diff mbox series

[v1,2/4] mpi3mr: Support PCIe Error Recovery callback handlers

Message ID 20231206152513.71253-3-ranjan.kumar@broadcom.com (mailing list archive)
State Changes Requested
Headers show
Series mpi3mr: Few Enhancements and minor fix | expand

Commit Message

Ranjan Kumar Dec. 6, 2023, 3:25 p.m. UTC
The driver has been upgraded to include support for the
PCIe error recovery callback handler which is crucial for
the recovery of the controllers. This feature is
necessary for addressing the errors reported by
the PCIe AER (Advanced Error Reporting) mechanism.

Signed-off-by: Sathya Prakash <sathya.prakash@broadcom.com>
Signed-off-by: Ranjan Kumar <ranjan.kumar@broadcom.com>
---
 drivers/scsi/mpi3mr/mpi3mr.h           |  31 +++
 drivers/scsi/mpi3mr/mpi3mr_app.c       |  64 +++--
 drivers/scsi/mpi3mr/mpi3mr_fw.c        |  30 ++-
 drivers/scsi/mpi3mr/mpi3mr_os.c        | 330 +++++++++++++++++++++----
 drivers/scsi/mpi3mr/mpi3mr_transport.c |  39 ++-
 5 files changed, 408 insertions(+), 86 deletions(-)

Comments

Bjorn Helgaas Dec. 6, 2023, 4:56 p.m. UTC | #1
On Wed, Dec 06, 2023 at 08:55:11PM +0530, Ranjan Kumar wrote:
> The driver has been upgraded to include support for the
> PCIe error recovery callback handler which is crucial for
> the recovery of the controllers. This feature is
> necessary for addressing the errors reported by
> the PCIe AER (Advanced Error Reporting) mechanism.
> 
> Signed-off-by: Sathya Prakash <sathya.prakash@broadcom.com>
> Signed-off-by: Ranjan Kumar <ranjan.kumar@broadcom.com>
> ...

> +static int
> +mpi3mr_get_shost_and_mrioc(struct pci_dev *pdev,
> +	struct Scsi_Host **shost, struct mpi3mr_ioc **mrioc)
> +{
> +	*shost = pci_get_drvdata(pdev);
> +	if (*shost == NULL) {
> +		dev_err(&pdev->dev, "pdev's driver data is null\n");
> +		return -1;
> +	}
> +
> +	*mrioc = shost_priv(*shost);
> +	if (*mrioc == NULL) {
> +		dev_err(&pdev->dev, "shost's private data is null\n");
> +		*shost = NULL;
> +		return -1;

I'm a little bit skeptical about these checks for NULL, although I do
see that the existing code has similar "if (!shost)" checks.

Usually these checks will only find memory corruption or logic errors,
and silently bailing out, as the previous "if (!shost)" checks do,
just masks a serious problem.  Logging errors, as you do here, is a
little better, but I think it's better to just take the exception when
we dereference the NULL pointer later because that's impossible to
ignore and usually gives more clues about what went wrong.

> +}
> +	return 0;
> +}

The addition and use of mpi3mr_get_shost_and_mrioc() looks like it
could be a separate patch.  If so, it might be nice to split this into
several smaller, simpler patches.

>  static int __maybe_unused
>  mpi3mr_suspend(struct device *dev)
>  {
>  	struct pci_dev *pdev = to_pci_dev(dev);
> -	struct Scsi_Host *shost = pci_get_drvdata(pdev);
> +	struct Scsi_Host *shost;
>  	struct mpi3mr_ioc *mrioc;
>  
> -	if (!shost)
> -		return 0;
> +	if (mpi3mr_get_shost_and_mrioc(pdev, &shost, &mrioc))
> +		return -1;

Is -1 really the best return value here?  It seems like usually a
negative errno is returned.

Bjorn
kernel test robot Dec. 7, 2023, 3:07 p.m. UTC | #2
Hi Ranjan,

kernel test robot noticed the following build warnings:

[auto build test WARNING on mkp-scsi/for-next]
[also build test WARNING on next-20231207]
[cannot apply to jejb-scsi/for-next linus/master v6.7-rc4]
[If your patch is applied to the wrong git tree, kindly drop us a note.
And when submitting patch, we suggest to use '--base' as documented in
https://git-scm.com/docs/git-format-patch#_base_tree_information]

url:    https://github.com/intel-lab-lkp/linux/commits/Ranjan-Kumar/mpi3mr-Improve-Shutdown-times-when-firmware-has-faulted/20231207-004256
base:   https://git.kernel.org/pub/scm/linux/kernel/git/mkp/scsi.git for-next
patch link:    https://lore.kernel.org/r/20231206152513.71253-3-ranjan.kumar%40broadcom.com
patch subject: [PATCH v1 2/4] mpi3mr: Support PCIe Error Recovery callback handlers
config: arc-randconfig-001-20231207 (https://download.01.org/0day-ci/archive/20231207/202312072213.uopHjdQV-lkp@intel.com/config)
compiler: arc-elf-gcc (GCC) 13.2.0
reproduce (this is a W=1 build): (https://download.01.org/0day-ci/archive/20231207/202312072213.uopHjdQV-lkp@intel.com/reproduce)

If you fix the issue in a separate patch/commit (i.e. not just a new version of
the same patch/commit), kindly add following tags
| Reported-by: kernel test robot <lkp@intel.com>
| Closes: https://lore.kernel.org/oe-kbuild-all/202312072213.uopHjdQV-lkp@intel.com/

All warnings (new ones prefixed by >>):

>> drivers/scsi/mpi3mr/mpi3mr_os.c:5272: warning: Function parameter or member 'mrioc' not described in 'mpi3mr_get_shost_and_mrioc'
>> drivers/scsi/mpi3mr/mpi3mr_os.c:5272: warning: Excess function parameter 'ioc' description in 'mpi3mr_get_shost_and_mrioc'


