > > > > We are going to have to err on the side of strictness, that is, having > > > the documentation place more requirements on the developer than the union > > > of the hardware does. Besides, I haven't heard any recent complaints > > > that memory-barriers.txt is too simple. ;-) > > > > Agreed ;-) > > > > For atomic operations, using isync in ACQUIRE operations does gaurantee > > that a pure RELEASE/ACQUIRE chain provides transitivity. So, again, I > > think we are fine here to use isync in ACQUIRE atomic operations, > > unless you think we need to be more strict, i.e, making ACQUIRE itself > > provide transitivy? > > As I understand it, either isync or lwsync suffices, with the choice > depending on the hardware. The kernel will rewrite itself at boot time > if you use the appropriate macro. ;-) > Yep ;-) Thank you and Will both for your comments. To be honest, I just began to learn about these transitivity and cumulativity things recently. That's a lot of fun to discuss these with you, Peter and Will, and the herdtools you suggested and the N2745 document you wrote really helped a lot. Thank you again! Regards, Boqun signature.asc Description: PGP signature ___ Linuxppc-dev mailing list Linuxppc-dev@lists.ozlabs.org https://lists.ozlabs.org/listinfo/linuxppc-dev
[PATCH v4 32/32] cxlflash: Fix to avoid potential deadlock on EEH
Ioctl threads that use scsi_execute() can run for an excessive amount of time due to the fact that they have lengthy timeouts and retry logic built in. Under normal operation this is not an issue. However, once EEH enters the picture, a long execution time coupled with the possibility that a timeout can trigger entry to the driver via registered reset callbacks becomes a liability. In particular, a deadlock can occur when an EEH event is encountered while in running in scsi_execute(). As part of the recovery, the EEH handler drains all currently running ioctls, waiting until they have completed before proceeding with a reset. As the scsi_execute()'s are situated on the ioctl path, the EEH handler will wait until they (and the remainder of the ioctl handler they're associated with) have completed. Normally this would not be much of an issue aside from the longer recovery period. Unfortunately, the scsi_execute() triggers a reset when it times out. The reset handler will see that the device is already being reset and wait until that reset completed. This creates a condition where the EEH handler becomes stuck, infinitely waiting for the ioctl thread to complete. To avoid this behavior, temporarily unmark the scsi_execute() threads as an ioctl thread by releasing the ioctl read semaphore. This allows the EEH handler to proceed with a recovery while the thread is still running. Once the scsi_execute() returns, the ioctl read semaphore is reacquired and the adapter state is rechecked in case it changed while inside of scsi_execute(). The state check will wait if the adapter is still being recovered or returns a failure if the recovery failed. In the event that the adapter reset failed, the failure is simply returned as the ioctl would be unable to continue. Reported-by: Brian King Signed-off-by: Matthew R. Ochs Signed-off-by: Manoj N. Kumar --- drivers/scsi/cxlflash/superpipe.c | 30 +- drivers/scsi/cxlflash/superpipe.h | 2 ++ drivers/scsi/cxlflash/vlun.c | 29 + 3 files changed, 60 insertions(+), 1 deletion(-) diff --git a/drivers/scsi/cxlflash/superpipe.c b/drivers/scsi/cxlflash/superpipe.c index f625e07..8af7cdc 100644 --- a/drivers/scsi/cxlflash/superpipe.c +++ b/drivers/scsi/cxlflash/superpipe.c @@ -283,6 +283,24 @@ out: * @sdev: SCSI device associated with LUN. * @lli: LUN destined for capacity request. * + * The READ_CAP16 can take quite a while to complete. Should an EEH occur while + * in scsi_execute(), the EEH handler will attempt to recover. As part of the + * recovery, the handler drains all currently running ioctls, waiting until they + * have completed before proceeding with a reset. As this routine is used on the + * ioctl path, this can create a condition where the EEH handler becomes stuck, + * infinitely waiting for this ioctl thread. To avoid this behavior, temporarily + * unmark this thread as an ioctl thread by releasing the ioctl read semaphore. + * This will allow the EEH handler to proceed with a recovery while this thread + * is still running. Once the scsi_execute() returns, reacquire the ioctl read + * semaphore and check the adapter state in case it changed while inside of + * scsi_execute(). The state check will wait if the adapter is still being + * recovered or return a failure if the recovery failed. In the event that the + * adapter reset failed, simply return the failure as the ioctl would be unable + * to continue. + * + * Note that the above puts a requirement on this routine to only be called on + * an ioctl thread. + * * Return: 0 on success, -errno on failure */ static int read_cap16(struct scsi_device *sdev, struct llun_info *lli) @@ -314,8 +332,18 @@ retry: dev_dbg(dev, "%s: %ssending cmd(0x%x)\n", __func__, retry_cnt ? "re" : "", scsi_cmd[0]); + /* Drop the ioctl read semahpore across lengthy call */ + up_read(&cfg->ioctl_rwsem); result = scsi_execute(sdev, scsi_cmd, DMA_FROM_DEVICE, cmd_buf, CMD_BUFSIZE, sense_buf, to, CMD_RETRIES, 0, NULL); + down_read(&cfg->ioctl_rwsem); + rc = check_state(cfg); + if (rc) { + dev_err(dev, "%s: Failed state! result=0x08%X\n", + __func__, result); + rc = -ENODEV; + goto out; + } if (driver_byte(result) == DRIVER_SENSE) { result &= ~(0xFF<<24); /* DRIVER_SENSE is not an error */ @@ -1221,7 +1249,7 @@ static const struct file_operations null_fops = { * * Return: 0 on success, -errno on failure */ -static int check_state(struct cxlflash_cfg *cfg) +int check_state(struct cxlflash_cfg *cfg) { struct device *dev = &cfg->dev->dev; int rc = 0; diff --git a/drivers/scsi/cxlflash/superpipe.h b/drivers/scsi/cxlflash/superpipe.h index 7df88ee..06a805a 100644 --- a/drivers/scsi/cxlflash/superpipe.h +++ b/drivers/scsi/cxlflash/superpipe.h @@ -
[PATCH v4 31/32] cxlflash: Correct trace string
The trace following the failure of alloc_mem() incorrectly identifies which function failed. This can lead to misdiagnosing a failure. Fix the string to correctly indicate that alloc_mem() failed. Reported-by: Brian King Signed-off-by: Matthew R. Ochs Signed-off-by: Manoj N. Kumar Reviewed-by: Brian King --- drivers/scsi/cxlflash/main.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/drivers/scsi/cxlflash/main.c b/drivers/scsi/cxlflash/main.c index 38e7edc..e8f8c7e 100644 --- a/drivers/scsi/cxlflash/main.c +++ b/drivers/scsi/cxlflash/main.c @@ -2377,7 +2377,7 @@ static int cxlflash_probe(struct pci_dev *pdev, cfg->host = host; rc = alloc_mem(cfg); if (rc) { - dev_err(&pdev->dev, "%s: call to scsi_host_alloc failed!\n", + dev_err(&pdev->dev, "%s: call to alloc_mem failed!\n", __func__); rc = -ENOMEM; scsi_host_put(cfg->host); -- 2.1.0 ___ Linuxppc-dev mailing list Linuxppc-dev@lists.ozlabs.org https://lists.ozlabs.org/listinfo/linuxppc-dev
[PATCH v4 30/32] cxlflash: Fix to avoid corrupting adapter fops
The fops owned by the adapter can be corrupted in certain scenarios, opening a window where certain fops are temporarily NULLed before being reset to their proper value. This can potentially lead software to make incorrect decisions, leaving the user with the inability to function as intended. An example of this behavior can be observed when there are a number of users with a high rate of turn around (attach to LUN, perform an I/O, detach from LUN, repeat). Every so often a user is given a valid context and adapter file descriptor, but the file associated with the descriptor lacks the correct read permission bit (FMODE_CAN_READ) and thus the read system call bails before calling the valid read fop. Background: The fops is stored in the adapter structure to provide the ability to lookup the adapter structure from within the fop handler. CXL services use the file's private_data and at present, the CXL context does not have a private section. In an effort to limit areas of the cxlflash driver with code specific the superpipe function, a design choice was made to keep the details of the fops situated away from the legacy portions of the driver. This drove the behavior that the adapter fops is set at the beginning of the disk attach ioctl handler when there are no users present. The corruption that this fix remedies is due to the fact that the fops is initially defaulted to values found within a static structure. When the fops is handed down to the CXL services later in the attach path, certain services are patched. The fops structure remains correct until the user count drops to 0 and the fops is reset, triggering the process to repeat again. The user counts are tightly coupled with the creation and deletion of the user context. If multiple users perform a disk attach at the same time, when the user count is currently 0, some users can be in the middle of obtaining a file descriptor and have not yet reached the context creation code that [in addition to creating the context] increments the user count. Subsequent users coming in to perform the attach see that the user count is still 0, and reinitialize the fops, temporarily removing the patched fops. The users that are in the middle obtaining their file descriptor may then receive an invalid descriptor. The fix simply removes the user count altogether and moves the fops initialization to probe time such that it is only performed one time for the life of the adapter. In the future, if the CXL services adopt a private member for their context, that could be used to store the adapter structure reference and cxlflash could revert to a model that does not require an embedded fops. Signed-off-by: Matthew R. Ochs Signed-off-by: Manoj N. Kumar --- drivers/scsi/cxlflash/common.h| 3 +-- drivers/scsi/cxlflash/main.c | 1 + drivers/scsi/cxlflash/superpipe.c | 11 +-- 3 files changed, 3 insertions(+), 12 deletions(-) diff --git a/drivers/scsi/cxlflash/common.h b/drivers/scsi/cxlflash/common.h index bbfe711..c11cd19 100644 --- a/drivers/scsi/cxlflash/common.h +++ b/drivers/scsi/cxlflash/common.h @@ -21,6 +21,7 @@ #include #include +extern const struct file_operations cxlflash_cxl_fops; #define MAX_CONTEXT CXLFLASH_MAX_CONTEXT /* num contexts per afu */ @@ -115,8 +116,6 @@ struct cxlflash_cfg { struct list_head ctx_err_recovery; /* contexts w/ recovery pending */ struct file_operations cxl_fops; - atomic_t num_user_contexts; - /* Parameters that are LUN table related */ int last_lun_index[CXLFLASH_NUM_FC_PORTS]; int promote_lun_index; diff --git a/drivers/scsi/cxlflash/main.c b/drivers/scsi/cxlflash/main.c index be78906..38e7edc 100644 --- a/drivers/scsi/cxlflash/main.c +++ b/drivers/scsi/cxlflash/main.c @@ -2386,6 +2386,7 @@ static int cxlflash_probe(struct pci_dev *pdev, cfg->init_state = INIT_STATE_NONE; cfg->dev = pdev; + cfg->cxl_fops = cxlflash_cxl_fops; /* * The promoted LUNs move to the top of the LUN table. The rest stay diff --git a/drivers/scsi/cxlflash/superpipe.c b/drivers/scsi/cxlflash/superpipe.c index 3cc8609..f625e07 100644 --- a/drivers/scsi/cxlflash/superpipe.c +++ b/drivers/scsi/cxlflash/superpipe.c @@ -712,7 +712,6 @@ static void destroy_context(struct cxlflash_cfg *cfg, kfree(ctxi->rht_needs_ws); kfree(ctxi->rht_lun); kfree(ctxi); - atomic_dec_if_positive(&cfg->num_user_contexts); } /** @@ -769,7 +768,6 @@ static struct ctx_info *create_context(struct cxlflash_cfg *cfg, INIT_LIST_HEAD(&ctxi->luns); INIT_LIST_HEAD(&ctxi->list); /* initialize for list_empty() */ - atomic_inc(&cfg->num_user_contexts); mutex_lock(&ctxi->mutex); out: return ctxi; @@ -1164,10 +1162,7 @@ out: return rc; } -/* - * Local fops for adapter file descriptor - */ -static const struct file_operations cxlflash_cxl_fops = { +const struct file_operations cxlflash_cx
[PATCH v4 29/32] cxlflash: Fix to double the delay each time
From: Manoj Kumar The operator used to double the delay is incorrect and does not result in delay doubling. To fix, use a left shift instead of the XOR operator. Reported-by: Tomas Henzl Signed-off-by: Matthew R. Ochs Signed-off-by: Manoj N. Kumar Reviewed-by: Brian King --- drivers/scsi/cxlflash/main.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/drivers/scsi/cxlflash/main.c b/drivers/scsi/cxlflash/main.c index ab11ee6..be78906 100644 --- a/drivers/scsi/cxlflash/main.c +++ b/drivers/scsi/cxlflash/main.c @@ -303,7 +303,7 @@ write_rrin: if (rrin != 0x1) break; /* Double delay each time */ - udelay(2 ^ nretry); + udelay(2 << nretry); } while (nretry++ < MC_ROOM_RETRY_CNT); } -- 2.1.0 ___ Linuxppc-dev mailing list Linuxppc-dev@lists.ozlabs.org https://lists.ozlabs.org/listinfo/linuxppc-dev
[PATCH v4 23/32] cxlflash: Fix function prolog parameters and return codes
Several function prologs have incorrect parameter names and return code descriptions. This can lead to confusion when reviewing the source and creates inaccurate documentation. To remedy, update the function prologs to properly reflect parameter names and return codes. Signed-off-by: Matthew R. Ochs Signed-off-by: Manoj N. Kumar Reviewed-by: Brian King --- drivers/scsi/cxlflash/main.c | 68 1 file changed, 25 insertions(+), 43 deletions(-) diff --git a/drivers/scsi/cxlflash/main.c b/drivers/scsi/cxlflash/main.c index 729f742..7c76227 100644 --- a/drivers/scsi/cxlflash/main.c +++ b/drivers/scsi/cxlflash/main.c @@ -401,8 +401,7 @@ static void wait_resp(struct afu *afu, struct afu_cmd *cmd) * @tmfcmd:TMF command to send. * * Return: - * 0 on success - * SCSI_MLQUEUE_HOST_BUSY when host is busy + * 0 on success or SCSI_MLQUEUE_HOST_BUSY */ static int send_tmf(struct afu *afu, struct scsi_cmnd *scp, u64 tmfcmd) { @@ -491,9 +490,7 @@ static const char *cxlflash_driver_info(struct Scsi_Host *host) * @host: SCSI host associated with device. * @scp: SCSI command to send. * - * Return: - * 0 on success - * SCSI_MLQUEUE_HOST_BUSY when host is busy + * Return: 0 on success or SCSI_MLQUEUE_HOST_BUSY */ static int cxlflash_queuecommand(struct Scsi_Host *host, struct scsi_cmnd *scp) { @@ -597,7 +594,7 @@ out: /** * cxlflash_wait_for_pci_err_recovery() - wait for error recovery during probe - * @cxlflash: Internal structure associated with the host. + * @cfg: Internal structure associated with the host. */ static void cxlflash_wait_for_pci_err_recovery(struct cxlflash_cfg *cfg) { @@ -611,7 +608,7 @@ static void cxlflash_wait_for_pci_err_recovery(struct cxlflash_cfg *cfg) /** * free_mem() - free memory associated with the AFU - * @cxlflash: Internal structure associated with the host. + * @cfg: Internal structure associated with the host. */ static void free_mem(struct cxlflash_cfg *cfg) { @@ -633,7 +630,7 @@ static void free_mem(struct cxlflash_cfg *cfg) /** * stop_afu() - stops the AFU command timers and unmaps the MMIO space - * @cxlflash: Internal structure associated with the host. + * @cfg: Internal structure associated with the host. * * Safe to call with AFU in a partially allocated/initialized state. */ @@ -655,7 +652,7 @@ static void stop_afu(struct cxlflash_cfg *cfg) /** * term_mc() - terminates the master context - * @cxlflash: Internal structure associated with the host. + * @cfg: Internal structure associated with the host. * @level: Depth of allocation, where to begin waterfall tear down. * * Safe to call with AFU/MC in partially allocated/initialized state. @@ -691,7 +688,7 @@ static void term_mc(struct cxlflash_cfg *cfg, enum undo_level level) /** * term_afu() - terminates the AFU - * @cxlflash: Internal structure associated with the host. + * @cfg: Internal structure associated with the host. * * Safe to call with AFU/MC in partially allocated/initialized state. */ @@ -751,7 +748,7 @@ static void cxlflash_remove(struct pci_dev *pdev) /** * alloc_mem() - allocates the AFU and its command pool - * @cxlflash: Internal structure associated with the host. + * @cfg: Internal structure associated with the host. * * A partially allocated state remains on failure. * @@ -804,12 +801,9 @@ out: /** * init_pci() - initializes the host as a PCI device - * @cxlflash: Internal structure associated with the host. + * @cfg: Internal structure associated with the host. * - * Return: - * 0 on success - * -EIO on unable to communicate with device - * A return code from the PCI sub-routines + * Return: 0 on success, -errno on failure */ static int init_pci(struct cxlflash_cfg *cfg) { @@ -889,11 +883,9 @@ out_release_regions: /** * init_scsi() - adds the host to the SCSI stack and kicks off host scan - * @cxlflash: Internal structure associated with the host. + * @cfg: Internal structure associated with the host. * - * Return: - * 0 on success - * A return code from adding the host + * Return: 0 on success, -errno on failure */ static int init_scsi(struct cxlflash_cfg *cfg) { @@ -1357,7 +1349,7 @@ out: /** * start_context() - starts the master context - * @cxlflash: Internal structure associated with the host. + * @cfg: Internal structure associated with the host. * * Return: A success or failure value from CXL services. */ @@ -1375,12 +1367,10 @@ static int start_context(struct cxlflash_cfg *cfg) /** * read_vpd() - obtains the WWPNs from VPD - * @cxlflash: Internal structure associated with the host. + * @cfg: Internal structure associated with the host. * @wwpn: Array of size NUM_FC_PORTS to pass back WWPNs * - * Return: - * 0 on success - * -ENODEV when VPD or WWPN keywords not found + * Return: 0 on success,
