releases/4.19.34/scsi-core-replace-gfp_atomic-with-gfp_kernel-in-scsi.patch

   1 From b82e5b9ad86533fd2b65a3966a0133d4b2d1e1da Mon Sep 17 00:00:00 2001
   2 From: Benjamin Block <bblock@linux.ibm.com>
   3 Date: Thu, 21 Feb 2019 10:18:00 +0100
   4 Subject: scsi: core: replace GFP_ATOMIC with GFP_KERNEL in scsi_scan.c
   5
   6 [ Upstream commit 1749ef00f7312679f76d5e9104c5d1e22a829038 ]
   7
   8 We had a test-report where, under memory pressure, adding LUNs to the
   9 systems would fail (the tests add LUNs strictly in sequence):
  10
  11 [ 5525.853432] scsi 0:0:1:1088045124: Direct-Access     IBM      2107900          .148 PQ: 0 ANSI: 5
  12 [ 5525.853826] scsi 0:0:1:1088045124: alua: supports implicit TPGS
  13 [ 5525.853830] scsi 0:0:1:1088045124: alua: device naa.6005076303ffd32700000000000044da port group 0 rel port 43
  14 [ 5525.853931] sd 0:0:1:1088045124: Attached scsi generic sg10 type 0
  15 [ 5525.854075] sd 0:0:1:1088045124: [sdk] Disabling DIF Type 1 protection
  16 [ 5525.855495] sd 0:0:1:1088045124: [sdk] 2097152 512-byte logical blocks: (1.07 GB/1.00 GiB)
  17 [ 5525.855606] sd 0:0:1:1088045124: [sdk] Write Protect is off
  18 [ 5525.855609] sd 0:0:1:1088045124: [sdk] Mode Sense: ed 00 00 08
  19 [ 5525.855795] sd 0:0:1:1088045124: [sdk] Write cache: enabled, read cache: enabled, doesn't support DPO or FUA
  20 [ 5525.857838]  sdk: sdk1
  21 [ 5525.859468] sd 0:0:1:1088045124: [sdk] Attached SCSI disk
  22 [ 5525.865073] sd 0:0:1:1088045124: alua: transition timeout set to 60 seconds
  23 [ 5525.865078] sd 0:0:1:1088045124: alua: port group 00 state A preferred supports tolusnA
  24 [ 5526.015070] sd 0:0:1:1088045124: alua: port group 00 state A preferred supports tolusnA
  25 [ 5526.015213] sd 0:0:1:1088045124: alua: port group 00 state A preferred supports tolusnA
  26 [ 5526.587439] scsi_alloc_sdev: Allocation failure during SCSI scanning, some SCSI devices might not be configured
  27 [ 5526.588562] scsi_alloc_sdev: Allocation failure during SCSI scanning, some SCSI devices might not be configured
  28
  29 Looking at the code of scsi_alloc_sdev(), and all the calling contexts,
  30 there seems to be no reason to use GFP_ATMOIC here. All the different
  31 call-contexts use a mutex at some point, and nothing in between that
  32 requires no sleeping, as far as I could see. Additionally, the code that
  33 later allocates the block queue for the device (scsi_mq_alloc_queue())
  34 already uses GFP_KERNEL.
  35
  36 There are similar allocations in two other functions:
  37 scsi_probe_and_add_lun(), and scsi_add_lun(),; that can also be done with
  38 GFP_KERNEL.
  39
  40 Here is the contexts for the three functions so far:
  41
  42     scsi_alloc_sdev()
  43         scsi_probe_and_add_lun()
  44             scsi_sequential_lun_scan()
  45                 __scsi_scan_target()
  46                     scsi_scan_target()
  47                         mutex_lock()
  48                     scsi_scan_channel()
  49                         scsi_scan_host_selected()
  50                             mutex_lock()
  51             scsi_report_lun_scan()
  52                 __scsi_scan_target()
  53                     ...
  54             __scsi_add_device()
  55                 mutex_lock()
  56             __scsi_scan_target()
  57                 ...
  58         scsi_report_lun_scan()
  59             ...
  60         scsi_get_host_dev()
  61             mutex_lock()
  62
  63     scsi_probe_and_add_lun()
  64         ...
  65
  66     scsi_add_lun()
  67         scsi_probe_and_add_lun()
  68             ...
  69
  70 So replace all these, and give them a bit of a better chance to succeed,
  71 with more chances of reclaim.
  72
  73 Signed-off-by: Benjamin Block <bblock@linux.ibm.com>
  74 Reviewed-by: Bart Van Assche <bvanassche@acm.org>
  75 Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
  76 Signed-off-by: Sasha Levin <sashal@kernel.org>
  77 ---
  78  drivers/scsi/scsi_scan.c | 6 +++---
  79  1 file changed, 3 insertions(+), 3 deletions(-)
  80
  81 diff --git a/drivers/scsi/scsi_scan.c b/drivers/scsi/scsi_scan.c
  82 index 78ca63dfba4a..9a7e3a3bd5ce 100644
  83 --- a/drivers/scsi/scsi_scan.c
  84 +++ b/drivers/scsi/scsi_scan.c
  85 @@ -220,7 +220,7 @@ static struct scsi_device *scsi_alloc_sdev(struct scsi_target *starget,
  86         struct Scsi_Host *shost = dev_to_shost(starget->dev.parent);
  87
  88         sdev = kzalloc(sizeof(*sdev) + shost->transportt->device_size,
  89 -                      GFP_ATOMIC);
  90 +                      GFP_KERNEL);
  91         if (!sdev)
  92                 goto out;
  93
  94 @@ -796,7 +796,7 @@ static int scsi_add_lun(struct scsi_device *sdev, unsigned char *inq_result,
  95          */
  96         sdev->inquiry = kmemdup(inq_result,
  97                                 max_t(size_t, sdev->inquiry_len, 36),
  98 -                               GFP_ATOMIC);
  99 +                               GFP_KERNEL);
 100         if (sdev->inquiry == NULL)
 101                 return SCSI_SCAN_NO_RESPONSE;
 102
 103 @@ -1087,7 +1087,7 @@ static int scsi_probe_and_add_lun(struct scsi_target *starget,
 104         if (!sdev)
 105                 goto out;
 106
 107 -       result = kmalloc(result_len, GFP_ATOMIC |
 108 +       result = kmalloc(result_len, GFP_KERNEL |
 109                         ((shost->unchecked_isa_dma) ? __GFP_DMA : 0));
 110         if (!result)
 111                 goto out_free_sdev;
 112 --
 113 2.19.1
 114