vim +5272 drivers/scsi/mpi3mr/mpi3mr_os.c

  5258	
  5259	/**
  5260	 * mpi3mr_get_shost_and_mrioc - get shost and ioc reference if
  5261	 *			they are valid
  5262	 * @pdev: PCI device struct
  5263	 * @shost: address to store scsi host reference
  5264	 * @ioc: address store HBA adapter reference
  5265	 *
  5266	 * Return: 0 if *shost and *ioc are not NULL otherwise -1.
  5267	 */
  5268	
  5269	static int
  5270	mpi3mr_get_shost_and_mrioc(struct pci_dev *pdev,
  5271		struct Scsi_Host **shost, struct mpi3mr_ioc **mrioc)
> 5272	{
  5273		*shost = pci_get_drvdata(pdev);
  5274		if (*shost == NULL) {
  5275			dev_err(&pdev->dev, "pdev's driver data is null\n");
  5276			return -1;
  5277		}
  5278	
  5279		*mrioc = shost_priv(*shost);
  5280		if (*mrioc == NULL) {
  5281			dev_err(&pdev->dev, "shost's private data is null\n");
  5282			*shost = NULL;
  5283			return -1;
  5284	}
  5285		return 0;
  5286	}
  5287
Sathya Prakash Veerichetty Dec. 7, 2023, 4:55 p.m. UTC | #3
On Wed, Dec 6, 2023 at 9:56 AM Bjorn Helgaas <helgaas@kernel.org> wrote:
>
> On Wed, Dec 06, 2023 at 08:55:11PM +0530, Ranjan Kumar wrote:
> > The driver has been upgraded to include support for the
> > PCIe error recovery callback handler which is crucial for
> > the recovery of the controllers. This feature is
> > necessary for addressing the errors reported by
> > the PCIe AER (Advanced Error Reporting) mechanism.
> >
> > Signed-off-by: Sathya Prakash <sathya.prakash@broadcom.com>
> > Signed-off-by: Ranjan Kumar <ranjan.kumar@broadcom.com>
> > ...
>
> > +static int
> > +mpi3mr_get_shost_and_mrioc(struct pci_dev *pdev,
> > +     struct Scsi_Host **shost, struct mpi3mr_ioc **mrioc)
> > +{
> > +     *shost = pci_get_drvdata(pdev);
> > +     if (*shost == NULL) {
> > +             dev_err(&pdev->dev, "pdev's driver data is null\n");
> > +             return -1;
> > +     }
> > +
> > +     *mrioc = shost_priv(*shost);
> > +     if (*mrioc == NULL) {
> > +             dev_err(&pdev->dev, "shost's private data is null\n");
> > +             *shost = NULL;
> > +             return -1;
>
> I'm a little bit skeptical about these checks for NULL, although I do
> see that the existing code has similar "if (!shost)" checks.
>
> Usually these checks will only find memory corruption or logic errors,
> and silently bailing out, as the previous "if (!shost)" checks do,
> just masks a serious problem.  Logging errors, as you do here, is a
> little better, but I think it's better to just take the exception when
> we dereference the NULL pointer later because that's impossible to
> ignore and usually gives more clues about what went wrong.
>
Agree, however, I wouldn't want to crash the system from this driver.
Will add WARN notices and move this into a separate patch later. For
now, will check this inside the PM functions as we do currently for
other cases.
> > +}
> > +     return 0;
> > +}
>
> The addition and use of mpi3mr_get_shost_and_mrioc() looks like it
> could be a separate patch.  If so, it might be nice to split this into
> several smaller, simpler patches
Will detach this from this patch and will submit another separate patch.

>
> >  static int __maybe_unused
> >  mpi3mr_suspend(struct device *dev)
> >  {
> >       struct pci_dev *pdev = to_pci_dev(dev);
> > -     struct Scsi_Host *shost = pci_get_drvdata(pdev);
> > +     struct Scsi_Host *shost;
> >       struct mpi3mr_ioc *mrioc;
> >
> > -     if (!shost)
> > -             return 0;
> > +     if (mpi3mr_get_shost_and_mrioc(pdev, &shost, &mrioc))
> > +             return -1;
>
> Is -1 really the best return value here?  It seems like usually a
> negative errno is returned.
The __suspend_report_result just prints the return value, so i thought
-1 is fine as we already print a error message. Do you recommend to
use another negative (non -1) number for differentiation between
generic error and this?


>
> Bjorn
>
Bjorn Helgaas Dec. 7, 2023, 8:24 p.m. UTC | #4
On Thu, Dec 07, 2023 at 09:55:09AM -0700, Sathya Prakash Veerichetty wrote:
> On Wed, Dec 6, 2023 at 9:56 AM Bjorn Helgaas <helgaas@kernel.org> wrote:
> > On Wed, Dec 06, 2023 at 08:55:11PM +0530, Ranjan Kumar wrote:
> ...

> > >  static int __maybe_unused
> > >  mpi3mr_suspend(struct device *dev)
> > >  {
> > >       struct pci_dev *pdev = to_pci_dev(dev);
> > > -     struct Scsi_Host *shost = pci_get_drvdata(pdev);
> > > +     struct Scsi_Host *shost;
> > >       struct mpi3mr_ioc *mrioc;
> > >
> > > -     if (!shost)
> > > -             return 0;
> > > +     if (mpi3mr_get_shost_and_mrioc(pdev, &shost, &mrioc))
> > > +             return -1;
> >
> > Is -1 really the best return value here?  It seems like usually a
> > negative errno is returned.
>
> The __suspend_report_result just prints the return value, so i thought
> -1 is fine as we already print a error message. Do you recommend to
> use another negative (non -1) number for differentiation between
> generic error and this?

In addition to being printed by __suspend_report_result() (if the
return value is non-zero), the return value is returned up the chain
by legacy_suspend(), pci_pm_suspend(), etc.

I would probably make mpi3mr_get_shost_and_mrioc() return an errno
like -ENODEV and then have mpi3mr_suspend() capture that and return
it.

These function pointers go in struct dev_pm_ops, and these functions
are documented as returning "error codes", which I interpret as errno
values:
https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/tree/include/linux/pm.h?id=v6.6#n258

Bjorn
diff mbox series

Patch

diff --git a/drivers/scsi/mpi3mr/mpi3mr.h b/drivers/scsi/mpi3mr/mpi3mr.h
index e44e262748ea..de953eb055d0 100644
--- a/drivers/scsi/mpi3mr/mpi3mr.h
+++ b/drivers/scsi/mpi3mr/mpi3mr.h
@@ -23,6 +23,7 @@ 
 #include <linux/miscdevice.h>
 #include <linux/module.h>
 #include <linux/pci.h>
+#include <linux/aer.h>
 #include <linux/poll.h>
 #include <linux/sched.h>
 #include <linux/slab.h>
@@ -481,6 +482,7 @@  struct mpi3mr_throttle_group_info {
 
 /* HBA port flags */
 #define MPI3MR_HBA_PORT_FLAG_DIRTY	0x01
+#define MPI3MR_HBA_PORT_FLAG_NEW       0x02
 
 /* IOCTL data transfer sge*/
 #define MPI3MR_NUM_IOCTL_SGE		256
@@ -900,6 +902,29 @@  struct scmd_priv {
 	u8 mpi3mr_scsiio_req[MPI3MR_ADMIN_REQ_FRAME_SZ];
 };
 