[PATCH v4 28/32] MAINTAINERS: Add cxlflash driver
Add stanza for cxlflash SCSI driver. Signed-off-by: Matthew R. Ochs Signed-off-by: Manoj N. Kumar Reviewed-by: Brian King Reviewed-by: Andrew Donnellan --- MAINTAINERS | 9 + 1 file changed, 9 insertions(+) diff --git a/MAINTAINERS b/MAINTAINERS index 274f854..f2f3046 100644 --- a/MAINTAINERS +++ b/MAINTAINERS @@ -3153,6 +3153,15 @@ F: Documentation/powerpc/cxl.txt F: Documentation/powerpc/cxl.txt F: Documentation/ABI/testing/sysfs-class-cxl +CXLFLASH (IBM Coherent Accelerator Processor Interface CAPI Flash) SCSI DRIVER +M: Manoj N. Kumar +M: Matthew R. Ochs +L: linux-s...@vger.kernel.org +S: Supported +F: drivers/scsi/cxlflash/ +F: include/uapi/scsi/cxlflash_ioctls.h +F: Documentation/powerpc/cxlflash.txt + STMMAC ETHERNET DRIVER M: Giuseppe Cavallaro L: net...@vger.kernel.org -- 2.1.0 ___ Linuxppc-dev mailing list Linuxppc-dev@lists.ozlabs.org https://lists.ozlabs.org/listinfo/linuxppc-dev
[PATCH v4 27/32] cxlflash: Fix to prevent stale AFU RRQ
Following an adapter reset, the AFU RRQ that resides in host memory holds stale data. This can lead to a condition where the RRQ interrupt handler tries to process stale entries and/or endlessly loops due to an out of sync generation bit. To fix, the AFU RRQ in host memory needs to be cleared after each reset. Signed-off-by: Matthew R. Ochs Signed-off-by: Manoj N. Kumar Reviewed-by: Brian King --- drivers/scsi/cxlflash/main.c | 3 +++ 1 file changed, 3 insertions(+) diff --git a/drivers/scsi/cxlflash/main.c b/drivers/scsi/cxlflash/main.c index 24aedfb..ab11ee6 100644 --- a/drivers/scsi/cxlflash/main.c +++ b/drivers/scsi/cxlflash/main.c @@ -1598,6 +1598,9 @@ static int start_afu(struct cxlflash_cfg *cfg) init_pcr(cfg); + /* After an AFU reset, RRQ entries are stale, clear them */ + memset(&afu->rrq_entry, 0, sizeof(afu->rrq_entry)); + /* Initialize RRQ pointers */ afu->hrrq_start = &afu->rrq_entry[0]; afu->hrrq_end = &afu->rrq_entry[NUM_RRQ_ENTRY - 1]; -- 2.1.0 ___ Linuxppc-dev mailing list Linuxppc-dev@lists.ozlabs.org https://lists.ozlabs.org/listinfo/linuxppc-dev
[PATCH v4 26/32] cxlflash: Correct spelling, grammar, and alignment mistakes
There are several spelling and grammar mistakes throughout the driver. Additionally there are a handful of places where there are extra lines and unnecessary variables/statements. These are a nuisance and pollute the driver. Fix spelling and grammar issues. Update some comments for clarity and consistency. Remove extra lines and a few unneeded variables/statements. Signed-off-by: Matthew R. Ochs Signed-off-by: Manoj N. Kumar Reviewed-by: Brian King --- drivers/scsi/cxlflash/common.h| 2 -- drivers/scsi/cxlflash/main.c | 62 +-- drivers/scsi/cxlflash/sislite.h | 6 ++-- drivers/scsi/cxlflash/superpipe.c | 2 +- drivers/scsi/cxlflash/vlun.c | 14 - 5 files changed, 38 insertions(+), 48 deletions(-) diff --git a/drivers/scsi/cxlflash/common.h b/drivers/scsi/cxlflash/common.h index a810585..bbfe711 100644 --- a/drivers/scsi/cxlflash/common.h +++ b/drivers/scsi/cxlflash/common.h @@ -105,8 +105,6 @@ struct cxlflash_cfg { atomic_t scan_host_needed; struct cxl_afu *cxl_afu; - - struct pci_pool *cxlflash_cmd_pool; struct pci_dev *parent_dev; atomic_t recovery_threads; diff --git a/drivers/scsi/cxlflash/main.c b/drivers/scsi/cxlflash/main.c index 4893516..24aedfb 100644 --- a/drivers/scsi/cxlflash/main.c +++ b/drivers/scsi/cxlflash/main.c @@ -34,7 +34,6 @@ MODULE_AUTHOR("Manoj N. Kumar "); MODULE_AUTHOR("Matthew R. Ochs "); MODULE_LICENSE("GPL"); - /** * cmd_checkout() - checks out an AFU command * @afu: AFU to checkout from. @@ -730,7 +729,7 @@ static void cxlflash_remove(struct pci_dev *pdev) case INIT_STATE_SCSI: cxlflash_term_local_luns(cfg); scsi_remove_host(cfg->host); - /* Fall through */ + /* fall through */ case INIT_STATE_AFU: term_afu(cfg); cancel_work_sync(&cfg->work_q); @@ -763,9 +762,7 @@ static int alloc_mem(struct cxlflash_cfg *cfg) char *buf = NULL; struct device *dev = &cfg->dev->dev; - /* This allocation is about 12K, i.e. only 1 64k page -* and upto 4 4k pages -*/ + /* AFU is ~12k, i.e. only one 64k page or up to four 4k pages */ cfg->afu = (void *)__get_free_pages(GFP_KERNEL | __GFP_ZERO, get_order(sizeof(struct afu))); if (unlikely(!cfg->afu)) { @@ -1295,10 +1292,10 @@ static irqreturn_t cxlflash_async_err_irq(int irq, void *data) goto out; } - /* it is OK to clear AFU status before FC_ERROR */ + /* FYI, it is 'okay' to clear AFU status before FC_ERROR */ writeq_be(reg_unmasked, &global->regs.aintr_clear); - /* check each bit that is on */ + /* Check each bit that is on */ for (i = 0; reg_unmasked; i++, reg_unmasked = (reg_unmasked >> 1)) { info = find_ainfo(1ULL << i); if (((reg_unmasked & 0x1) == 0) || !info) @@ -1311,7 +1308,7 @@ static irqreturn_t cxlflash_async_err_irq(int irq, void *data) readq_be(&global->fc_regs[port][FC_STATUS / 8])); /* -* do link reset first, some OTHER errors will set FC_ERROR +* Do link reset first, some OTHER errors will set FC_ERROR * again if cleared before or w/o a reset */ if (info->action & LINK_RESET) { @@ -1326,7 +1323,7 @@ static irqreturn_t cxlflash_async_err_irq(int irq, void *data) reg = readq_be(&global->fc_regs[port][FC_ERROR / 8]); /* -* since all errors are unmasked, FC_ERROR and FC_ERRCAP +* Since all errors are unmasked, FC_ERROR and FC_ERRCAP * should be the same and tracing one is sufficient. */ @@ -1472,23 +1469,22 @@ static void init_pcr(struct cxlflash_cfg *cfg) for (i = 0; i < MAX_CONTEXT; i++) { ctrl_map = &afu->afu_map->ctrls[i].ctrl; - /* disrupt any clients that could be running */ - /* e. g. clients that survived a master restart */ + /* Disrupt any clients that could be running */ + /* e.g. clients that survived a master restart */ writeq_be(0, &ctrl_map->rht_start); writeq_be(0, &ctrl_map->rht_cnt_id); writeq_be(0, &ctrl_map->ctx_cap); } - /* copy frequently used fields into afu */ + /* Copy frequently used fields into afu */ afu->ctx_hndl = (u16) cxl_process_element(cfg->mcctx); - /* ctx_hndl is 16 bits in CAIA */ afu->host_map = &afu->afu_map->hosts[afu->ctx_hndl].host; afu->ctrl_map = &afu->afu_map->ctrls[afu->ctx_hndl].ctrl; /* Program the Endian Control for the master context */ writeq_be(SISL_ENDIAN_CTRL, &afu->host_map->endian_c
[PATCH v4 25/32] cxlflash: Fix to prevent EEH recovery failure
The process_sense() routine can perform a read capacity which can take some time to complete. If an EEH occurs while waiting on the read capacity, the EEH handler is unable to obtain the context's mutex in order to put the context in an error state. The EEH handler will sit and wait until the context is free, but this wait can last longer than the EEH handler tolerates, leading to a failed recovery. To address this issue, make the context unavailable to new, non-system owned threads and release the context while calling into process_sense(). After returning from process_sense() the context mutex is reacquired and the context is made available again. The context can be safely moved to the error state if needed during the unavailable window as no other threads will hold its reference. Signed-off-by: Matthew R. Ochs Signed-off-by: Manoj N. Kumar Reviewed-by: Brian King --- drivers/scsi/cxlflash/superpipe.c | 9 + 1 file changed, 9 insertions(+) diff --git a/drivers/scsi/cxlflash/superpipe.c b/drivers/scsi/cxlflash/superpipe.c index a6316f5..7283e83 100644 --- a/drivers/scsi/cxlflash/superpipe.c +++ b/drivers/scsi/cxlflash/superpipe.c @@ -1787,12 +1787,21 @@ static int cxlflash_disk_verify(struct scsi_device *sdev, * inquiry (i.e. the Unit attention is due to the WWN changing). */ if (verify->hint & DK_CXLFLASH_VERIFY_HINT_SENSE) { + /* Can't hold mutex across process_sense/read_cap16, +* since we could have an intervening EEH event. +*/ + ctxi->unavail = true; + mutex_unlock(&ctxi->mutex); rc = process_sense(sdev, verify); if (unlikely(rc)) { dev_err(dev, "%s: Failed to validate sense data (%d)\n", __func__, rc); + mutex_lock(&ctxi->mutex); + ctxi->unavail = false; goto out; } + mutex_lock(&ctxi->mutex); + ctxi->unavail = false; } switch (gli->mode) { -- 2.1.0 ___ Linuxppc-dev mailing list Linuxppc-dev@lists.ozlabs.org https://lists.ozlabs.org/listinfo/linuxppc-dev
[PATCH v4 24/32] cxlflash: Fix MMIO and endianness errors
Sparse uncovered several errors with MMIO operations (accessing directly) and handling endianness. These can cause issues when running in different environments. Introduce __iomem and proper endianness tags/swaps where appropriate to make driver sparse clean. Signed-off-by: Matthew R. Ochs Signed-off-by: Manoj N. Kumar Reviewed-by: Brian King --- drivers/scsi/cxlflash/common.h| 10 +- drivers/scsi/cxlflash/main.c | 25 + drivers/scsi/cxlflash/superpipe.c | 6 +++--- drivers/scsi/cxlflash/superpipe.h | 2 +- drivers/scsi/cxlflash/vlun.c | 4 ++-- 5 files changed, 24 insertions(+), 23 deletions(-) diff --git a/drivers/scsi/cxlflash/common.h b/drivers/scsi/cxlflash/common.h index 3be5754..a810585 100644 --- a/drivers/scsi/cxlflash/common.h +++ b/drivers/scsi/cxlflash/common.h @@ -164,9 +164,9 @@ struct afu { /* AFU HW */ struct cxl_ioctl_start_work work; - struct cxlflash_afu_map *afu_map; /* entire MMIO map */ - struct sisl_host_map *host_map; /* MC host map */ - struct sisl_ctrl_map *ctrl_map; /* MC control map */ + struct cxlflash_afu_map __iomem *afu_map; /* entire MMIO map */ + struct sisl_host_map __iomem *host_map; /* MC host map */ + struct sisl_ctrl_map __iomem *ctrl_map; /* MC control map */ ctx_hndl_t ctx_hndl;/* master's context handle */ u64 *hrrq_start; @@ -188,10 +188,10 @@ struct afu { static inline u64 lun_to_lunid(u64 lun) { - u64 lun_id; + __be64 lun_id; int_to_scsilun(lun, (struct scsi_lun *)&lun_id); - return swab64(lun_id); + return be64_to_cpu(lun_id); } int cxlflash_afu_sync(struct afu *, ctx_hndl_t, res_hndl_t, u8); diff --git a/drivers/scsi/cxlflash/main.c b/drivers/scsi/cxlflash/main.c index 7c76227..4893516 100644 --- a/drivers/scsi/cxlflash/main.c +++ b/drivers/scsi/cxlflash/main.c @@ -644,7 +644,7 @@ static void stop_afu(struct cxlflash_cfg *cfg) complete(&afu->cmd[i].cevent); if (likely(afu->afu_map)) { - cxl_psa_unmap((void *)afu->afu_map); + cxl_psa_unmap((void __iomem *)afu->afu_map); afu->afu_map = NULL; } } @@ -914,7 +914,7 @@ out: * that the FC link layer has synced, completed the handshaking process, and * is ready for login to start. */ -static void set_port_online(u64 *fc_regs) +static void set_port_online(__be64 __iomem *fc_regs) { u64 cmdcfg; @@ -930,7 +930,7 @@ static void set_port_online(u64 *fc_regs) * * The provided MMIO region must be mapped prior to call. */ -static void set_port_offline(u64 *fc_regs) +static void set_port_offline(__be64 __iomem *fc_regs) { u64 cmdcfg; @@ -954,7 +954,7 @@ static void set_port_offline(u64 *fc_regs) * FALSE (0) when the specified port fails to come online after timeout * -EINVAL when @delay_us is less than 1000 */ -static int wait_port_online(u64 *fc_regs, u32 delay_us, u32 nretry) +static int wait_port_online(__be64 __iomem *fc_regs, u32 delay_us, u32 nretry) { u64 status; @@ -985,7 +985,7 @@ static int wait_port_online(u64 *fc_regs, u32 delay_us, u32 nretry) * FALSE (0) when the specified port fails to go offline after timeout * -EINVAL when @delay_us is less than 1000 */ -static int wait_port_offline(u64 *fc_regs, u32 delay_us, u32 nretry) +static int wait_port_offline(__be64 __iomem *fc_regs, u32 delay_us, u32 nretry) { u64 status; @@ -1020,7 +1020,8 @@ static int wait_port_offline(u64 *fc_regs, u32 delay_us, u32 nretry) * 0 when the WWPN is successfully written and the port comes back online * -1 when the port fails to go offline or come back up online */ -static int afu_set_wwpn(struct afu *afu, int port, u64 *fc_regs, u64 wwpn) +static int afu_set_wwpn(struct afu *afu, int port, __be64 __iomem *fc_regs, + u64 wwpn) { int rc = 0; @@ -1065,7 +1066,7 @@ static int afu_set_wwpn(struct afu *afu, int port, u64 *fc_regs, u64 wwpn) * the alternate port exclusively while the reset takes place. * failure to come online is overridden. */ -static void afu_link_reset(struct afu *afu, int port, u64 *fc_regs) +static void afu_link_reset(struct afu *afu, int port, __be64 __iomem *fc_regs) { u64 port_sel; @@ -1280,7 +1281,7 @@ static irqreturn_t cxlflash_async_err_irq(int irq, void *data) struct device *dev = &cfg->dev->dev; u64 reg_unmasked; const struct asyc_intr_info *info; - struct sisl_global_map *global = &afu->afu_map->global; + struct sisl_global_map __iomem *global = &afu->afu_map->global; u64 reg; u8 port; int i; @@ -1466,7 +1467,7 @@ out: static void init_pcr(struct cxlflash_cfg *cfg) { struct afu *afu = cfg->afu; - struct sisl_ctrl_map *ctrl_map; + struc
[PATCH v4 22/32] cxlflash: Remove unnecessary scsi_block_requests
The host reset handler is called with I/O already blocked, thus there is no need to explicitly block and unblock I/O in the handler. Signed-off-by: Matthew R. Ochs Signed-off-by: Manoj N. Kumar Reviewed-by: Brian King --- drivers/scsi/cxlflash/main.c | 2 -- 1 file changed, 2 deletions(-) diff --git a/drivers/scsi/cxlflash/main.c b/drivers/scsi/cxlflash/main.c index cf6a067..729f742 100644 --- a/drivers/scsi/cxlflash/main.c +++ b/drivers/scsi/cxlflash/main.c @@ -1966,7 +1966,6 @@ static int cxlflash_eh_host_reset_handler(struct scsi_cmnd *scp) switch (cfg->state) { case STATE_NORMAL: cfg->state = STATE_RESET; - scsi_block_requests(cfg->host); cxlflash_mark_contexts_error(cfg); rcr = afu_reset(cfg); if (rcr) { @@ -1975,7 +1974,6 @@ static int cxlflash_eh_host_reset_handler(struct scsi_cmnd *scp) } else cfg->state = STATE_NORMAL; wake_up_all(&cfg->reset_waitq); - scsi_unblock_requests(cfg->host); break; case STATE_RESET: wait_event(cfg->reset_waitq, cfg->state != STATE_RESET); -- 2.1.0 ___ Linuxppc-dev mailing list Linuxppc-dev@lists.ozlabs.org https://lists.ozlabs.org/listinfo/linuxppc-dev
[PATCH v4 21/32] cxlflash: Correct behavior in device reset handler following EEH
When the device reset handler is entered while a reset operation is taking place, the handler exits without actually sending a reset (TMF) to the targeted device. This behavior is incorrect as the device is not reset. Further complicating matters is the fact that a success is returned even when the TMF was not sent. To fix, the state is rechecked after coming out of the reset state. When the state is normal, a TMF will be sent out. Signed-off-by: Matthew R. Ochs Signed-off-by: Manoj N. Kumar Reviewed-by: Brian King --- drivers/scsi/cxlflash/main.c | 5 ++--- 1 file changed, 2 insertions(+), 3 deletions(-) diff --git a/drivers/scsi/cxlflash/main.c b/drivers/scsi/cxlflash/main.c index fada355..cf6a067 100644 --- a/drivers/scsi/cxlflash/main.c +++ b/drivers/scsi/cxlflash/main.c @@ -1920,6 +1920,7 @@ static int cxlflash_eh_device_reset_handler(struct scsi_cmnd *scp) get_unaligned_be32(&((u32 *)scp->cmnd)[2]), get_unaligned_be32(&((u32 *)scp->cmnd)[3])); +retry: switch (cfg->state) { case STATE_NORMAL: rcr = send_tmf(afu, scp, TMF_LUN_RESET); @@ -1928,9 +1929,7 @@ static int cxlflash_eh_device_reset_handler(struct scsi_cmnd *scp) break; case STATE_RESET: wait_event(cfg->reset_waitq, cfg->state != STATE_RESET); - if (cfg->state == STATE_NORMAL) - break; - /* fall through */ + goto retry; default: rc = FAILED; break; -- 2.1.0 ___ Linuxppc-dev mailing list Linuxppc-dev@lists.ozlabs.org https://lists.ozlabs.org/listinfo/linuxppc-dev
[PATCH v4 20/32] cxlflash: Fix to prevent workq from accessing freed memory