+/**
+ * struct mpi3mr_pdevinfo - PCI device information
+ *
+ * @dev_id: PCI device ID of the adapter
+ * @dev_hw_rev: PCI revision of the adapter
+ * @subsys_dev_id: PCI subsystem device ID of the adapter
+ * @subsys_ven_id: PCI subsystem vendor ID of the adapter
+ * @dev: PCI device
+ * @func: PCI function
+ * @bus: PCI bus
+ * @seg_id: PCI segment ID
+ */
+struct mpi3mr_pdevinfo {
+	u16 id;
+	u16 ssid;
+	u16 ssvid;
+	u16 segment;
+	u8 dev:5;
+	u8 func:3;
+	u8 bus;
+	u8 revision;
+};
+
 /**
  * struct mpi3mr_ioc - Adapter anchor structure stored in shost
  * private data
@@ -1056,6 +1081,9 @@  struct scmd_priv {
  * @ioctl_chain_sge: DMA buffer descriptor for IOCTL chain
  * @ioctl_resp_sge: DMA buffer descriptor for Mgmt cmd response
  * @ioctl_sges_allocated: Flag for IOCTL SGEs allocated or not
+ * @pcie_err_recovery: PCIe error recovery in progress
+ * @block_on_pcie_err: Block IO during PCI error recovery
+ * @pdevinfo: PCI device information
  */
 struct mpi3mr_ioc {
 	struct list_head list;
@@ -1247,6 +1275,9 @@  struct mpi3mr_ioc {
 	struct dma_memory_desc ioctl_chain_sge;
 	struct dma_memory_desc ioctl_resp_sge;
 	bool ioctl_sges_allocated;
+	bool pcie_err_recovery;
+	bool block_on_pcie_err;
+	struct mpi3mr_pdevinfo pdevinfo;
 };
 
 /**
diff --git a/drivers/scsi/mpi3mr/mpi3mr_app.c b/drivers/scsi/mpi3mr/mpi3mr_app.c
index 4b93b7440da6..a11d6f026f0e 100644
--- a/drivers/scsi/mpi3mr/mpi3mr_app.c
+++ b/drivers/scsi/mpi3mr/mpi3mr_app.c
@@ -31,7 +31,7 @@  static int mpi3mr_bsg_pel_abort(struct mpi3mr_ioc *mrioc)
 		dprint_bsg_err(mrioc, "%s: reset in progress\n", __func__);
 		return -1;
 	}
-	if (mrioc->stop_bsgs) {
+	if (mrioc->stop_bsgs || mrioc->block_on_pcie_err) {
 		dprint_bsg_err(mrioc, "%s: bsgs are blocked\n", __func__);
 		return -1;
 	}
@@ -424,6 +424,9 @@  static long mpi3mr_bsg_adp_reset(struct mpi3mr_ioc *mrioc,
 		goto out;
 	}
 
+	if (mrioc->unrecoverable || mrioc->block_on_pcie_err)
+		return -EINVAL;
+
 	sg_copy_to_buffer(job->request_payload.sg_list,
 			  job->request_payload.sg_cnt,
 			  &adpreset, sizeof(adpreset));
@@ -470,25 +473,29 @@  static long mpi3mr_bsg_populate_adpinfo(struct mpi3mr_ioc *mrioc,
 
 	memset(&adpinfo, 0, sizeof(adpinfo));
 	adpinfo.adp_type = MPI3MR_BSG_ADPTYPE_AVGFAMILY;
-	adpinfo.pci_dev_id = mrioc->pdev->device;
-	adpinfo.pci_dev_hw_rev = mrioc->pdev->revision;
-	adpinfo.pci_subsys_dev_id = mrioc->pdev->subsystem_device;
-	adpinfo.pci_subsys_ven_id = mrioc->pdev->subsystem_vendor;
-	adpinfo.pci_bus = mrioc->pdev->bus->number;
-	adpinfo.pci_dev = PCI_SLOT(mrioc->pdev->devfn);
-	adpinfo.pci_func = PCI_FUNC(mrioc->pdev->devfn);
-	adpinfo.pci_seg_id = pci_domain_nr(mrioc->pdev->bus);
 	adpinfo.app_intfc_ver = MPI3MR_IOCTL_VERSION;
 
-	ioc_state = mpi3mr_get_iocstate(mrioc);
-	if (ioc_state == MRIOC_STATE_UNRECOVERABLE)
-		adpinfo.adp_state = MPI3MR_BSG_ADPSTATE_UNRECOVERABLE;
-	else if ((mrioc->reset_in_progress) || (mrioc->stop_bsgs))
+	if (mrioc->reset_in_progress || mrioc->stop_bsgs ||
+	    mrioc->block_on_pcie_err)
 		adpinfo.adp_state = MPI3MR_BSG_ADPSTATE_IN_RESET;
-	else if (ioc_state == MRIOC_STATE_FAULT)
-		adpinfo.adp_state = MPI3MR_BSG_ADPSTATE_FAULT;
-	else
-		adpinfo.adp_state = MPI3MR_BSG_ADPSTATE_OPERATIONAL;
+	else {
+		ioc_state = mpi3mr_get_iocstate(mrioc);
+		if (ioc_state == MRIOC_STATE_UNRECOVERABLE)
+			adpinfo.adp_state = MPI3MR_BSG_ADPSTATE_UNRECOVERABLE;
+		else if (ioc_state == MRIOC_STATE_FAULT)
+			adpinfo.adp_state = MPI3MR_BSG_ADPSTATE_FAULT;
+		else
+			adpinfo.adp_state = MPI3MR_BSG_ADPSTATE_OPERATIONAL;
+	}
+
+	adpinfo.pci_dev_id = mrioc->pdevinfo.id;
+	adpinfo.pci_dev_hw_rev = mrioc->pdevinfo.revision;
+	adpinfo.pci_subsys_dev_id = mrioc->pdevinfo.ssid;
+	adpinfo.pci_subsys_ven_id = mrioc->pdevinfo.ssvid;
+	adpinfo.pci_bus = mrioc->pdevinfo.bus;
+	adpinfo.pci_dev = mrioc->pdevinfo.dev;
+	adpinfo.pci_func = mrioc->pdevinfo.func;
+	adpinfo.pci_seg_id = mrioc->pdevinfo.segment;
 
 	memcpy((u8 *)&adpinfo.driver_info, (u8 *)&mrioc->driver_info,
 	    sizeof(adpinfo.driver_info));
@@ -1495,7 +1502,7 @@  static long mpi3mr_bsg_process_mpt_cmds(struct bsg_job *job)
 		mutex_unlock(&mrioc->bsg_cmds.mutex);
 		goto out;
 	}
-	if (mrioc->stop_bsgs) {
+	if (mrioc->stop_bsgs || mrioc->block_on_pcie_err) {
 		dprint_bsg_err(mrioc, "%s: bsgs are blocked\n", __func__);
 		rval = -EAGAIN;
 		mutex_unlock(&mrioc->bsg_cmds.mutex);
@@ -2020,17 +2027,20 @@  adp_state_show(struct device *dev, struct device_attribute *attr,
 	enum mpi3mr_iocstate ioc_state;
 	uint8_t adp_state;
 