The workq can process work in parallel with a remove event, leading to a condition where the workq handler can access freed memory. To remedy, the workq should be terminated prior to freeing memory. Move the termination call earlier in remove and use cancel_work_sync() instead of flush_work() as there is not a need to process any scheduled work when shutting down. Signed-off-by: Matthew R. Ochs Signed-off-by: Manoj N. Kumar Reviewed-by: Brian King --- drivers/scsi/cxlflash/main.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/drivers/scsi/cxlflash/main.c b/drivers/scsi/cxlflash/main.c index 0bb6629..fada355 100644 --- a/drivers/scsi/cxlflash/main.c +++ b/drivers/scsi/cxlflash/main.c @@ -736,11 +736,11 @@ static void cxlflash_remove(struct pci_dev *pdev) /* Fall through */ case INIT_STATE_AFU: term_afu(cfg); + cancel_work_sync(&cfg->work_q); case INIT_STATE_PCI: pci_release_regions(cfg->dev); pci_disable_device(pdev); case INIT_STATE_NONE: - flush_work(&cfg->work_q); free_mem(cfg); scsi_host_put(cfg->host); break; -- 2.1.0 ___ Linuxppc-dev mailing list Linuxppc-dev@lists.ozlabs.org https://lists.ozlabs.org/listinfo/linuxppc-dev
[PATCH v4 19/32] cxlflash: Correct usage of scsi_host_put()
Currently, scsi_host_put() is being called prematurely in the remove path and is missing entirely in an error cleanup path. The former can lead to memory being freed too early with subsequent access potentially corrupting data whilst the former would result in a memory leak. Move the usage on remove to be the last cleanup action taken and introduce a call to scsi_host_put() in the one initialization error path that does not use remove to cleanup. Signed-off-by: Matthew R. Ochs Signed-off-by: Manoj N. Kumar Reviewed-by: Brian King --- drivers/scsi/cxlflash/main.c | 3 ++- 1 file changed, 2 insertions(+), 1 deletion(-) diff --git a/drivers/scsi/cxlflash/main.c b/drivers/scsi/cxlflash/main.c index b6a2584..0bb6629 100644 --- a/drivers/scsi/cxlflash/main.c +++ b/drivers/scsi/cxlflash/main.c @@ -733,7 +733,6 @@ static void cxlflash_remove(struct pci_dev *pdev) case INIT_STATE_SCSI: cxlflash_term_local_luns(cfg); scsi_remove_host(cfg->host); - scsi_host_put(cfg->host); /* Fall through */ case INIT_STATE_AFU: term_afu(cfg); @@ -743,6 +742,7 @@ static void cxlflash_remove(struct pci_dev *pdev) case INIT_STATE_NONE: flush_work(&cfg->work_q); free_mem(cfg); + scsi_host_put(cfg->host); break; } @@ -2404,6 +2404,7 @@ static int cxlflash_probe(struct pci_dev *pdev, dev_err(&pdev->dev, "%s: call to scsi_host_alloc failed!\n", __func__); rc = -ENOMEM; + scsi_host_put(cfg->host); goto out; } -- 2.1.0 ___ Linuxppc-dev mailing list Linuxppc-dev@lists.ozlabs.org https://lists.ozlabs.org/listinfo/linuxppc-dev
[PATCH v4 13/32] cxlflash: Fix to avoid stall while waiting on TMF
Borrowing the TMF waitq's spinlock causes a stall condition when waiting for the TMF to complete. To remedy, introduce our own spin lock to serialize TMF and use the appropriate wait services. Also add a timeout while waiting for a TMF completion. When a TMF times out, report back a failure such that a bigger hammer reset can occur. Signed-off-by: Matthew R. Ochs Signed-off-by: Manoj N. Kumar Reviewed-by: Brian King --- drivers/scsi/cxlflash/common.h | 1 + drivers/scsi/cxlflash/main.c | 55 +- 2 files changed, 34 insertions(+), 22 deletions(-) diff --git a/drivers/scsi/cxlflash/common.h b/drivers/scsi/cxlflash/common.h index b038ac7..7a0cb5c 100644 --- a/drivers/scsi/cxlflash/common.h +++ b/drivers/scsi/cxlflash/common.h @@ -124,6 +124,7 @@ struct cxlflash_cfg { struct list_head lluns; /* list of llun_info structs */ wait_queue_head_t tmf_waitq; + spinlock_t tmf_slock; bool tmf_active; wait_queue_head_t reset_waitq; enum cxlflash_state state; diff --git a/drivers/scsi/cxlflash/main.c b/drivers/scsi/cxlflash/main.c index 527ff85..110037d 100644 --- a/drivers/scsi/cxlflash/main.c +++ b/drivers/scsi/cxlflash/main.c @@ -249,11 +249,10 @@ static void cmd_complete(struct afu_cmd *cmd) scp->scsi_done(scp); if (cmd_is_tmf) { - spin_lock_irqsave(&cfg->tmf_waitq.lock, lock_flags); + spin_lock_irqsave(&cfg->tmf_slock, lock_flags); cfg->tmf_active = false; wake_up_all_locked(&cfg->tmf_waitq); - spin_unlock_irqrestore(&cfg->tmf_waitq.lock, - lock_flags); + spin_unlock_irqrestore(&cfg->tmf_slock, lock_flags); } } else complete(&cmd->cevent); @@ -420,6 +419,7 @@ static int send_tmf(struct afu *afu, struct scsi_cmnd *scp, u64 tmfcmd) struct device *dev = &cfg->dev->dev; ulong lock_flags; int rc = 0; + ulong to; cmd = cmd_checkout(afu); if (unlikely(!cmd)) { @@ -428,15 +428,15 @@ static int send_tmf(struct afu *afu, struct scsi_cmnd *scp, u64 tmfcmd) goto out; } - /* If a Task Management Function is active, do not send one more. -*/ - spin_lock_irqsave(&cfg->tmf_waitq.lock, lock_flags); + /* When Task Management Function is active do not send another */ + spin_lock_irqsave(&cfg->tmf_slock, lock_flags); if (cfg->tmf_active) - wait_event_interruptible_locked_irq(cfg->tmf_waitq, - !cfg->tmf_active); + wait_event_interruptible_lock_irq(cfg->tmf_waitq, + !cfg->tmf_active, + cfg->tmf_slock); cfg->tmf_active = true; cmd->cmd_tmf = true; - spin_unlock_irqrestore(&cfg->tmf_waitq.lock, lock_flags); + spin_unlock_irqrestore(&cfg->tmf_slock, lock_flags); cmd->rcb.ctx_id = afu->ctx_hndl; cmd->rcb.port_sel = port_sel; @@ -457,15 +457,24 @@ static int send_tmf(struct afu *afu, struct scsi_cmnd *scp, u64 tmfcmd) rc = send_cmd(afu, cmd); if (unlikely(rc)) { cmd_checkin(cmd); - spin_lock_irqsave(&cfg->tmf_waitq.lock, lock_flags); + spin_lock_irqsave(&cfg->tmf_slock, lock_flags); cfg->tmf_active = false; - spin_unlock_irqrestore(&cfg->tmf_waitq.lock, lock_flags); + spin_unlock_irqrestore(&cfg->tmf_slock, lock_flags); goto out; } - spin_lock_irqsave(&cfg->tmf_waitq.lock, lock_flags); - wait_event_interruptible_locked_irq(cfg->tmf_waitq, !cfg->tmf_active); - spin_unlock_irqrestore(&cfg->tmf_waitq.lock, lock_flags); + spin_lock_irqsave(&cfg->tmf_slock, lock_flags); + to = msecs_to_jiffies(5000); + to = wait_event_interruptible_lock_irq_timeout(cfg->tmf_waitq, + !cfg->tmf_active, + cfg->tmf_slock, + to); + if (!to) { + cfg->tmf_active = false; + dev_err(dev, "%s: TMF timed out!\n", __func__); + rc = -1; + } + spin_unlock_irqrestore(&cfg->tmf_slock, lock_flags); out: return rc; } @@ -512,16 +521,17 @@ static int cxlflash_queuecommand(struct Scsi_Host *host, struct scsi_cmnd *scp) get_unaligned_be32(&((u32 *)scp->cmnd)[2]), get_unaligned_be32(&((u32 *)scp->cmnd)[3])); - /* If a Task Management Function is active, wait for it to complete + /* +* If a Task Management Function is active, wait for it to compl
[PATCH v4 18/32] cxlflash: Fix AFU version access/storage and add check
The AFU version is stored as a non-terminated string of bytes within a 64-bit little-endian register. Presently the value is read directly (no MMIO accessor) and is stored in a buffer that is not big enough to contain a NULL terminator. Additionally the version obtained is not evaluated against a known value to prevent usage with unsupported AFUs. All of these deficiencies can lead to a variety of problems. To remedy, use the correct MMIO accessor to read the version value into a null-terminated buffer and add a check to prevent an incompatible AFU from being used with this driver. Signed-off-by: Matthew R. Ochs Signed-off-by: Manoj N. Kumar Reviewed-by: Brian King --- drivers/scsi/cxlflash/common.h | 2 +- drivers/scsi/cxlflash/main.c| 18 -- drivers/scsi/cxlflash/sislite.h | 2 +- 3 files changed, 14 insertions(+), 8 deletions(-) diff --git a/drivers/scsi/cxlflash/common.h b/drivers/scsi/cxlflash/common.h index faf7f56..3be5754 100644 --- a/drivers/scsi/cxlflash/common.h +++ b/drivers/scsi/cxlflash/common.h @@ -179,7 +179,7 @@ struct afu { u32 cmd_couts; /* Number of command checkouts */ u32 internal_lun; /* User-desired LUN mode for this AFU */ - char version[8]; + char version[16]; u64 interface_version; struct cxlflash_cfg *parent; /* Pointer back to parent cxlflash_cfg */ diff --git a/drivers/scsi/cxlflash/main.c b/drivers/scsi/cxlflash/main.c index d45388f..b6a2584 100644 --- a/drivers/scsi/cxlflash/main.c +++ b/drivers/scsi/cxlflash/main.c @@ -1751,14 +1751,20 @@ static int init_afu(struct cxlflash_cfg *cfg) goto err1; } - /* don't byte reverse on reading afu_version, else the string form */ - /* will be backwards */ - reg = afu->afu_map->global.regs.afu_version; - memcpy(afu->version, ®, 8); + /* No byte reverse on reading afu_version or string will be backwards */ + reg = readq(&afu->afu_map->global.regs.afu_version); + memcpy(afu->version, ®, sizeof(reg)); afu->interface_version = readq_be(&afu->afu_map->global.regs.interface_version); - pr_debug("%s: afu version %s, interface version 0x%llX\n", -__func__, afu->version, afu->interface_version); + if ((afu->interface_version + 1) == 0) { + pr_err("Back level AFU, please upgrade. AFU version %s " + "interface version 0x%llx\n", afu->version, + afu->interface_version); + rc = -EINVAL; + goto err1; + } else + pr_debug("%s: afu version %s, interface version 0x%llX\n", +__func__, afu->version, afu->interface_version); rc = start_afu(cfg); if (rc) { diff --git a/drivers/scsi/cxlflash/sislite.h b/drivers/scsi/cxlflash/sislite.h index 63bf394..8425d1a 100644 --- a/drivers/scsi/cxlflash/sislite.h +++ b/drivers/scsi/cxlflash/sislite.h @@ -340,7 +340,7 @@ struct sisl_global_regs { #define SISL_AFUCONF_MBOX_CLR_READ 0x0010ULL __be64 afu_config; __be64 rsvd[0xf8]; - __be64 afu_version; + __le64 afu_version; __be64 interface_version; }; -- 2.1.0 ___ Linuxppc-dev mailing list Linuxppc-dev@lists.ozlabs.org https://lists.ozlabs.org/listinfo/linuxppc-dev
[PATCH v4 17/32] cxlflash: Remove dual port online dependency
At present, both ports must be online for the device to configure properly. Remove this dependency and the unnecessary internal LUN override logic as well. Additionally, as a refactoring measure, change the return code variable name to match that used throughout the driver. Signed-off-by: Matthew R. Ochs Signed-off-by: Manoj N. Kumar Reviewed-by: Brian King --- drivers/scsi/cxlflash/main.c | 23 --- 1 file changed, 8 insertions(+), 15 deletions(-) diff --git a/drivers/scsi/cxlflash/main.c b/drivers/scsi/cxlflash/main.c index ed9fd8c..d45388f 100644 --- a/drivers/scsi/cxlflash/main.c +++ b/drivers/scsi/cxlflash/main.c @@ -1030,7 +1030,7 @@ static int wait_port_offline(u64 *fc_regs, u32 delay_us, u32 nretry) */ static int afu_set_wwpn(struct afu *afu, int port, u64 *fc_regs, u64 wwpn) { - int ret = 0; + int rc = 0; set_port_offline(fc_regs); @@ -1038,33 +1038,26 @@ static int afu_set_wwpn(struct afu *afu, int port, u64 *fc_regs, u64 wwpn) FC_PORT_STATUS_RETRY_CNT)) { pr_debug("%s: wait on port %d to go offline timed out\n", __func__, port); - ret = -1; /* but continue on to leave the port back online */ + rc = -1; /* but continue on to leave the port back online */ } - if (ret == 0) + if (rc == 0) writeq_be(wwpn, &fc_regs[FC_PNAME / 8]); + /* Always return success after programming WWPN */ + rc = 0; + set_port_online(fc_regs); if (!wait_port_online(fc_regs, FC_PORT_STATUS_RETRY_INTERVAL_US, FC_PORT_STATUS_RETRY_CNT)) { pr_debug("%s: wait on port %d to go online timed out\n", __func__, port); - ret = -1; - - /* -* Override for internal lun!!! -*/ - if (afu->internal_lun) { - pr_debug("%s: Overriding port %d online timeout!!!\n", -__func__, port); - ret = 0; - } } - pr_debug("%s: returning rc=%d\n", __func__, ret); + pr_debug("%s: returning rc=%d\n", __func__, rc); - return ret; + return rc; } /** -- 2.1.0 ___ Linuxppc-dev mailing list Linuxppc-dev@lists.ozlabs.org https://lists.ozlabs.org/listinfo/linuxppc-dev
[PATCH v4 16/32] cxlflash: Fix async interrupt bypass logic
A bug was introduced earlier in the development cycle when cleaning up logic statements. Instead of skipping bits that are not set, set bits are skipped, causing async interrupts to not be handled correctly. To fix, simply add back in the proper evaluation for an unset bit. Signed-off-by: Matthew R. Ochs Signed-off-by: Manoj N. Kumar Reviewed-by: Brian King --- drivers/scsi/cxlflash/main.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/drivers/scsi/cxlflash/main.c b/drivers/scsi/cxlflash/main.c index 98fdac1..ed9fd8c 100644 --- a/drivers/scsi/cxlflash/main.c +++ b/drivers/scsi/cxlflash/main.c @@ -1315,7 +1315,7 @@ static irqreturn_t cxlflash_async_err_irq(int irq, void *data) /* check each bit that is on */ for (i = 0; reg_unmasked; i++, reg_unmasked = (reg_unmasked >> 1)) { info = find_ainfo(1ULL << i); - if ((reg_unmasked & 0x1) || !info) + if (((reg_unmasked & 0x1) == 0) || !info) continue; port = info->port; -- 2.1.0 ___ Linuxppc-dev mailing list Linuxppc-dev@lists.ozlabs.org https://lists.ozlabs.org/listinfo/linuxppc-dev
[PATCH v4 15/32] cxlflash: Fix host link up event handling
Following a link up event, the LUNs available to the host may have changed. Without rescanning the host, the LUN topology is unknown to the user. In such a state, the user would be unable to locate provisioned resources. To remedy, the host should be rescanned after a link up event. Signed-off-by: Matthew R. Ochs Signed-off-by: Manoj N. Kumar Reviewed-by: Brian King Conflicts: drivers/scsi/cxlflash/common.h --- drivers/scsi/cxlflash/common.h | 1 + drivers/scsi/cxlflash/main.c | 17 + drivers/scsi/cxlflash/main.h | 1 + 3 files changed, 15 insertions(+), 4 deletions(-) diff --git a/drivers/scsi/cxlflash/common.h b/drivers/scsi/cxlflash/common.h index 7a0cb5c..faf7f56 100644 --- a/drivers/scsi/cxlflash/common.h +++ b/drivers/scsi/cxlflash/common.h @@ -102,6 +102,7 @@ struct cxlflash_cfg { enum cxlflash_init_state init_state; enum cxlflash_lr_state lr_state; int lr_port; + atomic_t scan_host_needed; struct cxl_afu *cxl_afu; diff --git a/drivers/scsi/cxlflash/main.c b/drivers/scsi/cxlflash/main.c index 5503a40..98fdac1 100644 --- a/drivers/scsi/cxlflash/main.c +++ b/drivers/scsi/cxlflash/main.c @@ -1119,17 +1119,17 @@ static const struct asyc_intr_info ainfo[] = { {SISL_ASTATUS_FC0_CRC_T, "CRC threshold exceeded", 0, LINK_RESET}, {SISL_ASTATUS_FC0_LOGI_R, "login timed out, retrying", 0, 0}, {SISL_ASTATUS_FC0_LOGI_F, "login failed", 0, CLR_FC_ERROR}, - {SISL_ASTATUS_FC0_LOGI_S, "login succeeded", 0, 0}, + {SISL_ASTATUS_FC0_LOGI_S, "login succeeded", 0, SCAN_HOST}, {SISL_ASTATUS_FC0_LINK_DN, "link down", 0, 0}, - {SISL_ASTATUS_FC0_LINK_UP, "link up", 0, 0}, + {SISL_ASTATUS_FC0_LINK_UP, "link up", 0, SCAN_HOST}, {SISL_ASTATUS_FC1_OTHER, "other error", 1, CLR_FC_ERROR | LINK_RESET}, {SISL_ASTATUS_FC1_LOGO, "target initiated LOGO", 1, 0}, {SISL_ASTATUS_FC1_CRC_T, "CRC threshold exceeded", 1, LINK_RESET}, {SISL_ASTATUS_FC1_LOGI_R, "login timed out, retrying", 1, 0}, {SISL_ASTATUS_FC1_LOGI_F, "login failed", 1, CLR_FC_ERROR}, - {SISL_ASTATUS_FC1_LOGI_S, "login succeeded", 1, 0}, + {SISL_ASTATUS_FC1_LOGI_S, "login succeeded", 1, SCAN_HOST}, {SISL_ASTATUS_FC1_LINK_DN, "link down", 1, 0}, - {SISL_ASTATUS_FC1_LINK_UP, "link up", 1, 0}, + {SISL_ASTATUS_FC1_LINK_UP, "link up", 1, SCAN_HOST}, {0x0, "", 0, 0} /* terminator */ }; @@ -1350,6 +1350,11 @@ static irqreturn_t cxlflash_async_err_irq(int irq, void *data) writeq_be(reg, &global->fc_regs[port][FC_ERROR / 8]); writeq_be(0, &global->fc_regs[port][FC_ERRCAP / 8]); } + + if (info->action & SCAN_HOST) { + atomic_inc(&cfg->scan_host_needed); + schedule_work(&cfg->work_q); + } } out: @@ -2309,6 +2314,7 @@ MODULE_DEVICE_TABLE(pci, cxlflash_pci_table); * - Link reset which cannot be performed on interrupt context due to * blocking up to a few seconds * - Read AFU command room + * - Rescan the host */ static void cxlflash_worker_thread(struct work_struct *work) { @@ -2351,6 +2357,9 @@ static void cxlflash_worker_thread(struct work_struct *work) } spin_unlock_irqrestore(cfg->host->host_lock, lock_flags); + + if (atomic_dec_if_positive(&cfg->scan_host_needed) >= 0) + scsi_scan_host(cfg->host); } /** diff --git a/drivers/scsi/cxlflash/main.h b/drivers/scsi/cxlflash/main.h index cf0e809..6032456 100644 --- a/drivers/scsi/cxlflash/main.h +++ b/drivers/scsi/cxlflash/main.h @@ -99,6 +99,7 @@ struct asyc_intr_info { u8 action; #define CLR_FC_ERROR 0x01 #define LINK_RESET 0x02 +#define SCAN_HOST 0x04 }; #ifndef CONFIG_CXL_EEH -- 2.1.0 ___ Linuxppc-dev mailing list Linuxppc-dev@lists.ozlabs.org https://lists.ozlabs.org/listinfo/linuxppc-dev
[PATCH v4 14/32] cxlflash: Fix location of setting resid