-	ioc_state = mpi3mr_get_iocstate(mrioc);
-	if (ioc_state == MRIOC_STATE_UNRECOVERABLE)
-		adp_state = MPI3MR_BSG_ADPSTATE_UNRECOVERABLE;
-	else if ((mrioc->reset_in_progress) || (mrioc->stop_bsgs))
+	if (mrioc->reset_in_progress || mrioc->stop_bsgs ||
+		 mrioc->block_on_pcie_err)
 		adp_state = MPI3MR_BSG_ADPSTATE_IN_RESET;
-	else if (ioc_state == MRIOC_STATE_FAULT)
-		adp_state = MPI3MR_BSG_ADPSTATE_FAULT;
-	else
-		adp_state = MPI3MR_BSG_ADPSTATE_OPERATIONAL;
+	else {
+		ioc_state = mpi3mr_get_iocstate(mrioc);
+		if (ioc_state == MRIOC_STATE_UNRECOVERABLE)
+			adp_state = MPI3MR_BSG_ADPSTATE_UNRECOVERABLE;
+		else if (ioc_state == MRIOC_STATE_FAULT)
+			adp_state = MPI3MR_BSG_ADPSTATE_FAULT;
+		else
+			adp_state = MPI3MR_BSG_ADPSTATE_OPERATIONAL;
+	}
 
-	return sysfs_emit(buf, "%u\n", adp_state);
+	return snprintf(buf, PAGE_SIZE, "%u\n", adp_state);
 }
 
 static DEVICE_ATTR_RO(adp_state);
diff --git a/drivers/scsi/mpi3mr/mpi3mr_fw.c b/drivers/scsi/mpi3mr/mpi3mr_fw.c
index 9e4a075fd7f0..e0d69f147fce 100644
--- a/drivers/scsi/mpi3mr/mpi3mr_fw.c
+++ b/drivers/scsi/mpi3mr/mpi3mr_fw.c
@@ -595,7 +595,7 @@  int mpi3mr_blk_mq_poll(struct Scsi_Host *shost, unsigned int queue_num)
 	mrioc = (struct mpi3mr_ioc *)shost->hostdata;
 
 	if ((mrioc->reset_in_progress || mrioc->prepare_for_reset ||
-	    mrioc->unrecoverable))
+	    mrioc->unrecoverable || mrioc->pcie_err_recovery))
 		return 0;
 
 	num_entries = mpi3mr_process_op_reply_q(mrioc,
@@ -1037,14 +1037,16 @@  enum mpi3mr_iocstate mpi3mr_get_iocstate(struct mpi3mr_ioc *mrioc)
 	u32 ioc_status, ioc_config;
 	u8 ready, enabled;
 
-	ioc_status = readl(&mrioc->sysif_regs->ioc_status);
-	ioc_config = readl(&mrioc->sysif_regs->ioc_configuration);
-
 	if (mrioc->unrecoverable)
 		return MRIOC_STATE_UNRECOVERABLE;
+
+	ioc_status = readl(&mrioc->sysif_regs->ioc_status);
+
 	if (ioc_status & MPI3_SYSIF_IOC_STATUS_FAULT)
 		return MRIOC_STATE_FAULT;
 
+	ioc_config = readl(&mrioc->sysif_regs->ioc_configuration);
+
 	ready = (ioc_status & MPI3_SYSIF_IOC_STATUS_READY);
 	enabled = (ioc_config & MPI3_SYSIF_IOC_CONFIG_ENABLE_IOC);
 
@@ -1667,6 +1669,12 @@  int mpi3mr_admin_request_post(struct mpi3mr_ioc *mrioc, void *admin_req,
 		retval = -EAGAIN;
 		goto out;
 	}
+	if (mrioc->pcie_err_recovery) {
+		ioc_err(mrioc, "admin request queue submission failed due to pcie error recovery in progress\n");
+		retval = -EAGAIN;
+		goto out;
+	}
+
 	areq_entry = (u8 *)mrioc->admin_req_base +
 	    (areq_pi * MPI3MR_ADMIN_REQ_FRAME_SZ);
 	memset(areq_entry, 0, MPI3MR_ADMIN_REQ_FRAME_SZ);
@@ -2337,6 +2345,11 @@  int mpi3mr_op_request_post(struct mpi3mr_ioc *mrioc,
 		retval = -EAGAIN;
 		goto out;
 	}
+	if (mrioc->pcie_err_recovery) {
+		ioc_err(mrioc, "operational request queue submission failed due to pcie error recovery in progress\n");
+		retval = -EAGAIN;
+		goto out;
+	}
 
 	segment_base_addr = segments[pi / op_req_q->segment_qd].segment;
 	req_entry = (u8 *)segment_base_addr +
@@ -2585,7 +2598,7 @@  static void mpi3mr_watchdog_work(struct work_struct *work)
 	u32 fault, host_diagnostic, ioc_status;
 	u32 reset_reason = MPI3MR_RESET_FROM_FAULT_WATCH;
 
-	if (mrioc->reset_in_progress)
+	if (mrioc->reset_in_progress || mrioc->pcie_err_recovery)
 		return;
 
 	if (!mrioc->unrecoverable && !pci_device_is_present(mrioc->pdev)) {
@@ -4111,7 +4124,7 @@  int mpi3mr_reinit_ioc(struct mpi3mr_ioc *mrioc, u8 is_resume)
 		goto out_failed_noretry;
 	}
 
-	if (is_resume) {
+	if (is_resume || mrioc->block_on_pcie_err) {
 		dprint_reset(mrioc, "setting up single ISR\n");
 		retval = mpi3mr_setup_isr(mrioc, 1);
 		if (retval) {
@@ -4151,7 +4164,7 @@  int mpi3mr_reinit_ioc(struct mpi3mr_ioc *mrioc, u8 is_resume)
 		goto out_failed;
 	}
 
-	if (is_resume) {
+	if (is_resume || mrioc->block_on_pcie_err) {
 		dprint_reset(mrioc, "setting up multiple ISR\n");
 		retval = mpi3mr_setup_isr(mrioc, 0);
 		if (retval) {
@@ -4645,7 +4658,8 @@  void mpi3mr_cleanup_ioc(struct mpi3mr_ioc *mrioc)
 
 	ioc_state = mpi3mr_get_iocstate(mrioc);
 