The resid is incorrectly set which can lead to unnecessary retry attempts by the stack. This is due to resid _always_ being set using a value returned from the adapter. Instead, the value should only be interpreted and set when in an underrun scenario. Signed-off-by: Matthew R. Ochs Signed-off-by: Manoj N. Kumar Reviewed-by: Brian King --- drivers/scsi/cxlflash/main.c | 20 1 file changed, 8 insertions(+), 12 deletions(-) diff --git a/drivers/scsi/cxlflash/main.c b/drivers/scsi/cxlflash/main.c index 110037d..5503a40 100644 --- a/drivers/scsi/cxlflash/main.c +++ b/drivers/scsi/cxlflash/main.c @@ -107,6 +107,7 @@ static void process_cmd_err(struct afu_cmd *cmd, struct scsi_cmnd *scp) { struct sisl_ioarcb *ioarcb; struct sisl_ioasa *ioasa; + u32 resid; if (unlikely(!cmd)) return; @@ -115,9 +116,10 @@ static void process_cmd_err(struct afu_cmd *cmd, struct scsi_cmnd *scp) ioasa = &(cmd->sa); if (ioasa->rc.flags & SISL_RC_FLAGS_UNDERRUN) { - pr_debug("%s: cmd underrun cmd = %p scp = %p\n", -__func__, cmd, scp); - scp->result = (DID_ERROR << 16); + resid = ioasa->resid; + scsi_set_resid(scp, resid); + pr_debug("%s: cmd underrun cmd = %p scp = %p, resid = %d\n", +__func__, cmd, scp, resid); } if (ioasa->rc.flags & SISL_RC_FLAGS_OVERRUN) { @@ -158,8 +160,7 @@ static void process_cmd_err(struct afu_cmd *cmd, struct scsi_cmnd *scp) /* If the SISL_RC_FLAGS_OVERRUN flag was set, * then we will handle this error else where. * If not then we must handle it here. -* This is probably an AFU bug. We will -* attempt a retry to see if that resolves it. +* This is probably an AFU bug. */ scp->result = (DID_ERROR << 16); } @@ -183,7 +184,7 @@ static void process_cmd_err(struct afu_cmd *cmd, struct scsi_cmnd *scp) /* We have an AFU error */ switch (ioasa->rc.afu_rc) { case SISL_AFU_RC_NO_CHANNELS: - scp->result = (DID_MEDIUM_ERROR << 16); + scp->result = (DID_NO_CONNECT << 16); break; case SISL_AFU_RC_DATA_DMA_ERR: switch (ioasa->afu_extra) { @@ -217,7 +218,6 @@ static void process_cmd_err(struct afu_cmd *cmd, struct scsi_cmnd *scp) static void cmd_complete(struct afu_cmd *cmd) { struct scsi_cmnd *scp; - u32 resid; ulong lock_flags; struct afu *afu = cmd->parent; struct cxlflash_cfg *cfg = afu->parent; @@ -229,14 +229,11 @@ static void cmd_complete(struct afu_cmd *cmd) if (cmd->rcb.scp) { scp = cmd->rcb.scp; - if (unlikely(cmd->sa.rc.afu_rc || -cmd->sa.rc.scsi_rc || -cmd->sa.rc.fc_rc)) + if (unlikely(cmd->sa.ioasc)) process_cmd_err(cmd, scp); else scp->result = (DID_OK << 16); - resid = cmd->sa.resid; cmd_is_tmf = cmd->cmd_tmf; cmd_checkin(cmd); /* Don't use cmd after here */ @@ -244,7 +241,6 @@ static void cmd_complete(struct afu_cmd *cmd) "ioasc=%d\n", __func__, scp, scp->result, cmd->sa.ioasc); - scsi_set_resid(scp, resid); scsi_dma_unmap(scp); scp->scsi_done(scp); -- 2.1.0 ___ Linuxppc-dev mailing list Linuxppc-dev@lists.ozlabs.org https://lists.ozlabs.org/listinfo/linuxppc-dev
[PATCH v4 12/32] cxlflash: Fix to avoid spamming the kernel log
During run-time the driver can be very chatty and spam the system kernel log. Various print statements can be limited and/or moved to development-only mode. Additionally, numerous prints can be converted to trace the corresponding device. The following changes were made: - pr_debug to pr_devel - pr_debug to pr_debug_ratelimited - pr_err to dev_err - pr_debug to dev_dbg Signed-off-by: Matthew R. Ochs Signed-off-by: Manoj N. Kumar Reviewed-by: Brian King Conflicts: drivers/scsi/cxlflash/main.c --- drivers/scsi/cxlflash/main.c | 109 +++ 1 file changed, 59 insertions(+), 50 deletions(-) diff --git a/drivers/scsi/cxlflash/main.c b/drivers/scsi/cxlflash/main.c index b44212b..527ff85 100644 --- a/drivers/scsi/cxlflash/main.c +++ b/drivers/scsi/cxlflash/main.c @@ -58,8 +58,8 @@ static struct afu_cmd *cmd_checkout(struct afu *afu) cmd = &afu->cmd[k]; if (!atomic_dec_if_positive(&cmd->free)) { - pr_debug("%s: returning found index=%d\n", -__func__, cmd->slot); + pr_devel("%s: returning found index=%d cmd=%p\n", +__func__, cmd->slot, cmd); memset(cmd->buf, 0, CMD_BUFSIZE); memset(cmd->rcb.cdb, 0, sizeof(cmd->rcb.cdb)); return cmd; @@ -93,7 +93,7 @@ static void cmd_checkin(struct afu_cmd *cmd) return; } - pr_debug("%s: released cmd %p index=%d\n", __func__, cmd, cmd->slot); + pr_devel("%s: released cmd %p index=%d\n", __func__, cmd, cmd->slot); } /** @@ -127,7 +127,7 @@ static void process_cmd_err(struct afu_cmd *cmd, struct scsi_cmnd *scp) } pr_debug("%s: cmd failed afu_rc=%d scsi_rc=%d fc_rc=%d " -"afu_extra=0x%X, scsi_entra=0x%X, fc_extra=0x%X\n", +"afu_extra=0x%X, scsi_extra=0x%X, fc_extra=0x%X\n", __func__, ioasa->rc.afu_rc, ioasa->rc.scsi_rc, ioasa->rc.fc_rc, ioasa->afu_extra, ioasa->scsi_extra, ioasa->fc_extra); @@ -240,9 +240,9 @@ static void cmd_complete(struct afu_cmd *cmd) cmd_is_tmf = cmd->cmd_tmf; cmd_checkin(cmd); /* Don't use cmd after here */ - pr_debug("%s: calling scsi_set_resid, scp=%p " -"result=%X resid=%d\n", __func__, -scp, scp->result, resid); + pr_debug_ratelimited("%s: calling scsi_done scp=%p result=%X " +"ioasc=%d\n", __func__, scp, scp->result, +cmd->sa.ioasc); scsi_set_resid(scp, resid); scsi_dma_unmap(scp); @@ -417,12 +417,13 @@ static int send_tmf(struct afu *afu, struct scsi_cmnd *scp, u64 tmfcmd) short lflag = 0; struct Scsi_Host *host = scp->device->host; struct cxlflash_cfg *cfg = (struct cxlflash_cfg *)host->hostdata; + struct device *dev = &cfg->dev->dev; ulong lock_flags; int rc = 0; cmd = cmd_checkout(afu); if (unlikely(!cmd)) { - pr_err("%s: could not get a free command\n", __func__); + dev_err(dev, "%s: could not get a free command\n", __func__); rc = SCSI_MLQUEUE_HOST_BUSY; goto out; } @@ -493,7 +494,7 @@ static int cxlflash_queuecommand(struct Scsi_Host *host, struct scsi_cmnd *scp) { struct cxlflash_cfg *cfg = (struct cxlflash_cfg *)host->hostdata; struct afu *afu = cfg->afu; - struct pci_dev *pdev = cfg->dev; + struct device *dev = &cfg->dev->dev; struct afu_cmd *cmd; u32 port_sel = scp->device->channel + 1; int nseg, i, ncount; @@ -502,13 +503,14 @@ static int cxlflash_queuecommand(struct Scsi_Host *host, struct scsi_cmnd *scp) short lflag = 0; int rc = 0; - pr_debug("%s: (scp=%p) %d/%d/%d/%llu cdb=(%08X-%08X-%08X-%08X)\n", -__func__, scp, host->host_no, scp->device->channel, -scp->device->id, scp->device->lun, -get_unaligned_be32(&((u32 *)scp->cmnd)[0]), -get_unaligned_be32(&((u32 *)scp->cmnd)[1]), -get_unaligned_be32(&((u32 *)scp->cmnd)[2]), -get_unaligned_be32(&((u32 *)scp->cmnd)[3])); + dev_dbg_ratelimited(dev, "%s: (scp=%p) %d/%d/%d/%llu " + "cdb=(%08X-%08X-%08X-%08X)\n", + __func__, scp, host->host_no, scp->device->channel, + scp->device->id, scp->device->lun, + get_unaligned_be32(&((u32 *)scp->cmnd)[0]), + get_unaligned_be32(&((u32 *)scp->cmnd)[1]), + get_unaligned_be32(&((u32 *)scp->cmnd)[2]), + get_unaligned_be32(&((u32 *)scp->cmnd)[3])); /* If
[PATCH v4 11/32] cxlflash: Refine host/device attributes
Implement the following suggestions and add two new attributes to allow for debugging the port LUN table. - use scnprintf() instead of snprintf() - use DEVICE_ATTR_RO and DEVICE_ATTR_RW Suggested-by: Shane Seymour Signed-off-by: Matthew R. Ochs Signed-off-by: Manoj N. Kumar Reviewed-by: Brian King --- drivers/scsi/cxlflash/main.c | 180 +-- 1 file changed, 138 insertions(+), 42 deletions(-) diff --git a/drivers/scsi/cxlflash/main.c b/drivers/scsi/cxlflash/main.c index 226cefe..b44212b 100644 --- a/drivers/scsi/cxlflash/main.c +++ b/drivers/scsi/cxlflash/main.c @@ -1995,33 +1995,24 @@ static int cxlflash_change_queue_depth(struct scsi_device *sdev, int qdepth) /** * cxlflash_show_port_status() - queries and presents the current port status - * @dev: Generic device associated with the host owning the port. - * @attr: Device attribute representing the port. + * @port: Desired port for status reporting. + * @afu: AFU owning the specified port. * @buf: Buffer of length PAGE_SIZE to report back port status in ASCII. * * Return: The size of the ASCII string returned in @buf. */ -static ssize_t cxlflash_show_port_status(struct device *dev, -struct device_attribute *attr, -char *buf) +static ssize_t cxlflash_show_port_status(u32 port, struct afu *afu, char *buf) { - struct Scsi_Host *shost = class_to_shost(dev); - struct cxlflash_cfg *cfg = (struct cxlflash_cfg *)shost->hostdata; - struct afu *afu = cfg->afu; - char *disp_status; - int rc; - u32 port; u64 status; - u64 *fc_regs; + __be64 __iomem *fc_regs; - rc = kstrtouint((attr->attr.name + 4), 10, &port); - if (rc || (port >= NUM_FC_PORTS)) + if (port >= NUM_FC_PORTS) return 0; fc_regs = &afu->afu_map->global.fc_regs[port][0]; - status = - (readq_be(&fc_regs[FC_MTIP_STATUS / 8]) & FC_MTIP_STATUS_MASK); + status = readq_be(&fc_regs[FC_MTIP_STATUS / 8]); + status &= FC_MTIP_STATUS_MASK; if (status == FC_MTIP_STATUS_ONLINE) disp_status = "online"; @@ -2030,31 +2021,69 @@ static ssize_t cxlflash_show_port_status(struct device *dev, else disp_status = "unknown"; - return snprintf(buf, PAGE_SIZE, "%s\n", disp_status); + return scnprintf(buf, PAGE_SIZE, "%s\n", disp_status); +} + +/** + * port0_show() - queries and presents the current status of port 0 + * @dev: Generic device associated with the host owning the port. + * @attr: Device attribute representing the port. + * @buf: Buffer of length PAGE_SIZE to report back port status in ASCII. + * + * Return: The size of the ASCII string returned in @buf. + */ +static ssize_t port0_show(struct device *dev, + struct device_attribute *attr, + char *buf) +{ + struct Scsi_Host *shost = class_to_shost(dev); + struct cxlflash_cfg *cfg = (struct cxlflash_cfg *)shost->hostdata; + struct afu *afu = cfg->afu; + + return cxlflash_show_port_status(0, afu, buf); } /** - * cxlflash_show_lun_mode() - presents the current LUN mode of the host + * port1_show() - queries and presents the current status of port 1 + * @dev: Generic device associated with the host owning the port. + * @attr: Device attribute representing the port. + * @buf: Buffer of length PAGE_SIZE to report back port status in ASCII. + * + * Return: The size of the ASCII string returned in @buf. + */ +static ssize_t port1_show(struct device *dev, + struct device_attribute *attr, + char *buf) +{ + struct Scsi_Host *shost = class_to_shost(dev); + struct cxlflash_cfg *cfg = (struct cxlflash_cfg *)shost->hostdata; + struct afu *afu = cfg->afu; + + return cxlflash_show_port_status(1, afu, buf); +} + +/** + * lun_mode_show() - presents the current LUN mode of the host * @dev: Generic device associated with the host. - * @attr: Device attribute representing the lun mode. + * @attr: Device attribute representing the LUN mode. * @buf: Buffer of length PAGE_SIZE to report back the LUN mode in ASCII. * * Return: The size of the ASCII string returned in @buf. */ -static ssize_t cxlflash_show_lun_mode(struct device *dev, - struct device_attribute *attr, char *buf) +static ssize_t lun_mode_show(struct device *dev, +struct device_attribute *attr, char *buf) { struct Scsi_Host *shost = class_to_shost(dev); struct cxlflash_cfg *cfg = (struct cxlflash_cfg *)shost->hostdata; struct afu *afu = cfg->afu; - return snprintf(buf, PAGE_SIZE, "%u\n", afu->internal_lun); + return scnprintf(buf, PAGE_SIZE, "%u\n", afu->internal
[PATCH v4 10/32] cxlflash: Make functions static
Found during code inspection, that the following functions are not being used outside of the file where they are defined. Make them static. int cxlflash_send_cmd(struct afu *, struct afu_cmd *); void cxlflash_wait_resp(struct afu *, struct afu_cmd *); int cxlflash_afu_reset(struct cxlflash_cfg *); struct afu_cmd *cxlflash_cmd_checkout(struct afu *); void cxlflash_cmd_checkin(struct afu_cmd *); void init_pcr(struct cxlflash_cfg *); int init_global(struct cxlflash_cfg *); Signed-off-by: Matthew R. Ochs Signed-off-by: Manoj N. Kumar Reviewed-by: Brian King --- drivers/scsi/cxlflash/common.h |5 - drivers/scsi/cxlflash/main.c | 1018 2 files changed, 509 insertions(+), 514 deletions(-) diff --git a/drivers/scsi/cxlflash/common.h b/drivers/scsi/cxlflash/common.h index 11318de..b038ac7 100644 --- a/drivers/scsi/cxlflash/common.h +++ b/drivers/scsi/cxlflash/common.h @@ -192,11 +192,6 @@ static inline u64 lun_to_lunid(u64 lun) return swab64(lun_id); } -int cxlflash_send_cmd(struct afu *, struct afu_cmd *); -void cxlflash_wait_resp(struct afu *, struct afu_cmd *); -int cxlflash_afu_reset(struct cxlflash_cfg *); -struct afu_cmd *cxlflash_cmd_checkout(struct afu *); -void cxlflash_cmd_checkin(struct afu_cmd *); int cxlflash_afu_sync(struct afu *, ctx_hndl_t, res_hndl_t, u8); void cxlflash_list_init(void); void cxlflash_term_global_luns(void); diff --git a/drivers/scsi/cxlflash/main.c b/drivers/scsi/cxlflash/main.c index 8940336..226cefe 100644 --- a/drivers/scsi/cxlflash/main.c +++ b/drivers/scsi/cxlflash/main.c @@ -36,7 +36,7 @@ MODULE_LICENSE("GPL"); /** - * cxlflash_cmd_checkout() - checks out an AFU command + * cmd_checkout() - checks out an AFU command * @afu: AFU to checkout from. * * Commands are checked out in a round-robin fashion. Note that since @@ -47,7 +47,7 @@ MODULE_LICENSE("GPL"); * * Return: The checked out command or NULL when command pool is empty. */ -struct afu_cmd *cxlflash_cmd_checkout(struct afu *afu) +static struct afu_cmd *cmd_checkout(struct afu *afu) { int k, dec = CXLFLASH_NUM_CMDS; struct afu_cmd *cmd; @@ -70,7 +70,7 @@ struct afu_cmd *cxlflash_cmd_checkout(struct afu *afu) } /** - * cxlflash_cmd_checkin() - checks in an AFU command + * cmd_checkin() - checks in an AFU command * @cmd: AFU command to checkin. * * Safe to pass commands that have already been checked in. Several @@ -79,7 +79,7 @@ struct afu_cmd *cxlflash_cmd_checkout(struct afu *afu) * to avoid clobbering values in the event that the command is checked * out right away. */ -void cxlflash_cmd_checkin(struct afu_cmd *cmd) +static void cmd_checkin(struct afu_cmd *cmd) { cmd->rcb.scp = NULL; cmd->rcb.timeout = 0; @@ -238,7 +238,7 @@ static void cmd_complete(struct afu_cmd *cmd) resid = cmd->sa.resid; cmd_is_tmf = cmd->cmd_tmf; - cxlflash_cmd_checkin(cmd); /* Don't use cmd after here */ + cmd_checkin(cmd); /* Don't use cmd after here */ pr_debug("%s: calling scsi_set_resid, scp=%p " "result=%X resid=%d\n", __func__, @@ -260,6 +260,146 @@ static void cmd_complete(struct afu_cmd *cmd) } /** + * context_reset() - timeout handler for AFU commands + * @cmd: AFU command that timed out. + * + * Sends a reset to the AFU. + */ +static void context_reset(struct afu_cmd *cmd) +{ + int nretry = 0; + u64 rrin = 0x1; + u64 room = 0; + struct afu *afu = cmd->parent; + ulong lock_flags; + + pr_debug("%s: cmd=%p\n", __func__, cmd); + + spin_lock_irqsave(&cmd->slock, lock_flags); + + /* Already completed? */ + if (cmd->sa.host_use_b[0] & B_DONE) { + spin_unlock_irqrestore(&cmd->slock, lock_flags); + return; + } + + cmd->sa.host_use_b[0] |= (B_DONE | B_ERROR | B_TIMEOUT); + spin_unlock_irqrestore(&cmd->slock, lock_flags); + + /* +* We really want to send this reset at all costs, so spread +* out wait time on successive retries for available room. +*/ + do { + room = readq_be(&afu->host_map->cmd_room); + atomic64_set(&afu->room, room); + if (room) + goto write_rrin; + udelay(nretry); + } while (nretry++ < MC_ROOM_RETRY_CNT); + + pr_err("%s: no cmd_room to send reset\n", __func__); + return; + +write_rrin: + nretry = 0; + writeq_be(rrin, &afu->host_map->ioarrin); + do { + rrin = readq_be(&afu->host_map->ioarrin); + if (rrin != 0x1) + break; + /* Double delay each time */ + udelay(2 ^ nretry); + } while (nretry++ < MC_ROOM_RETRY_CNT); +} + +/** + * send_cmd() - sends an AFU command + * @afu: AFU associated with the host. + * @cmd: AFU command to send. +
[PATCH v4 04/32] cxlflash: Fix potential oops following LUN removal