-	if ((!mrioc->unrecoverable) && (!mrioc->reset_in_progress) &&
+	if (!mrioc->unrecoverable && !mrioc->reset_in_progress &&
+	    !mrioc->pcie_err_recovery &&
 	    (ioc_state == MRIOC_STATE_READY)) {
 		if (mpi3mr_issue_and_process_mur(mrioc,
 		    MPI3MR_RESET_FROM_CTLR_CLEANUP))
diff --git a/drivers/scsi/mpi3mr/mpi3mr_os.c b/drivers/scsi/mpi3mr/mpi3mr_os.c
index 1bffd629c124..d58aa1ecf9b1 100644
--- a/drivers/scsi/mpi3mr/mpi3mr_os.c
+++ b/drivers/scsi/mpi3mr/mpi3mr_os.c
@@ -919,7 +919,7 @@  static int mpi3mr_report_tgtdev_to_host(struct mpi3mr_ioc *mrioc,
 	int retval = 0;
 	struct mpi3mr_tgt_dev *tgtdev;
 
-	if (mrioc->reset_in_progress)
+	if (mrioc->reset_in_progress || mrioc->pcie_err_recovery)
 		return -1;
 
 	tgtdev = mpi3mr_get_tgtdev_by_perst_id(mrioc, perst_id);
@@ -2000,9 +2000,13 @@  static void mpi3mr_fwevt_bh(struct mpi3mr_ioc *mrioc,
 	}
 	case MPI3_EVENT_WAIT_FOR_DEVICES_TO_REFRESH:
 	{
-		while (mrioc->device_refresh_on)
+		while ((mrioc->device_refresh_on || mrioc->block_on_pcie_err) &&
+		    !mrioc->unrecoverable && !mrioc->pcie_err_recovery) {
 			msleep(500);
+		}
 
+		if (mrioc->unrecoverable || mrioc->pcie_err_recovery)
+			break;
 		dprint_event_bh(mrioc,
 		    "scan for non responding and newly added devices after soft reset started\n");
 		if (mrioc->sas_transport_enabled) {
@@ -3680,6 +3684,13 @@  int mpi3mr_issue_tm(struct mpi3mr_ioc *mrioc, u8 tm_type,
 		mutex_unlock(&drv_cmd->mutex);
 		goto out;
 	}
+	if (mrioc->block_on_pcie_err) {
+		retval = -1;
+		dprint_tm(mrioc, "sending task management failed due to\n"
+				"pcie error recovery in progress\n");
+		mutex_unlock(&drv_cmd->mutex);
+		goto out;
+	}
 
 	drv_cmd->state = MPI3MR_CMD_PENDING;
 	drv_cmd->is_waiting = 1;
@@ -4073,12 +4084,19 @@  static int mpi3mr_eh_bus_reset(struct scsi_cmnd *scmd)
 	if (dev_type == MPI3_DEVICE_DEVFORM_VD) {
 		mpi3mr_wait_for_host_io(mrioc,
 			MPI3MR_RAID_ERRREC_RESET_TIMEOUT);
-		if (!mpi3mr_get_fw_pending_ios(mrioc))
+		if (!mpi3mr_get_fw_pending_ios(mrioc)) {
+			while (mrioc->reset_in_progress ||
+			       mrioc->prepare_for_reset ||
+			       mrioc->block_on_pcie_err)
+				ssleep(1);
 			retval = SUCCESS;
+			goto out;
+		}
 	}
 	if (retval == FAILED)
 		mpi3mr_print_pending_host_io(mrioc);
 
+out:
 	sdev_printk(KERN_INFO, scmd->device,
 		"Bus reset is %s for scmd(%p)\n",
 		((retval == SUCCESS) ? "SUCCESS" : "FAILED"), scmd);
@@ -4779,7 +4797,8 @@  static int mpi3mr_qcmd(struct Scsi_Host *shost,
 		goto out;
 	}
 
-	if (mrioc->reset_in_progress) {
+	if (mrioc->reset_in_progress || mrioc->prepare_for_reset
+	    || mrioc->block_on_pcie_err) {
 		retval = SCSI_MLQUEUE_HOST_BUSY;
 		goto out;
 	}
@@ -5123,8 +5142,15 @@  mpi3mr_probe(struct pci_dev *pdev, const struct pci_device_id *id)
 	mrioc->logging_level = logging_level;
 	mrioc->shost = shost;
 	mrioc->pdev = pdev;
+	mrioc->pdevinfo.id = pdev->device;
+	mrioc->pdevinfo.revision = pdev->revision;
+	mrioc->pdevinfo.ssid = pdev->subsystem_device;
+	mrioc->pdevinfo.ssvid = pdev->subsystem_vendor;
+	mrioc->pdevinfo.bus = pdev->bus->number;
+	mrioc->pdevinfo.dev = PCI_SLOT(pdev->devfn);
+	mrioc->pdevinfo.func = PCI_FUNC(pdev->devfn);
+	mrioc->pdevinfo.segment = pci_domain_nr(pdev->bus);
 	mrioc->stop_bsgs = 1;
-
 	mrioc->max_sgl_entries = max_sgl_entries;
 	if (max_sgl_entries > MPI3MR_MAX_SGL_ENTRIES)
 		mrioc->max_sgl_entries = MPI3MR_MAX_SGL_ENTRIES;
@@ -5230,6 +5256,35 @@  mpi3mr_probe(struct pci_dev *pdev, const struct pci_device_id *id)
 	return retval;
 }
 
+/**
+ * mpi3mr_get_shost_and_mrioc - get shost and ioc reference if
+ *			they are valid
+ * @pdev: PCI device struct
+ * @shost: address to store scsi host reference
+ * @ioc: address store HBA adapter reference
+ *
+ * Return: 0 if *shost and *ioc are not NULL otherwise -1.
+ */
+
+static int
+mpi3mr_get_shost_and_mrioc(struct pci_dev *pdev,
+	struct Scsi_Host **shost, struct mpi3mr_ioc **mrioc)
+{
+	*shost = pci_get_drvdata(pdev);
+	if (*shost == NULL) {
+		dev_err(&pdev->dev, "pdev's driver data is null\n");
+		return -1;
+	}
+
+	*mrioc = shost_priv(*shost);
+	if (*mrioc == NULL) {
+		dev_err(&pdev->dev, "shost's private data is null\n");
+		*shost = NULL;
+		return -1;
+}
+	return 0;
+}
+
 /**
  * mpi3mr_remove - PCI remove callback
  * @pdev: PCI device instance
@@ -5242,22 +5297,26 @@  mpi3mr_probe(struct pci_dev *pdev, const struct pci_device_id *id)
  */
 static void mpi3mr_remove(struct pci_dev *pdev)
 {
-	struct Scsi_Host *shost = pci_get_drvdata(pdev);
+	struct Scsi_Host *shost;
 	struct mpi3mr_ioc *mrioc;
 	struct workqueue_struct	*wq;
 	unsigned long flags;
 	struct mpi3mr_tgt_dev *tgtdev, *tgtdev_next;
-	struct mpi3mr_hba_port *port, *hba_port_next;
-	struct mpi3mr_sas_node *sas_expander, *sas_expander_next;
 