When a LUN is removed, the sdev that is associated with the LUN remains intact until its reference count drops to 0. In order to prevent an sdev from being removed while a context is still associated with it, obtain an additional reference per-context for each LUN attached to the context. This resolves a potential Oops in the release handler when a dealing with a LUN that has already been removed. Signed-off-by: Matthew R. Ochs Signed-off-by: Manoj N. Kumar Reviewed-by: Brian King --- drivers/scsi/cxlflash/superpipe.c | 35 +++ 1 file changed, 23 insertions(+), 12 deletions(-) diff --git a/drivers/scsi/cxlflash/superpipe.c b/drivers/scsi/cxlflash/superpipe.c index 4e44a48..ffa68cc 100644 --- a/drivers/scsi/cxlflash/superpipe.c +++ b/drivers/scsi/cxlflash/superpipe.c @@ -880,6 +880,9 @@ static int _cxlflash_disk_detach(struct scsi_device *sdev, sys_close(lfd); } + /* Release the sdev reference that bound this LUN to the context */ + scsi_device_put(sdev); + out: if (put_ctx) put_context(ctxi); @@ -1287,11 +1290,17 @@ static int cxlflash_disk_attach(struct scsi_device *sdev, } } + rc = scsi_device_get(sdev); + if (unlikely(rc)) { + dev_err(dev, "%s: Unable to get sdev reference!\n", __func__); + goto out; + } + lun_access = kzalloc(sizeof(*lun_access), GFP_KERNEL); if (unlikely(!lun_access)) { dev_err(dev, "%s: Unable to allocate lun_access!\n", __func__); rc = -ENOMEM; - goto out; + goto err0; } lun_access->lli = lli; @@ -1311,21 +1320,21 @@ static int cxlflash_disk_attach(struct scsi_device *sdev, dev_err(dev, "%s: Could not initialize context %p\n", __func__, ctx); rc = -ENODEV; - goto err0; + goto err1; } ctxid = cxl_process_element(ctx); if (unlikely((ctxid >= MAX_CONTEXT) || (ctxid < 0))) { dev_err(dev, "%s: ctxid (%d) invalid!\n", __func__, ctxid); rc = -EPERM; - goto err1; + goto err2; } file = cxl_get_fd(ctx, &cfg->cxl_fops, &fd); if (unlikely(fd < 0)) { rc = -ENODEV; dev_err(dev, "%s: Could not get file descriptor\n", __func__); - goto err1; + goto err2; } /* Translate read/write O_* flags from fcntl.h to AFU permission bits */ @@ -1335,7 +1344,7 @@ static int cxlflash_disk_attach(struct scsi_device *sdev, if (unlikely(!ctxi)) { dev_err(dev, "%s: Failed to create context! (%d)\n", __func__, ctxid); - goto err2; + goto err3; } work = &ctxi->work; @@ -1346,13 +1355,13 @@ static int cxlflash_disk_attach(struct scsi_device *sdev, if (unlikely(rc)) { dev_dbg(dev, "%s: Could not start context rc=%d\n", __func__, rc); - goto err3; + goto err4; } rc = afu_attach(cfg, ctxi); if (unlikely(rc)) { dev_err(dev, "%s: Could not attach AFU rc %d\n", __func__, rc); - goto err4; + goto err5; } /* @@ -1388,13 +1397,13 @@ out: __func__, ctxid, fd, attach->block_size, rc, attach->last_lba); return rc; -err4: +err5: cxl_stop_context(ctx); -err3: +err4: put_context(ctxi); destroy_context(cfg, ctxi); ctxi = NULL; -err2: +err3: /* * Here, we're overriding the fops with a dummy all-NULL fops because * fput() calls the release fop, which will cause us to mistakenly @@ -1406,10 +1415,12 @@ err2: fput(file); put_unused_fd(fd); fd = -1; -err1: +err2: cxl_release_context(ctx); -err0: +err1: kfree(lun_access); +err0: + scsi_device_put(sdev); goto out; } -- 2.1.0 ___ Linuxppc-dev mailing list Linuxppc-dev@lists.ozlabs.org https://lists.ozlabs.org/listinfo/linuxppc-dev
[PATCH v4 09/32] cxlflash: Correct naming of limbo state and waitq
Limbo is not an accurate representation of this state and is also not consistent with the terminology that other drivers use to represent this concept. Rename the state and and its associated waitq to 'reset'. Signed-off-by: Matthew R. Ochs Signed-off-by: Manoj N. Kumar Reviewed-by: Brian King --- drivers/scsi/cxlflash/common.h| 4 ++-- drivers/scsi/cxlflash/main.c | 26 +- drivers/scsi/cxlflash/superpipe.c | 14 +++--- 3 files changed, 22 insertions(+), 22 deletions(-) diff --git a/drivers/scsi/cxlflash/common.h b/drivers/scsi/cxlflash/common.h index 1abe4e0..11318de 100644 --- a/drivers/scsi/cxlflash/common.h +++ b/drivers/scsi/cxlflash/common.h @@ -79,7 +79,7 @@ enum cxlflash_init_state { enum cxlflash_state { STATE_NORMAL, /* Normal running state, everything good */ - STATE_LIMBO,/* Limbo running state, trying to reset/recover */ + STATE_RESET,/* Reset state, trying to reset/recover */ STATE_FAILTERM /* Failed/terminating state, error out users/threads */ }; @@ -125,7 +125,7 @@ struct cxlflash_cfg { wait_queue_head_t tmf_waitq; bool tmf_active; - wait_queue_head_t limbo_waitq; + wait_queue_head_t reset_waitq; enum cxlflash_state state; }; diff --git a/drivers/scsi/cxlflash/main.c b/drivers/scsi/cxlflash/main.c index 6e85c77..8940336 100644 --- a/drivers/scsi/cxlflash/main.c +++ b/drivers/scsi/cxlflash/main.c @@ -382,8 +382,8 @@ static int cxlflash_queuecommand(struct Scsi_Host *host, struct scsi_cmnd *scp) spin_unlock_irqrestore(&cfg->tmf_waitq.lock, lock_flags); switch (cfg->state) { - case STATE_LIMBO: - dev_dbg_ratelimited(&cfg->dev->dev, "%s: device in limbo!\n", + case STATE_RESET: + dev_dbg_ratelimited(&cfg->dev->dev, "%s: device is in reset!\n", __func__); rc = SCSI_MLQUEUE_HOST_BUSY; goto out; @@ -479,8 +479,8 @@ static int cxlflash_eh_device_reset_handler(struct scsi_cmnd *scp) if (unlikely(rcr)) rc = FAILED; break; - case STATE_LIMBO: - wait_event(cfg->limbo_waitq, cfg->state != STATE_LIMBO); + case STATE_RESET: + wait_event(cfg->reset_waitq, cfg->state != STATE_RESET); if (cfg->state == STATE_NORMAL) break; /* fall through */ @@ -519,7 +519,7 @@ static int cxlflash_eh_host_reset_handler(struct scsi_cmnd *scp) switch (cfg->state) { case STATE_NORMAL: - cfg->state = STATE_LIMBO; + cfg->state = STATE_RESET; scsi_block_requests(cfg->host); cxlflash_mark_contexts_error(cfg); rcr = cxlflash_afu_reset(cfg); @@ -528,11 +528,11 @@ static int cxlflash_eh_host_reset_handler(struct scsi_cmnd *scp) cfg->state = STATE_FAILTERM; } else cfg->state = STATE_NORMAL; - wake_up_all(&cfg->limbo_waitq); + wake_up_all(&cfg->reset_waitq); scsi_unblock_requests(cfg->host); break; - case STATE_LIMBO: - wait_event(cfg->limbo_waitq, cfg->state != STATE_LIMBO); + case STATE_RESET: + wait_event(cfg->reset_waitq, cfg->state != STATE_RESET); if (cfg->state == STATE_NORMAL) break; /* fall through */ @@ -705,7 +705,7 @@ static void cxlflash_wait_for_pci_err_recovery(struct cxlflash_cfg *cfg) struct pci_dev *pdev = cfg->dev; if (pci_channel_offline(pdev)) - wait_event_timeout(cfg->limbo_waitq, + wait_event_timeout(cfg->reset_waitq, !pci_channel_offline(pdev), CXLFLASH_PCI_ERROR_RECOVERY_TIMEOUT); } @@ -2304,7 +2304,7 @@ static int cxlflash_probe(struct pci_dev *pdev, cfg->mcctx = NULL; init_waitqueue_head(&cfg->tmf_waitq); - init_waitqueue_head(&cfg->limbo_waitq); + init_waitqueue_head(&cfg->reset_waitq); INIT_WORK(&cfg->work_q, cxlflash_worker_thread); cfg->lr_state = LINK_RESET_INVALID; @@ -2396,7 +2396,7 @@ static pci_ers_result_t cxlflash_pci_error_detected(struct pci_dev *pdev, switch (state) { case pci_channel_io_frozen: - cfg->state = STATE_LIMBO; + cfg->state = STATE_RESET; scsi_block_requests(cfg->host); drain_ioctls(cfg); rc = cxlflash_mark_contexts_error(cfg); @@ -2408,7 +2408,7 @@ static pci_ers_result_t cxlflash_pci_error_detected(struct pci_dev *pdev, return PCI_ERS_RESULT_NEED_RESET; case pci_channel_io_perm_failure: cfg->state = STATE_FAILTERM; - wake_up_all(&cfg->limbo_waitq); +
[PATCH v4 08/32] cxlflash: Fix to avoid CXL services during EEH
During an EEH freeze event, certain CXL services should not be called until after the hardware reset has taken place. Doing so can result in unnecessary failures and possibly cause other ill effects by triggering hardware accesses. This translates to a requirement to quiesce all threads that may potentially use CXL runtime service during this window. In particular, multiple ioctls make use of the CXL services when acting on contexts on behalf of the user. Thus, it is essential to 'drain' running ioctls _before_ proceeding with handling the EEH freeze event. Create the ability to drain ioctls by wrapping the ioctl handler call in a read semaphore and then implementing a small routine that obtains the write semaphore, effectively creating a wait point for all currently executing ioctls. Signed-off-by: Matthew R. Ochs Signed-off-by: Manoj N. Kumar --- drivers/scsi/cxlflash/common.h| 2 + drivers/scsi/cxlflash/main.c | 18 +-- drivers/scsi/cxlflash/superpipe.c | 98 --- 3 files changed, 77 insertions(+), 41 deletions(-) diff --git a/drivers/scsi/cxlflash/common.h b/drivers/scsi/cxlflash/common.h index 1c56037..1abe4e0 100644 --- a/drivers/scsi/cxlflash/common.h +++ b/drivers/scsi/cxlflash/common.h @@ -16,6 +16,7 @@ #define _CXLFLASH_COMMON_H #include +#include #include #include #include @@ -110,6 +111,7 @@ struct cxlflash_cfg { atomic_t recovery_threads; struct mutex ctx_recovery_mutex; struct mutex ctx_tbl_list_mutex; + struct rw_semaphore ioctl_rwsem; struct ctx_info *ctx_tbl[MAX_CONTEXT]; struct list_head ctx_err_recovery; /* contexts w/ recovery pending */ struct file_operations cxl_fops; diff --git a/drivers/scsi/cxlflash/main.c b/drivers/scsi/cxlflash/main.c index 3e3ccf1..6e85c77 100644 --- a/drivers/scsi/cxlflash/main.c +++ b/drivers/scsi/cxlflash/main.c @@ -2311,6 +2311,7 @@ static int cxlflash_probe(struct pci_dev *pdev, cfg->lr_port = -1; mutex_init(&cfg->ctx_tbl_list_mutex); mutex_init(&cfg->ctx_recovery_mutex); + init_rwsem(&cfg->ioctl_rwsem); INIT_LIST_HEAD(&cfg->ctx_err_recovery); INIT_LIST_HEAD(&cfg->lluns); @@ -2365,6 +2366,19 @@ out_remove: } /** + * drain_ioctls() - wait until all currently executing ioctls have completed + * @cfg: Internal structure associated with the host. + * + * Obtain write access to read/write semaphore that wraps ioctl + * handling to 'drain' ioctls currently executing. + */ +static void drain_ioctls(struct cxlflash_cfg *cfg) +{ + down_write(&cfg->ioctl_rwsem); + up_write(&cfg->ioctl_rwsem); +} + +/** * cxlflash_pci_error_detected() - called when a PCI error is detected * @pdev: PCI device struct. * @state: PCI channel state. @@ -2383,16 +2397,14 @@ static pci_ers_result_t cxlflash_pci_error_detected(struct pci_dev *pdev, switch (state) { case pci_channel_io_frozen: cfg->state = STATE_LIMBO; - - /* Turn off legacy I/O */ scsi_block_requests(cfg->host); + drain_ioctls(cfg); rc = cxlflash_mark_contexts_error(cfg); if (unlikely(rc)) dev_err(dev, "%s: Failed to mark user contexts!(%d)\n", __func__, rc); term_mc(cfg, UNDO_START); stop_afu(cfg); - return PCI_ERS_RESULT_NEED_RESET; case pci_channel_io_perm_failure: cfg->state = STATE_FAILTERM; diff --git a/drivers/scsi/cxlflash/superpipe.c b/drivers/scsi/cxlflash/superpipe.c index 28aa9d9..655cbf1 100644 --- a/drivers/scsi/cxlflash/superpipe.c +++ b/drivers/scsi/cxlflash/superpipe.c @@ -1214,6 +1214,46 @@ static const struct file_operations null_fops = { }; /** + * check_state() - checks and responds to the current adapter state + * @cfg: Internal structure associated with the host. + * + * This routine can block and should only be used on process context. + * It assumes that the caller is an ioctl thread and holding the ioctl + * read semaphore. This is temporarily let up across the wait to allow + * for draining actively running ioctls. Also note that when waking up + * from waiting in reset, the state is unknown and must be checked again + * before proceeding. + * + * Return: 0 on success, -errno on failure + */ +static int check_state(struct cxlflash_cfg *cfg) +{ + struct device *dev = &cfg->dev->dev; + int rc = 0; + +retry: + switch (cfg->state) { + case STATE_LIMBO: + dev_dbg(dev, "%s: Limbo state, going to wait...\n", __func__); + up_read(&cfg->ioctl_rwsem); + rc = wait_event_interruptible(cfg->limbo_waitq, + cfg->state != STATE_LIMBO); + down_read(&cfg->ioctl_rwsem); + if (unlikely(rc)) + break; + g
[PATCH v4 07/32] cxlflash: Fix context encode mask width
The context encode mask covers more than 32-bits, making it a long integer. This should be noted by appending the ULL width suffix to the mask. Signed-off-by: Matthew R. Ochs Signed-off-by: Manoj N. Kumar Reviewed-by: Brian King --- drivers/scsi/cxlflash/superpipe.h | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/drivers/scsi/cxlflash/superpipe.h b/drivers/scsi/cxlflash/superpipe.h index 72d53cf..7947091 100644 --- a/drivers/scsi/cxlflash/superpipe.h +++ b/drivers/scsi/cxlflash/superpipe.h @@ -87,7 +87,7 @@ enum ctx_ctrl { CTX_CTRL_FILE = (1 << 5) }; -#define ENCODE_CTXID(_ctx, _id)(u64)_ctx) & 0x0) << 28) | _id) +#define ENCODE_CTXID(_ctx, _id)(u64)_ctx) & 0x0ULL) << 28) | _id) #define DECODE_CTXID(_val) (_val & 0x) struct ctx_info { -- 2.1.0 ___ Linuxppc-dev mailing list Linuxppc-dev@lists.ozlabs.org https://lists.ozlabs.org/listinfo/linuxppc-dev
[PATCH v4 06/32] cxlflash: Fix to avoid sizeof(bool)
Using sizeof(bool) is considered poor form for various reasons and sparse warns us of that. Correct by changing type from bool to u8. Signed-off-by: Matthew R. Ochs Signed-off-by: Manoj N. Kumar Reviewed-by: Brian King --- drivers/scsi/cxlflash/superpipe.c | 2 +- drivers/scsi/cxlflash/superpipe.h | 2 +- 2 files changed, 2 insertions(+), 2 deletions(-) diff --git a/drivers/scsi/cxlflash/superpipe.c b/drivers/scsi/cxlflash/superpipe.c index ffa68cc..28aa9d9 100644 --- a/drivers/scsi/cxlflash/superpipe.c +++ b/drivers/scsi/cxlflash/superpipe.c @@ -737,7 +737,7 @@ static struct ctx_info *create_context(struct cxlflash_cfg *cfg, struct afu *afu = cfg->afu; struct ctx_info *ctxi = NULL; struct llun_info **lli = NULL; - bool *ws = NULL; + u8 *ws = NULL; struct sisl_rht_entry *rhte; ctxi = kzalloc(sizeof(*ctxi), GFP_KERNEL); diff --git a/drivers/scsi/cxlflash/superpipe.h b/drivers/scsi/cxlflash/superpipe.h index fffb179..72d53cf 100644 --- a/drivers/scsi/cxlflash/superpipe.h +++ b/drivers/scsi/cxlflash/superpipe.h @@ -97,7 +97,7 @@ struct ctx_info { u32 rht_out;/* Number of checked out RHT entries */ u32 rht_perms; /* User-defined permissions for RHT entries */ struct llun_info **rht_lun; /* Mapping of RHT entries to LUNs */ - bool *rht_needs_ws; /* User-desired write-same function per RHTE */ + u8 *rht_needs_ws; /* User-desired write-same function per RHTE */ struct cxl_ioctl_start_work work; u64 ctxid; -- 2.1.0 ___ Linuxppc-dev mailing list Linuxppc-dev@lists.ozlabs.org https://lists.ozlabs.org/listinfo/linuxppc-dev
[PATCH v4 05/32] cxlflash: Fix data corruption when vLUN used over multiple cards
If the same virtual LUN is accessed over multiple cards, only accesses made over the first card will be valid. Accesses made over the second card will go to the wrong LUN causing data corruption. This is because the global LUN's mode word was being used to determine whether the LUN table for that card needs to be programmed. The mode word would be setup by the first card, causing the LUN table for the second card to not be programmed. By unconditionally initializing the LUN table (not depending on the mode word), the problem is avoided. Signed-off-by: Matthew R. Ochs Signed-off-by: Manoj N. Kumar Reviewed-by: Brian King --- drivers/scsi/cxlflash/vlun.c | 16 1 file changed, 8 insertions(+), 8 deletions(-) diff --git a/drivers/scsi/cxlflash/vlun.c b/drivers/scsi/cxlflash/vlun.c index 68994c4..96b074f 100644 --- a/drivers/scsi/cxlflash/vlun.c +++ b/drivers/scsi/cxlflash/vlun.c @@ -915,16 +915,9 @@ int cxlflash_disk_virtual_open(struct scsi_device *sdev, void *arg) pr_debug("%s: ctxid=%llu ls=0x%llx\n", __func__, ctxid, lun_size); + /* Setup the LUNs block allocator on first call */ mutex_lock(&gli->mutex); if (gli->mode == MODE_NONE) { - /* Setup the LUN table and block allocator on first call */ - rc = init_luntable(cfg, lli); - if (rc) { - dev_err(dev, "%s: call to init_luntable failed " - "rc=%d!\n", __func__, rc); - goto err0; - } - rc = init_vlun(lli); if (rc) { dev_err(dev, "%s: call to init_vlun failed rc=%d!\n", @@ -942,6 +935,13 @@ int cxlflash_disk_virtual_open(struct scsi_device *sdev, void *arg) } mutex_unlock(&gli->mutex); + rc = init_luntable(cfg, lli); + if (rc) { + dev_err(dev, "%s: call to init_luntable failed rc=%d!\n", + __func__, rc); + goto err1; + } + ctxi = get_context(cfg, rctxid, lli, 0); if (unlikely(!ctxi)) { dev_err(dev, "%s: Bad context! (%llu)\n", __func__, ctxid); -- 2.1.0 ___ Linuxppc-dev mailing list Linuxppc-dev@lists.ozlabs.org https://lists.ozlabs.org/listinfo/linuxppc-dev
[PATCH v4 03/32] cxlflash: Fix read capacity timeout