-	if (!shost)
+	if (mpi3mr_get_shost_and_mrioc(pdev, &shost, &mrioc))
 		return;
 
-	mrioc = shost_priv(shost);
 	while (mrioc->reset_in_progress || mrioc->is_driver_loading)
 		ssleep(1);
 
-	if (!pci_device_is_present(mrioc->pdev)) {
+	if (mrioc->block_on_pcie_err) {
+		mrioc->block_on_pcie_err = false;
+		scsi_unblock_requests(shost);
+		mrioc->unrecoverable = 1;
+	}
+
+	if (!pci_device_is_present(mrioc->pdev) ||
+	    mrioc->pcie_err_recovery) {
 		mrioc->unrecoverable = 1;
 		mpi3mr_flush_cmds_for_unrecovered_controller(mrioc);
 	}
@@ -5287,29 +5346,6 @@  static void mpi3mr_remove(struct pci_dev *pdev)
 	mpi3mr_cleanup_ioc(mrioc);
 	mpi3mr_free_mem(mrioc);
 	mpi3mr_cleanup_resources(mrioc);
-
-	spin_lock_irqsave(&mrioc->sas_node_lock, flags);
-	list_for_each_entry_safe_reverse(sas_expander, sas_expander_next,
-	    &mrioc->sas_expander_list, list) {
-		spin_unlock_irqrestore(&mrioc->sas_node_lock, flags);
-		mpi3mr_expander_node_remove(mrioc, sas_expander);
-		spin_lock_irqsave(&mrioc->sas_node_lock, flags);
-	}
-	list_for_each_entry_safe(port, hba_port_next, &mrioc->hba_port_table_list, list) {
-		ioc_info(mrioc,
-		    "removing hba_port entry: %p port: %d from hba_port list\n",
-		    port, port->port_id);
-		list_del(&port->list);
-		kfree(port);
-	}
-	spin_unlock_irqrestore(&mrioc->sas_node_lock, flags);
-
-	if (mrioc->sas_hba.num_phys) {
-		kfree(mrioc->sas_hba.phy);
-		mrioc->sas_hba.phy = NULL;
-		mrioc->sas_hba.num_phys = 0;
-	}
-
 	spin_lock(&mrioc_list_lock);
 	list_del(&mrioc->list);
 	spin_unlock(&mrioc_list_lock);
@@ -5328,19 +5364,25 @@  static void mpi3mr_remove(struct pci_dev *pdev)
  */
 static void mpi3mr_shutdown(struct pci_dev *pdev)
 {
-	struct Scsi_Host *shost = pci_get_drvdata(pdev);
+	struct Scsi_Host *shost;
 	struct mpi3mr_ioc *mrioc;
 	struct workqueue_struct	*wq;
 	unsigned long flags;
 
-	if (!shost)
+	if (mpi3mr_get_shost_and_mrioc(pdev, &shost, &mrioc))
 		return;
 
-	mrioc = shost_priv(shost);
 	while (mrioc->reset_in_progress || mrioc->is_driver_loading)
 		ssleep(1);
 
 	mrioc->stop_drv_processing = 1;
+
+	if (mrioc->block_on_pcie_err) {
+		mrioc->block_on_pcie_err = false;
+		scsi_unblock_requests(shost);
+		mrioc->unrecoverable = 1;
+	}
+
 	mpi3mr_cleanup_fwevt_list(mrioc);
 	spin_lock_irqsave(&mrioc->fwevt_lock, flags);
 	wq = mrioc->fwevt_worker_thread;
@@ -5361,19 +5403,18 @@  static void mpi3mr_shutdown(struct pci_dev *pdev)
  * Change the power state to the given value and cleanup the IOC
  * by issuing MUR and shutdown notification
  *
- * Return: 0 always.
+ * Return: 0 on success, non-zero on failure
  */
 static int __maybe_unused
 mpi3mr_suspend(struct device *dev)
 {
 	struct pci_dev *pdev = to_pci_dev(dev);
-	struct Scsi_Host *shost = pci_get_drvdata(pdev);
+	struct Scsi_Host *shost;
 	struct mpi3mr_ioc *mrioc;
 
-	if (!shost)
-		return 0;
+	if (mpi3mr_get_shost_and_mrioc(pdev, &shost, &mrioc))
+		return -1;
 
-	mrioc = shost_priv(shost);
 	while (mrioc->reset_in_progress || mrioc->is_driver_loading)
 		ssleep(1);
 	mrioc->stop_drv_processing = 1;
@@ -5402,15 +5443,13 @@  static int __maybe_unused
 mpi3mr_resume(struct device *dev)
 {
 	struct pci_dev *pdev = to_pci_dev(dev);
-	struct Scsi_Host *shost = pci_get_drvdata(pdev);
+	struct Scsi_Host *shost;
 	struct mpi3mr_ioc *mrioc;
 	pci_power_t device_state = pdev->current_state;
 	int r;
 
-	if (!shost)
-		return 0;
-
-	mrioc = shost_priv(shost);
+	if (mpi3mr_get_shost_and_mrioc(pdev, &shost, &mrioc))
+		return -1;
 
 	ioc_info(mrioc, "pdev=0x%p, slot=%s, previous operating state [D%d]\n",
 	    pdev, pci_name(pdev), device_state);
@@ -5440,6 +5479,195 @@  mpi3mr_resume(struct device *dev)
 	return 0;
 }
 
+/**
+ * mpi3mr_pcierr_detected - PCI error detected callback
+ * @pdev: PCI device instance
+ * @state: channel state
+ *
+ * This function is called by the PCI error recovery driver and
+ * based on the state passed the driver decides what actions to
+ * be recommended back to PCI driver.
+ *
+ * For all of the states if there is no valid mrioc or scsi host
+ * references in the pci device then this function will retyrn
+ * the resul as disconnect.
+ *
+ * For normal state, this function will return the result as can
+ * recover.
+ *
+ * For frozen state, this function will block for any pennding
+ * controller initialization or re-initialization to complete,
+ * stop any new interactions with the controller and return
+ * status as reset required.
+ *
+ * For permanent failure state, this function will mark the
+ * controller as unrecoverable and return status as disconnect.
+ *
+ * Returns: PCI_ERS_RESULT_NEED_RESET or CAN_RECOVER or
+ * DISCONNECT based on the controller state.
+ */
+static pci_ers_result_t
+mpi3mr_pcierr_detected(struct pci_dev *pdev, pci_channel_state_t state)
+{
+	struct Scsi_Host *shost;
+	struct mpi3mr_ioc *mrioc;
+	pci_ers_result_t ret_val = PCI_ERS_RESULT_DISCONNECT;
+
+	dev_info(&pdev->dev, "%s: callback invoked state(%d)\n", __func__,
+	    state);
+
+	if (mpi3mr_get_shost_and_mrioc(pdev, &shost, &mrioc)) {
+		dev_err(&pdev->dev, "device not available\n");
+		return ret_val;
+	}
+
+	switch (state) {
+	case pci_channel_io_normal:
+		ret_val = PCI_ERS_RESULT_CAN_RECOVER;
+		break;
+	case pci_channel_io_frozen:
+		mrioc->pcie_err_recovery = true;
+		mrioc->block_on_pcie_err = true;
+		while (mrioc->reset_in_progress || mrioc->is_driver_loading)
+			ssleep(1);
+		scsi_block_requests(mrioc->shost);
+		mpi3mr_stop_watchdog(mrioc);
+		mpi3mr_cleanup_resources(mrioc);
+		mrioc->pdev = NULL;
+		ret_val = PCI_ERS_RESULT_NEED_RESET;
+		break;
+	case pci_channel_io_perm_failure:
+		mrioc->pcie_err_recovery = true;
+		mrioc->block_on_pcie_err = true;
+		mrioc->unrecoverable = 1;
+		mpi3mr_stop_watchdog(mrioc);
+		mpi3mr_flush_cmds_for_unrecovered_controller(mrioc);
+		ret_val = PCI_ERS_RESULT_DISCONNECT;
+		break;
+	default:
+		break;
+	}
+	return ret_val;
+}
+
+/**
+ * mpi3mr_pcierr_slot_reset_done - Post slot reset callback
+ * @pdev: PCI device instance
+ *
+ * This function is called by the PCI error recovery driver
+ * after a slot or link reset issued by it for the recovery, the
+ * driver is expected to bring back the controller and
+ * initialize it.
+ *
+ * This function restores pci state and reinitializes controller
+ * resoruces and the controller, this blocks for any pending
+ * reset to complete.
+ *
+ * Returns: PCI_ERS_RESULT_DISCONNECT on failure or
+ * PCI_ERS_RESULT_RECOVERED
+ */
+static pci_ers_result_t mpi3mr_pcierr_slot_reset_done(struct pci_dev *pdev)
+{
+	struct Scsi_Host *shost;
+	struct mpi3mr_ioc *mrioc;
+
+
+	dev_info(&pdev->dev, "%s: callback invoked\n", __func__);
+
+	if (mpi3mr_get_shost_and_mrioc(pdev, &shost, &mrioc)) {
+		dev_err(&pdev->dev, "device not available\n");
+		return PCI_ERS_RESULT_DISCONNECT;
+	}
+
+	while (mrioc->reset_in_progress)
+		ssleep(1);
+
+	mrioc->pdev = pdev;
+	pci_restore_state(pdev);
+
+	if (mpi3mr_setup_resources(mrioc)) {
+		ioc_err(mrioc, "setup resources failed\n");
+		goto out_failed;
+	}
+	mrioc->unrecoverable = 0;
+	mrioc->pcie_err_recovery = false;
+
+	if (mpi3mr_soft_reset_handler(mrioc, MPI3MR_RESET_FROM_FIRMWARE, 0))
+		goto out_failed;
+
+	return PCI_ERS_RESULT_RECOVERED;
+
+out_failed:
+	mrioc->unrecoverable = 1;
+	mrioc->block_on_pcie_err = false;
+	scsi_unblock_requests(shost);
+	mpi3mr_start_watchdog(mrioc);
+	return PCI_ERS_RESULT_DISCONNECT;
+}
+
+/**
+ * mpi3mr_pcierr_resume - PCI error recovery resume
+ * callback
+ * @pdev: PCI device instance
+ *
+ * This function enables all I/O and IOCTLs post reset issued as
+ * part of the PCI advacned error reporting and handling
+ *
+ * Return: Nothing.
+ */
+static void mpi3mr_pcierr_resume(struct pci_dev *pdev)
+{
+	struct Scsi_Host *shost;
+	struct mpi3mr_ioc *mrioc;
+
+	dev_info(&pdev->dev, "%s: callback invoked\n", __func__);
+
+	if (mpi3mr_get_shost_and_mrioc(pdev, &shost, &mrioc)) {
+		dev_err(&pdev->dev, "device not available\n");
+		return;
+	}
+
+	pci_aer_clear_nonfatal_status(pdev);
+
+	if (mrioc->block_on_pcie_err) {
+		mrioc->block_on_pcie_err = false;
+		scsi_unblock_requests(shost);
+		mpi3mr_start_watchdog(mrioc);
+	}
+
+}
+
+/**
+ * mpi3mr_pcierr_mmio_enabled - PCI error recovery callback
+ * @pdev: PCI device instance
+ *
+ * This is called only if _pcierr_error_detected returns
+ * PCI_ERS_RESULT_CAN_RECOVER.
+ *
+ * Return: PCI_ERS_RESULT_DISCONNECT when the controller is
+ * unrecoverable or when the shost/mnrioc reference cannot be
+ * found, else return PCI_ERS_RESULT_RECOVERED
+ */
+static pci_ers_result_t mpi3mr_pcierr_mmio_enabled(struct pci_dev *pdev)
+{
+
+	struct Scsi_Host *shost;
+	struct mpi3mr_ioc *mrioc;
+/*
+ *
+ */
+	dev_info(&pdev->dev, "%s: callback invoked\n", __func__);
+
+	if (mpi3mr_get_shost_and_mrioc(pdev, &shost, &mrioc)) {
+		dev_err(&pdev->dev, "device not available\n");
+		return PCI_ERS_RESULT_DISCONNECT;
+	}
+	if (mrioc->unrecoverable)
+		return PCI_ERS_RESULT_DISCONNECT;
+
+	return PCI_ERS_RESULT_RECOVERED;
+}
+
 static const struct pci_device_id mpi3mr_pci_id_table[] = {
 	{
 		PCI_DEVICE_SUB(MPI3_MFGPAGE_VENDORID_BROADCOM,
@@ -5457,6 +5685,13 @@  static const struct pci_device_id mpi3mr_pci_id_table[] = {
 };
 MODULE_DEVICE_TABLE(pci, mpi3mr_pci_id_table);
 