From: Manoj Kumar The timeout value for read capacity is too small. Certain devices may take longer to respond and thus the command may prematurely timeout. Additionally the literal used for the timeout is stale. Update the timeout to 30 seconds (matches the value used in sd.c) and rework the timeout literal to a more appropriate description. Signed-off-by: Matthew R. Ochs Signed-off-by: Manoj N. Kumar Reviewed-by: Brian King --- drivers/scsi/cxlflash/superpipe.c | 9 - drivers/scsi/cxlflash/superpipe.h | 2 +- drivers/scsi/cxlflash/vlun.c | 4 ++-- 3 files changed, 7 insertions(+), 8 deletions(-) diff --git a/drivers/scsi/cxlflash/superpipe.c b/drivers/scsi/cxlflash/superpipe.c index 4db15a4..4e44a48 100644 --- a/drivers/scsi/cxlflash/superpipe.c +++ b/drivers/scsi/cxlflash/superpipe.c @@ -296,7 +296,7 @@ static int read_cap16(struct scsi_device *sdev, struct llun_info *lli) int rc = 0; int result = 0; int retry_cnt = 0; - u32 tout = (MC_DISCOVERY_TIMEOUT * HZ); + u32 to = CMD_TIMEOUT * HZ; retry: cmd_buf = kzalloc(CMD_BUFSIZE, GFP_KERNEL); @@ -315,8 +315,7 @@ retry: retry_cnt ? "re" : "", scsi_cmd[0]); result = scsi_execute(sdev, scsi_cmd, DMA_FROM_DEVICE, cmd_buf, - CMD_BUFSIZE, sense_buf, tout, CMD_RETRIES, - 0, NULL); + CMD_BUFSIZE, sense_buf, to, CMD_RETRIES, 0, NULL); if (driver_byte(result) == DRIVER_SENSE) { result &= ~(0xFF<<24); /* DRIVER_SENSE is not an error */ @@ -1376,8 +1375,8 @@ out_attach: attach->block_size = gli->blk_len; attach->mmio_size = sizeof(afu->afu_map->hosts[0].harea); attach->last_lba = gli->max_lba; - attach->max_xfer = (sdev->host->max_sectors * MAX_SECTOR_UNIT) / - gli->blk_len; + attach->max_xfer = sdev->host->max_sectors * MAX_SECTOR_UNIT; + attach->max_xfer /= gli->blk_len; out: attach->adap_fd = fd; diff --git a/drivers/scsi/cxlflash/superpipe.h b/drivers/scsi/cxlflash/superpipe.h index 3f7856b..fffb179 100644 --- a/drivers/scsi/cxlflash/superpipe.h +++ b/drivers/scsi/cxlflash/superpipe.h @@ -28,7 +28,7 @@ extern struct cxlflash_global global; */ #define MC_CHUNK_SIZE (1 << MC_RHT_NMASK) /* in LBAs */ -#define MC_DISCOVERY_TIMEOUT 5 /* 5 secs */ +#define CMD_TIMEOUT 30 /* 30 secs */ #define CMD_RETRIES 5 /* 5 retries for scsi_execute */ #define MAX_SECTOR_UNIT 512 /* max_sector is in 512 byte multiples */ diff --git a/drivers/scsi/cxlflash/vlun.c b/drivers/scsi/cxlflash/vlun.c index 6d6608b..68994c4 100644 --- a/drivers/scsi/cxlflash/vlun.c +++ b/drivers/scsi/cxlflash/vlun.c @@ -414,7 +414,7 @@ static int write_same16(struct scsi_device *sdev, int ws_limit = SISLITE_MAX_WS_BLOCKS; u64 offset = lba; int left = nblks; - u32 tout = sdev->request_queue->rq_timeout; + u32 to = sdev->request_queue->rq_timeout; struct cxlflash_cfg *cfg = (struct cxlflash_cfg *)sdev->host->hostdata; struct device *dev = &cfg->dev->dev; @@ -434,7 +434,7 @@ static int write_same16(struct scsi_device *sdev, &scsi_cmd[10]); result = scsi_execute(sdev, scsi_cmd, DMA_TO_DEVICE, cmd_buf, - CMD_BUFSIZE, sense_buf, tout, CMD_RETRIES, + CMD_BUFSIZE, sense_buf, to, CMD_RETRIES, 0, NULL); if (result) { dev_err_ratelimited(dev, "%s: command failed for " -- 2.1.0 ___ Linuxppc-dev mailing list Linuxppc-dev@lists.ozlabs.org https://lists.ozlabs.org/listinfo/linuxppc-dev
[PATCH v4 02/32] cxlflash: Replace magic numbers with literals
From: Manoj Kumar Magic numbers are not meaningful and can create confusion. As a remedy, replace them with descriptive literals. Replace 512 with literal MAX_SECTOR_UNIT. Replace 5 with literal CMD_RETRIES. Signed-off-by: Matthew R. Ochs Signed-off-by: Manoj N. Kumar Reviewed-by: Brian King --- drivers/scsi/cxlflash/superpipe.c | 6 -- drivers/scsi/cxlflash/superpipe.h | 3 +++ drivers/scsi/cxlflash/vlun.c | 3 ++- 3 files changed, 9 insertions(+), 3 deletions(-) diff --git a/drivers/scsi/cxlflash/superpipe.c b/drivers/scsi/cxlflash/superpipe.c index 05e0ecf..4db15a4 100644 --- a/drivers/scsi/cxlflash/superpipe.c +++ b/drivers/scsi/cxlflash/superpipe.c @@ -315,7 +315,8 @@ retry: retry_cnt ? "re" : "", scsi_cmd[0]); result = scsi_execute(sdev, scsi_cmd, DMA_FROM_DEVICE, cmd_buf, - CMD_BUFSIZE, sense_buf, tout, 5, 0, NULL); + CMD_BUFSIZE, sense_buf, tout, CMD_RETRIES, + 0, NULL); if (driver_byte(result) == DRIVER_SENSE) { result &= ~(0xFF<<24); /* DRIVER_SENSE is not an error */ @@ -1375,7 +1376,8 @@ out_attach: attach->block_size = gli->blk_len; attach->mmio_size = sizeof(afu->afu_map->hosts[0].harea); attach->last_lba = gli->max_lba; - attach->max_xfer = (sdev->host->max_sectors * 512) / gli->blk_len; + attach->max_xfer = (sdev->host->max_sectors * MAX_SECTOR_UNIT) / + gli->blk_len; out: attach->adap_fd = fd; diff --git a/drivers/scsi/cxlflash/superpipe.h b/drivers/scsi/cxlflash/superpipe.h index d7dc88b..3f7856b 100644 --- a/drivers/scsi/cxlflash/superpipe.h +++ b/drivers/scsi/cxlflash/superpipe.h @@ -29,6 +29,9 @@ extern struct cxlflash_global global; #define MC_CHUNK_SIZE (1 << MC_RHT_NMASK) /* in LBAs */ #define MC_DISCOVERY_TIMEOUT 5 /* 5 secs */ +#define CMD_RETRIES 5 /* 5 retries for scsi_execute */ + +#define MAX_SECTOR_UNIT 512 /* max_sector is in 512 byte multiples */ #define CHAN2PORT(_x) ((_x) + 1) #define PORT2CHAN(_x) ((_x) - 1) diff --git a/drivers/scsi/cxlflash/vlun.c b/drivers/scsi/cxlflash/vlun.c index 6155cb1..6d6608b 100644 --- a/drivers/scsi/cxlflash/vlun.c +++ b/drivers/scsi/cxlflash/vlun.c @@ -434,7 +434,8 @@ static int write_same16(struct scsi_device *sdev, &scsi_cmd[10]); result = scsi_execute(sdev, scsi_cmd, DMA_TO_DEVICE, cmd_buf, - CMD_BUFSIZE, sense_buf, tout, 5, 0, NULL); + CMD_BUFSIZE, sense_buf, tout, CMD_RETRIES, + 0, NULL); if (result) { dev_err_ratelimited(dev, "%s: command failed for " "offset %lld result=0x%x\n", -- 2.1.0 ___ Linuxppc-dev mailing list Linuxppc-dev@lists.ozlabs.org https://lists.ozlabs.org/listinfo/linuxppc-dev
[PATCH v4 01/32] cxlflash: Fix to avoid invalid port_sel value
From: Manoj Kumar If two concurrent MANAGE_LUN ioctls are issued with the same WWID parameter, it would result in an incorrect value of port_sel. This is because port_sel is modified without any locks being held. If the first caller stalls after the return from find_and_create_lun(), the value of port_sel will be set incorrectly to indicate a single port, though in this case it should have been set to both ports. To fix, use the global mutex to serialize the lookup of the WWID and the subsequent modification of port_sel. Signed-off-by: Matthew R. Ochs Signed-off-by: Manoj N. Kumar Reviewed-by: Brian King --- drivers/scsi/cxlflash/lunmgt.c | 9 ++--- 1 file changed, 6 insertions(+), 3 deletions(-) diff --git a/drivers/scsi/cxlflash/lunmgt.c b/drivers/scsi/cxlflash/lunmgt.c index d98ad0f..8c372fc 100644 --- a/drivers/scsi/cxlflash/lunmgt.c +++ b/drivers/scsi/cxlflash/lunmgt.c @@ -120,7 +120,8 @@ static struct glun_info *lookup_global(u8 *wwid) * * The LUN is kept both in a local list (per adapter) and in a global list * (across all adapters). Certain attributes of the LUN are local to the - * adapter (such as index, port selection mask etc.). + * adapter (such as index, port selection mask, etc.). + * * The block allocation map is shared across all adapters (i.e. associated * wih the global list). Since different attributes are associated with * the per adapter and global entries, allocate two separate structures for each @@ -128,6 +129,8 @@ static struct glun_info *lookup_global(u8 *wwid) * * Keep a pointer back from the local to the global entry. * + * This routine assumes the caller holds the global mutex. + * * Return: Found/Allocated local lun_info structure on success, NULL on failure */ static struct llun_info *find_and_create_lun(struct scsi_device *sdev, u8 *wwid) @@ -137,7 +140,6 @@ static struct llun_info *find_and_create_lun(struct scsi_device *sdev, u8 *wwid) struct Scsi_Host *shost = sdev->host; struct cxlflash_cfg *cfg = shost_priv(shost); - mutex_lock(&global.mutex); if (unlikely(!wwid)) goto out; @@ -169,7 +171,6 @@ static struct llun_info *find_and_create_lun(struct scsi_device *sdev, u8 *wwid) list_add(&gli->list, &global.gluns); out: - mutex_unlock(&global.mutex); pr_debug("%s: returning %p\n", __func__, lli); return lli; } @@ -235,6 +236,7 @@ int cxlflash_manage_lun(struct scsi_device *sdev, u64 flags = manage->hdr.flags; u32 chan = sdev->channel; + mutex_lock(&global.mutex); lli = find_and_create_lun(sdev, manage->wwid); pr_debug("%s: ENTER: WWID = %016llX%016llX, flags = %016llX li = %p\n", __func__, get_unaligned_le64(&manage->wwid[0]), @@ -261,6 +263,7 @@ int cxlflash_manage_lun(struct scsi_device *sdev, } out: + mutex_unlock(&global.mutex); pr_debug("%s: returning rc=%d\n", __func__, rc); return rc; } -- 2.1.0 ___ Linuxppc-dev mailing list Linuxppc-dev@lists.ozlabs.org https://lists.ozlabs.org/listinfo/linuxppc-dev
[PATCH v4 00/32] cxlflash: Miscellaneous bug fixes and corrections
This patch set contains various fixes and corrections for issues that were found during test and code review. The series is based upon the code upstreamed in 4.3 and is intended for the rc phase. The entire set is bisectable. Please reference the changelog below for details on what has been altered from previous versions of this patch set. v4 Changes: - Incorporate comments from Brian King - Removed unnecessary check_state() parameter from "Fix to avoid CXL..." - Added patch to fix potential deadlock on EEH - Removed patch to avoid state change collision - Changed fops initialization location in "Fix to avoid corrupting..." v3 Changes: - Rebased the series on top of patch by Dan Carpenter ("a couple off...") - Incorporate comments from David Laight - Incorporate comments from Tomas Henzl - Incorporate comments from Brian King - Removed patch to stop interrupt processing on remove - Removed double scsi_device_put() from "Fix potential oops" - Fixed usage of scnprintf() in "Refine host/device attributes" - Removed unnecessary parenthesis from "Fix read capacity timeout" - Added patch to use correct operator for doubling delay - Changed location of cancel_work_sync() in "Fix to prevent workq..." - Removed local mutex from cxlflash_afu_sync() in "Fix to avoid state..." - Added patch to correctly identify a failed function in a trace - Added patch to fix a fops corruption bug v2 Changes: - Incorporate comments from Ian Munsie - Rework commit messages to be more descriptive - Add state change serialization patch Manoj Kumar (4): cxlflash: Fix to avoid invalid port_sel value cxlflash: Replace magic numbers with literals cxlflash: Fix read capacity timeout cxlflash: Fix to double the delay each time Matthew R. Ochs (28): cxlflash: Fix potential oops following LUN removal cxlflash: Fix data corruption when vLUN used over multiple cards cxlflash: Fix to avoid sizeof(bool) cxlflash: Fix context encode mask width cxlflash: Fix to avoid CXL services during EEH cxlflash: Correct naming of limbo state and waitq cxlflash: Make functions static cxlflash: Refine host/device attributes cxlflash: Fix to avoid spamming the kernel log cxlflash: Fix to avoid stall while waiting on TMF cxlflash: Fix location of setting resid cxlflash: Fix host link up event handling cxlflash: Fix async interrupt bypass logic cxlflash: Remove dual port online dependency cxlflash: Fix AFU version access/storage and add check cxlflash: Correct usage of scsi_host_put() cxlflash: Fix to prevent workq from accessing freed memory cxlflash: Correct behavior in device reset handler following EEH cxlflash: Remove unnecessary scsi_block_requests cxlflash: Fix function prolog parameters and return codes cxlflash: Fix MMIO and endianness errors cxlflash: Fix to prevent EEH recovery failure cxlflash: Correct spelling, grammar, and alignment mistakes cxlflash: Fix to prevent stale AFU RRQ MAINTAINERS: Add cxlflash driver cxlflash: Fix to avoid corrupting adapter fops cxlflash: Correct trace string cxlflash: Fix to avoid potential deadlock on EEH MAINTAINERS |9 + drivers/scsi/cxlflash/common.h| 30 +- drivers/scsi/cxlflash/lunmgt.c|9 +- drivers/scsi/cxlflash/main.c | 1529 - drivers/scsi/cxlflash/main.h |1 + drivers/scsi/cxlflash/sislite.h |8 +- drivers/scsi/cxlflash/superpipe.c | 204 +++-- drivers/scsi/cxlflash/superpipe.h | 13 +- drivers/scsi/cxlflash/vlun.c | 68 +- 9 files changed, 1043 insertions(+), 828 deletions(-) -- 2.1.0 ___ Linuxppc-dev mailing list Linuxppc-dev@lists.ozlabs.org https://lists.ozlabs.org/listinfo/linuxppc-dev
Re: [V5, 2/6] fsl/fman: Add FMan support
On Mon, Sep 21, 2015 at 02:52:34PM +0300, Igal.Liberman wrote: > diff --git a/drivers/net/ethernet/freescale/fman/fman.c > b/drivers/net/ethernet/freescale/fman/fman.c > new file mode 100644 > index 000..924685f > --- /dev/null > +++ b/drivers/net/ethernet/freescale/fman/fman.c > @@ -0,0 +1,2738 @@ > +/* > + * Copyright 2008-2015 Freescale Semiconductor Inc. > + * > + * Redistribution and use in source and binary forms, with or without > + * modification, are permitted provided that the following conditions are > met: > + * * Redistributions of source code must retain the above copyright > + * notice, this list of conditions and the following disclaimer. > + * * Redistributions in binary form must reproduce the above copyright > + * notice, this list of conditions and the following disclaimer in the > + * documentation and/or other materials provided with the distribution. > + * * Neither the name of Freescale Semiconductor nor the > + * names of its contributors may be used to endorse or promote products > + * derived from this software without specific prior written > permission. > + * > +// * > + * ALTERNATIVELY, this software may be distributed under the terms of the > + * GNU General Public License ("GPL") as published by the Free Software > + * Foundation, either version 2 of that License or (at your option) any > + * later version. What is that // doing there? > +/* Exceptions bit map */ > +#define EX_DMA_BUS_ERROR 0x8000 > +#define EX_DMA_READ_ECC 0x4000 > +#define EX_DMA_SYSTEM_WRITE_ECC 0x2000 > +#define EX_DMA_FM_WRITE_ECC 0x1000 > +#define EX_FPM_STALL_ON_TASKS0x0800 > +#define EX_FPM_SINGLE_ECC0x0400 > +#define EX_FPM_DOUBLE_ECC0x0200 > +#define EX_QMI_SINGLE_ECC0x0100 > +#define EX_QMI_DEQ_FROM_UNKNOWN_PORTID 0x0080 > +#define EX_QMI_DOUBLE_ECC0x0040 > +#define EX_BMI_LIST_RAM_ECC 0x0020 > +#define EX_BMI_STORAGE_PROFILE_ECC 0x0010 > +#define EX_BMI_STATISTICS_RAM_ECC0x0008 > +#define EX_IRAM_ECC 0x0004 > +#define EX_MURAM_ECC 0x0002 > +#define EX_BMI_DISPATCH_RAM_ECC 0x0001 > +#define EX_DMA_SINGLE_PORT_ECC 0x8000 > + > +#define DFLT_EXCEPTIONS \ > + ((EX_DMA_BUS_ERROR)| \ > + (EX_DMA_READ_ECC) | \ > + (EX_DMA_SYSTEM_WRITE_ECC) | \ > + (EX_DMA_FM_WRITE_ECC) | \ > + (EX_FPM_STALL_ON_TASKS)| \ > + (EX_FPM_SINGLE_ECC)| \ > + (EX_FPM_DOUBLE_ECC)| \ > + (EX_QMI_DEQ_FROM_UNKNOWN_PORTID) | \ > + (EX_BMI_LIST_RAM_ECC) | \ > + (EX_BMI_STORAGE_PROFILE_ECC) | \ > + (EX_BMI_STATISTICS_RAM_ECC)| \ > + (EX_MURAM_ECC) | \ > + (EX_BMI_DISPATCH_RAM_ECC) | \ > + (EX_QMI_DOUBLE_ECC)| \ > + (EX_QMI_SINGLE_ECC)) You don't need parentheses around each symbol. This is only used in one place -- why put the list here rather than in the place where it's used? > +struct fman_state_struct { > + u8 fm_id; > + u16 fm_clk_freq; > + struct fman_rev_info rev_info; > + bool enabled_time_stamp; > + u8 count1_micro_bit; > + u8 total_num_of_tasks; > + u8 accumulated_num_of_tasks; > + u32 accumulated_fifo_size; > + u8 accumulated_num_of_open_dmas; > + u8 accumulated_num_of_deq_tnums; > + bool low_end_restriction; > + u32 exceptions; > + u32 extra_fifo_pool_size; > + u8 extra_tasks_pool_size; > + u8 extra_open_dmas_pool_size; > + u16 port_mfl[MAX_NUM_OF_MACS]; > + u16 mac_mfl[MAX_NUM_OF_MACS]; > + > + /* SOC specific */ > + u32 fm_iram_size; > + /* DMA */ > + u32 dma_thresh_max_commq; > + u32 dma_thresh_max_buf; > + u32 max_num_of_open_dmas; > + /* QMI */ > + u32 qmi_max_num_of_tnums; > + u32 qmi_def_tnums_thresh; > + /* BMI */ > + u32 bmi_max_num_of_tasks; > + u32 bmi_max_fifo_size; > + /* General */ > + u32 fm_port_num_of_cg; > + u32 num_of_rx_ports; > + u32 total_fifo_size; > + > + u32 qman_channel_base; > + u32 num_of_qman_channels; > + > + struct resource *res; > +}; > + > +struct fman_cfg { > + u8 disp_limit_tsh; > + u8 prs_disp_tsh; > + u8 plcr_disp_tsh; > + u8 kg_disp_tsh; > + u8 bmi_disp_tsh; > + u8 qmi_enq_disp_tsh; > + u8 qmi_deq_disp_tsh; > + u8 fm_ctl1_disp_tsh; > + u8 fm_ctl2_disp_tsh; > + int dma_cache_override; > + enum fman_dma_aid_mode dma_aid_mode; > + bool dma_aid_override; > + u32 dma_axi_dbg_num_of_beats; > + u32 dma_cam_num_of_entries; > + u32 dma_watchdog; > + u8 dma_comm_qtsh_asrt_emer; > + u32 dma_write_buf_tsh_asrt_emer; > + u32 dma_read_buf_tsh_asrt_emer; > + u8 dma_comm_qtsh_
Re: [PATCH v3 31/32] cxlflash: Fix to avoid corrupting adapter fops
> On Sep 25, 2015, at 4:23 PM, Brian King wrote: > On 09/24/2015 02:44 PM, Matthew R. Ochs wrote: >> @@ -1293,8 +1291,8 @@ static int cxlflash_disk_attach(struct scsi_device >> *sdev, >> >> int fd = -1; >> >> -/* On first attach set fileops */ >> -if (atomic_read(&cfg->num_user_contexts) == 0) >> +/* On very first attach set fileops for adapter */ >> +if (cfg->cxl_fops.owner != THIS_MODULE) >> cfg->cxl_fops = cxlflash_cxl_fops; > > Hmm... Why not just set this up once at probe time instead? I tried to explain this in the commit message. I left it out of probe time in an attempt to keep it separate from the legacy support as it's only needed for superpipe. I'll fix up this commit and just move it to probe as that seems easiest. -matt ___ Linuxppc-dev mailing list Linuxppc-dev@lists.ozlabs.org https://lists.ozlabs.org/listinfo/linuxppc-dev
Re: [PATCH v3 28/32] cxlflash: Fix to avoid state change collision