+static struct pci_error_handlers mpi3mr_err_handler = {
+	.error_detected = mpi3mr_pcierr_detected,
+	.mmio_enabled = mpi3mr_pcierr_mmio_enabled,
+	.slot_reset = mpi3mr_pcierr_slot_reset_done,
+	.resume = mpi3mr_pcierr_resume,
+};
+
 static SIMPLE_DEV_PM_OPS(mpi3mr_pm_ops, mpi3mr_suspend, mpi3mr_resume);
 
 static struct pci_driver mpi3mr_pci_driver = {
@@ -5465,6 +5700,7 @@  static struct pci_driver mpi3mr_pci_driver = {
 	.probe = mpi3mr_probe,
 	.remove = mpi3mr_remove,
 	.shutdown = mpi3mr_shutdown,
+	.err_handler = &mpi3mr_err_handler,
 	.driver.pm = &mpi3mr_pm_ops,
 };
 
diff --git a/drivers/scsi/mpi3mr/mpi3mr_transport.c b/drivers/scsi/mpi3mr/mpi3mr_transport.c
index c0c8ab586957..8c8368104a27 100644
--- a/drivers/scsi/mpi3mr/mpi3mr_transport.c
+++ b/drivers/scsi/mpi3mr/mpi3mr_transport.c
@@ -149,6 +149,11 @@  static int mpi3mr_report_manufacture(struct mpi3mr_ioc *mrioc,
 		return -EFAULT;
 	}
 
+	if (mrioc->pcie_err_recovery) {
+		ioc_err(mrioc, "%s: pcie error recovery in progress!\n", __func__);
+		return -EFAULT;
+	}
+
 	data_out_sz = sizeof(struct rep_manu_request);
 	data_in_sz = sizeof(struct rep_manu_reply);
 	data_out = dma_alloc_coherent(&mrioc->pdev->dev,
@@ -792,6 +797,12 @@  static int mpi3mr_set_identify(struct mpi3mr_ioc *mrioc, u16 handle,
 		return -EFAULT;
 	}
 
+	if (mrioc->pcie_err_recovery) {
+		ioc_err(mrioc, "%s: pcie error recovery in progress!\n",
+		    __func__);
+		return -EFAULT;
+	}
+
 	if ((mpi3mr_cfg_get_dev_pg0(mrioc, &ioc_status, &device_pg0,
 	    sizeof(device_pg0), MPI3_DEVICE_PGAD_FORM_HANDLE, handle))) {
 		ioc_err(mrioc, "%s: device page0 read failed\n", __func__);
@@ -1009,6 +1020,9 @@  mpi3mr_alloc_hba_port(struct mpi3mr_ioc *mrioc, u16 port_id)
 	hba_port->port_id = port_id;
 	ioc_info(mrioc, "hba_port entry: %p, port: %d is added to hba_port list\n",
 	    hba_port, hba_port->port_id);
+	if (mrioc->reset_in_progress ||
+		mrioc->pcie_err_recovery)
+		hba_port->flags = MPI3MR_HBA_PORT_FLAG_NEW;
 	list_add_tail(&hba_port->list, &mrioc->hba_port_table_list);
 	return hba_port;
 }
@@ -1057,7 +1071,7 @@  void mpi3mr_update_links(struct mpi3mr_ioc *mrioc,
 	struct mpi3mr_sas_node *mr_sas_node;
 	struct mpi3mr_sas_phy *mr_sas_phy;
 
-	if (mrioc->reset_in_progress)
+	if (mrioc->reset_in_progress || mrioc->pcie_err_recovery)
 		return;
 
 	spin_lock_irqsave(&mrioc->sas_node_lock, flags);
@@ -1965,7 +1979,7 @@  int mpi3mr_expander_add(struct mpi3mr_ioc *mrioc, u16 handle)
 	if (!handle)
 		return -1;
 
-	if (mrioc->reset_in_progress)
+	if (mrioc->reset_in_progress || mrioc->pcie_err_recovery)
 		return -1;
 
 	if ((mpi3mr_cfg_get_sas_exp_pg0(mrioc, &ioc_status, &expander_pg0,
@@ -2171,7 +2185,7 @@  void mpi3mr_expander_node_remove(struct mpi3mr_ioc *mrioc,
 	/* remove sibling ports attached to this expander */
 	list_for_each_entry_safe(mr_sas_port, next,
 	   &sas_expander->sas_port_list, port_list) {
-		if (mrioc->reset_in_progress)
+		if (mrioc->reset_in_progress || mrioc->pcie_err_recovery)
 			return;
 		if (mr_sas_port->remote_identify.device_type ==
 		    SAS_END_DEVICE)
@@ -2221,7 +2235,7 @@  void mpi3mr_expander_remove(struct mpi3mr_ioc *mrioc, u64 sas_address,
 	struct mpi3mr_sas_node *sas_expander;
 	unsigned long flags;
 
-	if (mrioc->reset_in_progress)
+	if (mrioc->reset_in_progress || mrioc->pcie_err_recovery)
 		return;
 
 	if (!hba_port)
@@ -2532,6 +2546,11 @@  static int mpi3mr_get_expander_phy_error_log(struct mpi3mr_ioc *mrioc,
 		return -EFAULT;
 	}
 
+	if (mrioc->pcie_err_recovery) {
+		ioc_err(mrioc, "%s: pcie error recovery in progress!\n", __func__);
+		return -EFAULT;
+	}
+
 	data_out_sz = sizeof(struct phy_error_log_request);
 	data_in_sz = sizeof(struct phy_error_log_reply);
 	sz = data_out_sz + data_in_sz;
@@ -2791,6 +2810,12 @@  mpi3mr_expander_phy_control(struct mpi3mr_ioc *mrioc,
 		return -EFAULT;
 	}
 
+	if (mrioc->pcie_err_recovery) {
+		ioc_err(mrioc, "%s: pcie error recovery in progress!\n",
+		    __func__);
+		return -EFAULT;
+	}
+
 	data_out_sz = sizeof(struct phy_control_request);
 	data_in_sz = sizeof(struct phy_control_reply);
 	sz = data_out_sz + data_in_sz;
@@ -3214,6 +3239,12 @@  mpi3mr_transport_smp_handler(struct bsg_job *job, struct Scsi_Host *shost,
 		goto out;
 	}
 
+	if (mrioc->pcie_err_recovery) {
+		ioc_err(mrioc, "%s: pcie error recovery in progress!\n", __func__);
+		rc = -EFAULT;
+		goto out;
+	}
+
 	rc = mpi3mr_map_smp_buffer(&mrioc->pdev->dev, &job->request_payload,
 	    &dma_addr_out, &dma_len_out, &addr_out);
 	if (rc)