> On Sep 25, 2015, at 4:10 PM, Brian King wrote: > On 09/24/2015 02:42 PM, Matthew R. Ochs wrote: >> diff --git a/drivers/scsi/cxlflash/main.c b/drivers/scsi/cxlflash/main.c >> index ab11ee6..325ba31 100644 >> --- a/drivers/scsi/cxlflash/main.c >> +++ b/drivers/scsi/cxlflash/main.c >> @@ -496,6 +496,7 @@ static int cxlflash_queuecommand(struct Scsi_Host *host, >> struct scsi_cmnd *scp) >> struct cxlflash_cfg *cfg = (struct cxlflash_cfg *)host->hostdata; >> struct afu *afu = cfg->afu; >> struct device *dev = &cfg->dev->dev; >> +enum cxlflash_state state; >> struct afu_cmd *cmd; >> u32 port_sel = scp->device->channel + 1; >> int nseg, i, ncount; >> @@ -525,7 +526,11 @@ static int cxlflash_queuecommand(struct Scsi_Host >> *host, struct scsi_cmnd *scp) >> } >> spin_unlock_irqrestore(&cfg->tmf_slock, lock_flags); >> >> -switch (cfg->state) { >> +mutex_lock(&cfg->mutex); >> +state = cfg->state; >> +mutex_unlock(&cfg->mutex); > > You can't grab a mutex in queuecommand, since it can sleep and queuecommand > can be called from soft irq context. > > Also, I'm not sure what to think about this patch in general. Obviously state > can change immediately after > you drop the mutex, and according to Documentation/memory-barriers.txt, > memory operations after the unlock can > occur before the unlock occurs. Is this a problem? You bring up some good points. I'm going to remove this patch from the set and rework it. -matt ___ Linuxppc-dev mailing list Linuxppc-dev@lists.ozlabs.org https://lists.ozlabs.org/listinfo/linuxppc-dev
Re: [RFC v2 3/7] powerpc: atomic: Implement atomic{,64}_{add,sub}_return_* variants
On Wed, Sep 23, 2015 at 08:07:55AM +0800, Boqun Feng wrote: > On Tue, Sep 22, 2015 at 08:25:40AM -0700, Paul E. McKenney wrote: > > On Tue, Sep 22, 2015 at 07:37:04AM +0800, Boqun Feng wrote: > > > On Tue, Sep 22, 2015 at 07:26:56AM +0800, Boqun Feng wrote: > > > > On Mon, Sep 21, 2015 at 11:24:27PM +0100, Will Deacon wrote: > > > > > Hi Boqun, > > > > > > > > > > On Sun, Sep 20, 2015 at 09:23:03AM +0100, Boqun Feng wrote: > > > > > > On Sat, Sep 19, 2015 at 11:33:10PM +0800, Boqun Feng wrote: > > > > > > > On Fri, Sep 18, 2015 at 05:59:02PM +0100, Will Deacon wrote: > > > > > > > > On Wed, Sep 16, 2015 at 04:49:31PM +0100, Boqun Feng wrote: > > > > > > > > > On powerpc, we don't need a general memory barrier to achieve > > > > > > > > > acquire and > > > > > > > > > release semantics, so __atomic_op_{acquire,release} can be > > > > > > > > > implemented > > > > > > > > > using "lwsync" and "isync". > > > > > > > > > > > > > > > > I'm assuming isync+ctrl isn't transitive, so we need to get to > > > > > > > > the bottom > > > > > > > > > > > > > > Actually the transitivity is still guaranteed here, I think ;-) > > > > > > > > > > The litmus test I'm thinking of is: > > > > > > > > > > > > > > > { > > > > > 0:r2=x; > > > > > 1:r2=x; 1:r5=z; > > > > > 2:r2=z; 2:r4=x; > > > > > } > > > > > P0 | P1| P2 ; > > > > > li r1,1 | lwz r1,0(r2) | lwz r1,0(r2) ; > > > > > stw r1,0(r2) | cmpw r1,r1| cmpw r1,r1 ; > > > > > | beq LC00 | beq LC01; > > > > > | LC00: | LC01:; > > > > > | isync | isync; > > > > > | li r4,1 | lwz r3,0(r4) ; > > > > > | stw r4,0(r5) | ; > > > > > exists > > > > > (1:r1=1 /\ 2:r1=1 /\ 2:r3=0) > > > > > > > > > > > > > > > Which appears to be allowed. I don't think you need to worry about > > > > > backwards > > > > > branches for the ctrl+isync construction (none of the current example > > > > > do, > > > > > afaict). > > > > > > > > > > > > > Yes.. my care of backwards branches is not quite related to the topic, I > > > > concerned that mostly because my test is using atomic operation, and I > > > > just want to test the exact asm code. > > > > > > > > > Anyway, all the problematic cases seem to arise when we start mixing > > > > > ACQUIRE/RELEASE accesses with relaxed accesses (i.e. where an access > > > > > from > > > > > one group reads from an access in the other group). It would be > > > > > simplest > > > > > to say that this doesn't provide any transitivity guarantees, and that > > > > > an ACQUIRE must always read from a RELEASE if transitivity is > > > > > required. > > > > > > > > > > > > > Agreed. RELEASE alone doesn't provide transitivity and transitivity is > > > ^^^ > > > This should be ACQUIRE... > > > > > > > guaranteed only if an ACQUIRE read from a RELEASE. That's exactly the > > > > direction which the link (https://lkml.org/lkml/2015/9/15/836) is > > > > heading to. So I think we are fine here to use ctrl+isync here, right? > > > > We are going to have to err on the side of strictness, that is, having > > the documentation place more requirements on the developer than the union > > of the hardware does. Besides, I haven't heard any recent complaints > > that memory-barriers.txt is too simple. ;-) > > Agreed ;-) > > For atomic operations, using isync in ACQUIRE operations does gaurantee > that a pure RELEASE/ACQUIRE chain provides transitivity. So, again, I > think we are fine here to use isync in ACQUIRE atomic operations, > unless you think we need to be more strict, i.e, making ACQUIRE itself > provide transitivy? As I understand it, either isync or lwsync suffices, with the choice depending on the hardware. The kernel will rewrite itself at boot time if you use the appropriate macro. ;-) Thanx, Paul ___ Linuxppc-dev mailing list Linuxppc-dev@lists.ozlabs.org https://lists.ozlabs.org/listinfo/linuxppc-dev
Re: [PATCH v3 32/32] cxlflash: Correct trace string
Reviewed-by: Brian King -- Brian King Power Linux I/O IBM Linux Technology Center ___ Linuxppc-dev mailing list Linuxppc-dev@lists.ozlabs.org https://lists.ozlabs.org/listinfo/linuxppc-dev
Re: [PATCH v3 31/32] cxlflash: Fix to avoid corrupting adapter fops
On 09/24/2015 02:44 PM, Matthew R. Ochs wrote: > @@ -1293,8 +1291,8 @@ static int cxlflash_disk_attach(struct scsi_device > *sdev, > > int fd = -1; > > - /* On first attach set fileops */ > - if (atomic_read(&cfg->num_user_contexts) == 0) > + /* On very first attach set fileops for adapter */ > + if (cfg->cxl_fops.owner != THIS_MODULE) > cfg->cxl_fops = cxlflash_cxl_fops; Hmm... Why not just set this up once at probe time instead? > > if (attach->num_interrupts > 4) { > -- Brian King Power Linux I/O IBM Linux Technology Center ___ Linuxppc-dev mailing list Linuxppc-dev@lists.ozlabs.org https://lists.ozlabs.org/listinfo/linuxppc-dev
Re: [PATCH v3 5/5] cpufreq: qoriq: Don't look at clock implementation details
On Friday, September 25, 2015 04:17:07 PM Scott Wood wrote: > On Fri, 2015-09-25 at 23:42 +0200, Rafael J. Wysocki wrote: > > On Tuesday, September 22, 2015 12:46:54 PM Viresh Kumar wrote: > > > On 19-09-15, 23:29, Scott Wood wrote: > > > > Get the CPU clock's potential parent clocks from the clock interface > > > > itself, rather than manually parsing the clocks property to find a > > > > phandle, looking at the clock-names property of that, and assuming that > > > > those are valid parent clocks for the cpu clock. > > > > > > > > This is necessary now that the clocks are generated based on the clock > > > > driver's knowledge of the chip rather than a fragile device-tree > > > > description of the mux options. > > > > > > > > We can now rely on the clock driver to ensure that the mux only exposes > > > > options that are valid. The cpufreq driver was currently being overly > > > > conservative in some cases -- for example, the "min_cpufreq = > > > > get_bus_freq()" restriction only applies to chips with erratum > > > > A-004510, and whether the freq_mask used on p5020 is needed depends on > > > > the actual frequencies of the PLLs (FWIW, p5040 has a similar > > > > limitation but its .freq_mask was zero) -- and the frequency mask > > > > mechanism made assumptions about particular parent clock indices that > > > > are no longer valid. > > > > > > > > Signed-off-by: Scott Wood > > > > --- > > > > v3: was patch 1/5 and patch 4/5, plus blacklist e6500 and changes > > > > to clk api usage > > > > > > > > drivers/cpufreq/qoriq-cpufreq.c | 137 - > > > > --- > > > > 1 file changed, 40 insertions(+), 97 deletions(-) > > > > > > Acked-by: Viresh Kumar > > > > I'm wondering who's supposed to be merging this set? > > As I noted in the cover letter, I'm looking for acks so that I can apply > these to a topic branch which can be pulled through the PPC and ARM trees, > each of which will have patches that depend on it. OK, so no objections from the cpufreq side and you have the ACK from Viresh. Thanks, Rafael ___ Linuxppc-dev mailing list Linuxppc-dev@lists.ozlabs.org https://lists.ozlabs.org/listinfo/linuxppc-dev
Re: [PATCH v3 5/5] cpufreq: qoriq: Don't look at clock implementation details
On Tuesday, September 22, 2015 12:46:54 PM Viresh Kumar wrote: > On 19-09-15, 23:29, Scott Wood wrote: > > Get the CPU clock's potential parent clocks from the clock interface > > itself, rather than manually parsing the clocks property to find a > > phandle, looking at the clock-names property of that, and assuming that > > those are valid parent clocks for the cpu clock. > > > > This is necessary now that the clocks are generated based on the clock > > driver's knowledge of the chip rather than a fragile device-tree > > description of the mux options. > > > > We can now rely on the clock driver to ensure that the mux only exposes > > options that are valid. The cpufreq driver was currently being overly > > conservative in some cases -- for example, the "min_cpufreq = > > get_bus_freq()" restriction only applies to chips with erratum > > A-004510, and whether the freq_mask used on p5020 is needed depends on > > the actual frequencies of the PLLs (FWIW, p5040 has a similar > > limitation but its .freq_mask was zero) -- and the frequency mask > > mechanism made assumptions about particular parent clock indices that > > are no longer valid. > > > > Signed-off-by: Scott Wood > > --- > > v3: was patch 1/5 and patch 4/5, plus blacklist e6500 and changes > > to clk api usage > > > > drivers/cpufreq/qoriq-cpufreq.c | 137 > > > > 1 file changed, 40 insertions(+), 97 deletions(-) > > Acked-by: Viresh Kumar I'm wondering who's supposed to be merging this set? Rafael ___ Linuxppc-dev mailing list Linuxppc-dev@lists.ozlabs.org https://lists.ozlabs.org/listinfo/linuxppc-dev
Re: [PATCH v3 5/5] cpufreq: qoriq: Don't look at clock implementation details
On Fri, 2015-09-25 at 23:42 +0200, Rafael J. Wysocki wrote: > On Tuesday, September 22, 2015 12:46:54 PM Viresh Kumar wrote: > > On 19-09-15, 23:29, Scott Wood wrote: > > > Get the CPU clock's potential parent clocks from the clock interface > > > itself, rather than manually parsing the clocks property to find a > > > phandle, looking at the clock-names property of that, and assuming that > > > those are valid parent clocks for the cpu clock. > > > > > > This is necessary now that the clocks are generated based on the clock > > > driver's knowledge of the chip rather than a fragile device-tree > > > description of the mux options. > > > > > > We can now rely on the clock driver to ensure that the mux only exposes > > > options that are valid. The cpufreq driver was currently being overly > > > conservative in some cases -- for example, the "min_cpufreq = > > > get_bus_freq()" restriction only applies to chips with erratum > > > A-004510, and whether the freq_mask used on p5020 is needed depends on > > > the actual frequencies of the PLLs (FWIW, p5040 has a similar > > > limitation but its .freq_mask was zero) -- and the frequency mask > > > mechanism made assumptions about particular parent clock indices that > > > are no longer valid. > > > > > > Signed-off-by: Scott Wood > > > --- > > > v3: was patch 1/5 and patch 4/5, plus blacklist e6500 and changes > > > to clk api usage > > > > > > drivers/cpufreq/qoriq-cpufreq.c | 137 - > > > --- > > > 1 file changed, 40 insertions(+), 97 deletions(-) > > > > Acked-by: Viresh Kumar > > I'm wondering who's supposed to be merging this set? As I noted in the cover letter, I'm looking for acks so that I can apply these to a topic branch which can be pulled through the PPC and ARM trees, each of which will have patches that depend on it. -Scott ___ Linuxppc-dev mailing list Linuxppc-dev@lists.ozlabs.org https://lists.ozlabs.org/listinfo/linuxppc-dev
Re: [PATCH v3 30/32] cxlflash: Fix to double the delay each time
Reviewed-by: Brian King -- Brian King Power Linux I/O IBM Linux Technology Center ___ Linuxppc-dev mailing list Linuxppc-dev@lists.ozlabs.org https://lists.ozlabs.org/listinfo/linuxppc-dev
Re: [PATCH v3 28/32] cxlflash: Fix to avoid state change collision
On 09/24/2015 02:42 PM, Matthew R. Ochs wrote: > diff --git a/drivers/scsi/cxlflash/main.c b/drivers/scsi/cxlflash/main.c > index ab11ee6..325ba31 100644 > --- a/drivers/scsi/cxlflash/main.c > +++ b/drivers/scsi/cxlflash/main.c > @@ -496,6 +496,7 @@ static int cxlflash_queuecommand(struct Scsi_Host *host, > struct scsi_cmnd *scp) > struct cxlflash_cfg *cfg = (struct cxlflash_cfg *)host->hostdata; > struct afu *afu = cfg->afu; > struct device *dev = &cfg->dev->dev; > + enum cxlflash_state state; > struct afu_cmd *cmd; > u32 port_sel = scp->device->channel + 1; > int nseg, i, ncount; > @@ -525,7 +526,11 @@ static int cxlflash_queuecommand(struct Scsi_Host *host, > struct scsi_cmnd *scp) > } > spin_unlock_irqrestore(&cfg->tmf_slock, lock_flags); > > - switch (cfg->state) { > + mutex_lock(&cfg->mutex); > + state = cfg->state; > + mutex_unlock(&cfg->mutex); You can't grab a mutex in queuecommand, since it can sleep and queuecommand can be called from soft irq context. Also, I'm not sure what to think about this patch in general. Obviously state can change immediately after you drop the mutex, and according to Documentation/memory-barriers.txt, memory operations after the unlock can occur before the unlock occurs. Is this a problem? > + > + switch (state) { > case STATE_RESET: > dev_dbg_ratelimited(dev, "%s: device is in reset!\n", __func__); > rc = SCSI_MLQUEUE_HOST_BUSY; -- Brian King Power Linux I/O IBM Linux Technology Center ___ Linuxppc-dev mailing list Linuxppc-dev@lists.ozlabs.org https://lists.ozlabs.org/listinfo/linuxppc-dev
Re: [RFC PATCH 1/2] scripts/kconfig/Makefile: Allow KBUILD_DEFCONFIG to be a target
Dne 24.9.2015 v 00:16 Michael Ellerman napsal(a): > > > On 23 September 2015 19:50:52 GMT+10:00, Michal Marek > wrote: >> On 2015-09-23 07:40, Michael Ellerman wrote: >>> +else ifneq ($(wildcard >> arch/$(SRCARCH)/configs/$(KBUILD_DEFCONFIG)),) >>> @$(kecho) "*** Default configuration is based on >> '$(KBUILD_DEFCONFIG)'" >>> $(Q)$< $(silent) >> --defconfig=arch/$(SRCARCH)/configs/$(KBUILD_DEFCONFIG) $(Kconfig) >>> +else + @$(kecho) "*** Default configuration is based on target >> '$(KBUILD_DEFCONFIG)'" >>> + $(Q)$(MAKE) -f $(srctree)/Makefile $(KBUILD_DEFCONFIG) endif >> >> What is the anticipated usage of this? The patch is not needed to >> make >> >> make ppc64le_defconfig >> >> work with the second patch. If it was, this would create a loop >> anyway. > > The idea is to make 'make defconfig' work when KBUILD_DEFCONFIG is > ppc64le_defconfig (which happens for us when uname returns ppc64le) > and additionally when ppc64le_defconfig is not a real file. Ah, that makes sense. You can add Acked-by: Michal Marek if you want. Michal ___ Linuxppc-dev mailing list Linuxppc-dev@lists.ozlabs.org https://lists.ozlabs.org/listinfo/linuxppc-dev
Re: [PATCH v3 20/32] cxlflash: Fix to prevent workq from accessing freed memory
Reviewed-by: Brian King -- Brian King Power Linux I/O IBM Linux Technology Center ___ Linuxppc-dev mailing list Linuxppc-dev@lists.ozlabs.org https://lists.ozlabs.org/listinfo/linuxppc-dev
Re: [PATCH v3 19/32] cxlflash: Correct usage of scsi_host_put()
Reviewed-by: Brian King -- Brian King Power Linux I/O IBM Linux Technology Center ___ Linuxppc-dev mailing list Linuxppc-dev@lists.ozlabs.org https://lists.ozlabs.org/listinfo/linuxppc-dev
Re: [PATCH v3 13/32] cxlflash: Fix to avoid stall while waiting on TMF
Reviewed-by: Brian King -- Brian King Power Linux I/O IBM Linux Technology Center ___ Linuxppc-dev mailing list Linuxppc-dev@lists.ozlabs.org https://lists.ozlabs.org/listinfo/linuxppc-dev
Re: [v3 2/8] dpaa_eth: add support for DPAA Ethernet
From: Madalin Bucur Date: Thu, 24 Sep 2015 18:00:13 +0300 > +#define DPA_NAPI_WEIGHT 64 This is just the default, so simply use "NAPI_POLL_WEIGHT" rather than defining your own macro unnecessarily for this. > +static int dpa_eth_priv_stop(struct net_device *net_dev) > +{ > + int err; > + struct dpa_priv_s *priv; > + > + err = dpa_stop(net_dev); > + /* Allow NAPI to consume any frame still in the Rx/TxConfirm > + * ingress queues. This is to avoid a race between the current > + * context and ksoftirqd which could leave NAPI disabled while > + * in fact there's still Rx traffic to be processed. > + */ > + usleep_range(5000, 1); A wait never fixes a race. Instead you must use appropriate mutual exclusion mechanisms to ensure that NAPI is really completed it's run, and we definitely have a means by which to do that properly. > + net_dev->hw_features |= (NETIF_F_IP_CSUM | NETIF_F_IPV6_CSUM | > + NETIF_F_LLTX); This second line should start exactly at the first column after the openning parenthesis of the first line. ___ Linuxppc-dev mailing list Linuxppc-dev@lists.ozlabs.org https://lists.ozlabs.org/listinfo/linuxppc-dev
Re: [PATCH v3 08/32] cxlflash: Fix to avoid CXL services during EEH
> On Sep 24, 2015, at 8:23 PM, Brian King wrote: > On 09/24/2015 02:38 PM, Matthew R. Ochs wrote: >> diff --git a/drivers/scsi/cxlflash/main.c b/drivers/scsi/cxlflash/main.c >> index 3e3ccf1..6e85c77 100644 >> --- a/drivers/scsi/cxlflash/main.c >> +++ b/drivers/scsi/cxlflash/main.c >> @@ -2383,16 +2397,14 @@ static pci_ers_result_t >> cxlflash_pci_error_detected(struct pci_dev *pdev, >> switch (state) { >> case pci_channel_io_frozen: >> cfg->state = STATE_LIMBO; >> - >> -/* Turn off legacy I/O */ >> scsi_block_requests(cfg->host); >> +drain_ioctls(cfg); > > I don't see this addressing the deadlock with an outstanding read_cap16 > during EEH I identified > in the previous review. Am I missing something here? I'll submit this in a separate patch in v4. >> /** >> + * check_state() - checks and responds to the current adapter state >> + * @cfg:Internal structure associated with the host. >> + * @ioctl: Indicates if on an ioctl thread. >> + * >> + * This routine can block and should only be used on process context. >> + * When blocking on an ioctl thread, the ioctl read semaphore should be >> + * let up to allow for draining actively running ioctls. Also note that >> + * when waking up from waiting in reset, the state is unknown and must >> + * be checked again before proceeding. >> + * >> + * Return: 0 on success, -errno on failure >> + */ >> +static int check_state(struct cxlflash_cfg *cfg, bool ioctl) > > Looks like you missed this cleanup. The second parameter is not needed, since > all > the callers set it to true. Yep, this was my mistake. I'll correct in a v4 as the change will ripple to the next patch in the series. ___ Linuxppc-dev mailing list Linuxppc-dev@lists.ozlabs.org https://lists.ozlabs.org/listinfo/linuxppc-dev
Re: [PATCH v2 08/25] powerpc/8xx: Map IMMR area with 512k page at a fixed address
On Fri, 2015-09-25 at 14:46 +, David Laight wrote: > From: Scott Wood > > Sent: 24 September 2015 21:14 > > > Isn't this a more general problem? > > > > > > If there are multiple remap requests for the same physical page > > > shouldn't the kernel be just increasing a reference count somewhere > > > and returning address in the same virtual page? > > > This should probably happen regardless of the address. > > > I presume it must be done for cacheable mappings. > > > > Why would you assume that? > > Because 'really horrid (tm)' things happen on some cache > architectures if you map the same physical address to > multiple virtual addresses. PPC is not such an architecture. -Scott ___ Linuxppc-dev mailing list Linuxppc-dev@lists.ozlabs.org https://lists.ozlabs.org/listinfo/linuxppc-dev
RE: [PATCH v2 08/25] powerpc/8xx: Map IMMR area with 512k page at a fixed address
From: Scott Wood > Sent: 24 September 2015 21:14 > > Isn't this a more general problem? > > > > If there are multiple remap requests for the same physical page > > shouldn't the kernel be just increasing a reference count somewhere > > and returning address in the same virtual page? > > This should probably happen regardless of the address. > > I presume it must be done for cacheable mappings. > > Why would you assume that? Because 'really horrid (tm)' things happen on some cache architectures if you map the same physical address to multiple virtual addresses. David ___ Linuxppc-dev mailing list Linuxppc-dev@lists.ozlabs.org https://lists.ozlabs.org/listinfo/linuxppc-dev
Re: [PATCH v2 1/2] powerpc/selftest: Add gettimeofday() benchmark
On 9/25/15, Gabriel Paubert wrote: > On Fri, Sep 25, 2015 at 12:28:30PM +0300, Denis Kirjanov wrote: >> On 9/25/15, Arnd Bergmann wrote: >> > On Friday 25 September 2015 14:01:39 Michael Neuling wrote: >> >> This adds a benchmark directory to the powerpc selftests and adds a >> >> gettimeofday() benchmark to it. >> >> >> >> Suggested-by: Michael Ellerman >> >> Signed-off-by: Michael Neuling >> >> >> > >> > Any reason for keeping this powerpc specific? It seems generally >> > useful. >> > and portable. >> You're right. Moreover, we can put some comment to the benchmark why >> we've made such decision to add it (reference to the commit >> "powerpc/vdso: Avoid link stack corruption in __get_datapage()") > > Why gettimeofday? Isn't clock_gettime the modern variant? Mostly, I think, for historical reasons. Both of them are implemented as vsyscalls (on ppc, ) so why bother? BTW: Found an interesting vdso testsuite on github: https://github.com/nlynch-mentor/vdsotest under GPL. Not sure if we really need such similar thing in the kernel selftest sources. > > BTW: dows anyone receive 2 copies of every messge in this thread ? > > I do, and I suspect that this is due to the Cc: list having both > linuxppc-...@ozlabs.org and linuxppc-dev@lists.ozlabs.org. I removed the > former for this reply. > > Gabriel > ___ Linuxppc-dev mailing list Linuxppc-dev@lists.ozlabs.org https://lists.ozlabs.org/listinfo/linuxppc-dev
Re: [PATCH v2 1/2] powerpc/selftest: Add gettimeofday() benchmark
On Fri, Sep 25, 2015 at 12:28:30PM +0300, Denis Kirjanov wrote: > On 9/25/15, Arnd Bergmann wrote: > > On Friday 25 September 2015 14:01:39 Michael Neuling wrote: > >> This adds a benchmark directory to the powerpc selftests and adds a > >> gettimeofday() benchmark to it. > >> > >> Suggested-by: Michael Ellerman > >> Signed-off-by: Michael Neuling > >> > > > > Any reason for keeping this powerpc specific? It seems generally useful. > > and portable. > You're right. Moreover, we can put some comment to the benchmark why > we've made such decision to add it (reference to the commit > "powerpc/vdso: Avoid link stack corruption in __get_datapage()") Why gettimeofday? Isn't clock_gettime the modern variant? BTW: dows anyone receive 2 copies of every messge in this thread ? I do, and I suspect that this is due to the Cc: list having both linuxppc-...@ozlabs.org and linuxppc-dev@lists.ozlabs.org. I removed the former for this reply. Gabriel ___ Linuxppc-dev mailing list Linuxppc-dev@lists.ozlabs.org https://lists.ozlabs.org/listinfo/linuxppc-dev
Re: [PATCH v2 1/2] powerpc/selftest: Add gettimeofday() benchmark
On 9/25/15, Arnd Bergmann wrote: > On Friday 25 September 2015 14:01:39 Michael Neuling wrote: >> This adds a benchmark directory to the powerpc selftests and adds a >> gettimeofday() benchmark to it. >> >> Suggested-by: Michael Ellerman >> Signed-off-by: Michael Neuling >> > > Any reason for keeping this powerpc specific? It seems generally useful. > and portable. You're right. Moreover, we can put some comment to the benchmark why we've made such decision to add it (reference to the commit "powerpc/vdso: Avoid link stack corruption in __get_datapage()") > > Arnd > ___ Linuxppc-dev mailing list Linuxppc-dev@lists.ozlabs.org https://lists.ozlabs.org/listinfo/linuxppc-dev
Re: [PATCH] powerpc/eeh: Avoid to handle EEH on a passed Child PE
On Wed, Sep 23, 2015 at 09:07:41AM +1000, Gavin Shan wrote: >On Tue, Sep 22, 2015 at 12:43:03PM +0800, Wei Yang wrote: >>On Mon, Sep 21, 2015 at 09:49:45PM +1000, Gavin Shan wrote: >>>On Mon, Sep 21, 2015 at 05:29:48PM +0800, Wei Yang wrote: Current EEH infrastructure would avoid to handle EEH when a PE is passed to guest, while if this PE is a Child PE of the one hit EEH, host would handle this. By doing so, this would leads to guest hang. The correct way is avoid to handle it on host and let guest to recover. This patch avoids to handle EEH on a passed Child PE. >>> >>>Ok. It's fixing the problem the guest, which owns a VF, when its PF hitting >>>EEH error, right? If so, I'm not sure if you really tested this code. Does >>>it work for you? >> >>Yes, I inject error on Parent Bus PE. >> >>> >>>When the parent PE (PF) is stopped for EEH recovery, it sounds impossible >>>that the child PE can't be affected and just escape from the error. The >>>question is how the guest can continue to work after the EEH recovery on >>>parent PE? >> >>What I see is the PF is covering and VF in guest is recovering. >> > >What do you mean by "covering"? Which PE's error is detected first in your >testing? There is potentially race here: when the VF PE's error is detected >first and guest tries to recover it. After the recovery happened on guest >side, the host detects the PF PE's error and tries to recover it. During >the recovery, the VF PE is total unusable but guest doesn't know that and >operate like usual. I'm not sure what kinds of problems it can incur, but >it would incur issues. > >On the other hand, if PF PE's error is detected on host first, and then the >guest detects the error on VF PE. After that, the host and guest try to do >recovery at same time. Host issues PE reset to PF PE, which isn't finished >yet. Then guest issues PE reset to VF PE, which will cause another EEH error. > >So I'm not sure if had this patch fully tested. If so, it seems the result is >just achieved by luck, I guess. > It looks really has some race condition. So the expected order should be PF's PE recovered then VF's PE recover in guest, right? >>> Signed-off-by: Wei Yang --- arch/powerpc/kernel/eeh_pe.c | 5 + 1 file changed, 5 insertions(+) diff --git a/arch/powerpc/kernel/eeh_pe.c b/arch/powerpc/kernel/eeh_pe.c index 5cde950..c6d0e9f 100644 --- a/arch/powerpc/kernel/eeh_pe.c +++ b/arch/powerpc/kernel/eeh_pe.c @@ -172,6 +172,7 @@ static struct eeh_pe *eeh_pe_next(struct eeh_pe *pe, * callback returns something other than NULL, or no more PEs * to be traversed. */ +static void *__eeh_pe_get(void *data, void *flag); void *eeh_pe_traverse(struct eeh_pe *root, eeh_traverse_func fn, void *flag) { @@ -179,6 +180,8 @@ void *eeh_pe_traverse(struct eeh_pe *root, void *ret; for (pe = root; pe; pe = eeh_pe_next(pe, root)) { + if (eeh_pe_passed(pe) && (fn != __eeh_pe_get)) + continue; >>> >>>The code change here seems ugly. >>> >>>The "flag" can be extended to carry the information to skip pass-through >>>PEs or not. So the function calling eeh_pe_traverse() decides to skip >>>pass-through PEs or not. >> >>I don't get the point, which "flag" you mean? Add a flag in eeh_pe? >> > void *eeh_pe_traverse(struct eeh_pe *root, eeh_traverse_func fn, void *flag) >^^ >This one > >The code needn't to be changed in a hurry though. I don't think it's right >way to fix the issue as discussed as above. > >>> ret = fn(pe, flag); if (ret) return ret; } @@ -210,6 +213,8 @@ void *eeh_pe_dev_traverse(struct eeh_pe *root, >>> /* Traverse root PE */ for (pe = root; pe; pe = eeh_pe_next(pe, root)) { + if (eeh_pe_passed(pe)) + continue; eeh_pe_for_each_dev(pe, edev, tmp) { ret = fn(edev, flag); if (ret) -- 2.5.0 >> >>-- >>Richard Yang >>Help you, Help me -- Richard Yang Help you, Help me ___ Linuxppc-dev mailing list Linuxppc-dev@lists.ozlabs.org https://lists.ozlabs.org/listinfo/linuxppc-dev
Re: [PATCH v2 1/2] powerpc/selftest: Add gettimeofday() benchmark
On Friday 25 September 2015 14:01:39 Michael Neuling wrote: > This adds a benchmark directory to the powerpc selftests and adds a > gettimeofday() benchmark to it. > > Suggested-by: Michael Ellerman > Signed-off-by: Michael Neuling > Any reason for keeping this powerpc specific? It seems generally useful. and portable. Arnd ___ Linuxppc-dev mailing list Linuxppc-dev@lists.ozlabs.org https://lists.ozlabs.org/listinfo/linuxppc-dev
RE: [PATCH V3] thermal: qoriq: Add thermal management support
> -Original Message- > From: Wood Scott-B07421 > Sent: Friday, September 25, 2015 1:12 PM > To: Jia Hongtao-B38951 > Cc: edubez...@gmail.com; linux...@vger.kernel.org; linuxppc- > d...@lists.ozlabs.org > Subject: Re: [PATCH V3] thermal: qoriq: Add thermal management support > > On Thu, 2015-09-24 at 22:09 -0500, Jia Hongtao-B38951 wrote: > > > -Original Message- > > > From: Wood Scott-B07421 > > > Sent: Friday, September 25, 2015 6:10 AM > > > To: Jia Hongtao-B38951 > > > Cc: edubez...@gmail.com; linux...@vger.kernel.org; linuxppc- > > > d...@lists.ozlabs.org > > > Subject: Re: [PATCH V3] thermal: qoriq: Add thermal management > support > > > > > > On Wed, 2015-09-23 at 16:28 +0800, Jia Hongtao wrote: > > > > This driver add thermal management support by enabling TMU (Thermal > > > > Monitoring Unit) on QorIQ platform. > > > > > > > > It's based on thermal of framework: > > > > - Trip points defined in device tree. > > > > - Cpufreq as cooling device registered in qoriq cpufreq driver. > > > > > > I don't see any cooling device registered in the qoriq cpufreq driver. > > > Is this dependent on some other patch? > > > > It's not depend on any patch. But I saw your patch below: > > [PATCH v3 5/5] cpufreq: qoriq: Don't look at clock implementation > details > > So I hold my patch waiting for your patch merged or there will be > conflict. > > > > I could send it out too if you are fine with it. > > Yes, rebase it on that patch and send it. Mention below the --- that > you're > depending on that patch, and provide a patchwork/archive link. Ok. But I need rebase again if you have further changes on that patch. Anyway, this is not a big problem for me. > > > > > +static int tmu_get_temp(void *p, int *temp) { > > > > + u8 val; > > > > + struct qoriq_tmu_data *data = p; > > > > + > > > > + val = ioread32be(&data->regs->site[0].tritsr); > > > > + *temp = (int)val * 1000; > > > > > > Why don't you declare val as int in the first place? > > > > It's a 32bit register. > > Only the least significant 8 bits represent the temperature. > > The most significant bit shows the validness of the value. > > I use u8 type to remove the rest 24bits influence. > > That's even worse. Use an explicit & operation to extract the field > you're > interested in. Will change for next version. > > > > > + ret = qoriq_tmu_calibration(pdev); /* TMU calibration */ > > > > + if (ret < 0) { > > > > + dev_err(&pdev->dev, "TMU calibration failed.\n"); > > > > + ret = -ENODEV; > > > > + goto err_iomap; > > > > + } > > > > > > That function returns negative when device tree properties are > missing, > > > not when a calibration procedure fails. > > > > What's your suggestion here to return then? > > Remove this message and add a message in qoriq_tmu_calibration for each > error > condition. Also have qoriq_tmu_calibration pass back a proper error code, > not -1. Yes, this is more reasonable. > > > > > +static const struct of_device_id qoriq_tmu_match[] = { > > > > + { .compatible = "fsl,qoriq-tmu", }, > > > > + {}, > > > > +}; > > > > > > Binding? > > > > Not send out yet. > > The binding needs to come before the driver that uses it. Right. I will send out the whole patchset. > > -Scott > ___ Linuxppc-dev mailing list Linuxppc-dev@lists.ozlabs.org https://lists.ozlabs.org/listinfo/linuxppc-dev
[GIT PULL] Please pull powerpc/linux.git powerpc-4.3-3 tag
Hi Linus, Please pull two powerpc fixes for 4.3: The following changes since commit 1f93e4a96c9109378204c147b3eec0d0e8100fde: Linux 4.3-rc2 (2015-09-20 14:32:34 -0700) are available in the git repository at: git://git.kernel.org/pub/scm/linux/kernel/git/powerpc/linux.git tags/powerpc-4.3-3 for you to fetch changes up to d6eb71a6d2eda21c8cd7a4dcd6207a0d94eb6ae7: cxl: Fix lockdep warning while creating afu_err_buff attribute (2015-09-23 20:57:13 +1000) powerpc fixes for 4.3 #2 - Wire up sys_membarrier() - cxl: Fix lockdep warning while creating afu_err_buff from Vaibhav Michael Ellerman (1): powerpc: Wire up sys_membarrier() Vaibhav Jain (1): cxl: Fix lockdep warning while creating afu_err_buff attribute arch/powerpc/include/asm/systbl.h | 1 + arch/powerpc/include/asm/unistd.h | 2 +- arch/powerpc/include/uapi/asm/unistd.h | 1 + drivers/misc/cxl/sysfs.c | 2 ++ 4 files changed, 5 insertions(+), 1 deletion(-) signature.asc Description: This is a digitally signed message part ___ Linuxppc-dev mailing list Linuxppc-dev@lists.ozlabs.org https://lists.ozlabs.org/listinfo/linuxppc-dev