From patchwork Mon May 24 01:01:36 2021 Content-Type: text/plain; charset="utf-8" MIME-Version: 1.0 Content-Transfer-Encoding: 7bit X-Patchwork-Submitter: Douglas Gilbert X-Patchwork-Id: 12275287 Return-Path: X-Spam-Checker-Version: SpamAssassin 3.4.0 (2014-02-07) on aws-us-west-2-korg-lkml-1.web.codeaurora.org X-Spam-Level: X-Spam-Status: No, score=-16.8 required=3.0 tests=BAYES_00, HEADER_FROM_DIFFERENT_DOMAINS,INCLUDES_CR_TRAILER,INCLUDES_PATCH, MAILING_LIST_MULTI,SPF_HELO_NONE,SPF_PASS,USER_AGENT_GIT autolearn=ham autolearn_force=no version=3.4.0 Received: from mail.kernel.org (mail.kernel.org [198.145.29.99]) by smtp.lore.kernel.org (Postfix) with ESMTP id 92DD9C47082 for ; Mon, 24 May 2021 01:02:52 +0000 (UTC) Received: from vger.kernel.org (vger.kernel.org [23.128.96.18]) by mail.kernel.org (Postfix) with ESMTP id 6C6AC611CB for ; Mon, 24 May 2021 01:02:52 +0000 (UTC) Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S232295AbhEXBET (ORCPT ); Sun, 23 May 2021 21:04:19 -0400 Received: from smtp.infotech.no ([82.134.31.41]:33231 "EHLO smtp.infotech.no" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S232224AbhEXBEG (ORCPT ); Sun, 23 May 2021 21:04:06 -0400 Received: from localhost (localhost [127.0.0.1]) by smtp.infotech.no (Postfix) with ESMTP id EED952042A5; Mon, 24 May 2021 03:02:35 +0200 (CEST) X-Virus-Scanned: by amavisd-new-2.6.6 (20110518) (Debian) at infotech.no Received: from smtp.infotech.no ([127.0.0.1]) by localhost (smtp.infotech.no [127.0.0.1]) (amavisd-new, port 10024) with ESMTP id TnYFr-AeZJAR; Mon, 24 May 2021 03:02:34 +0200 (CEST) Received: from xtwo70.bingwo.ca (host-45-58-219-4.dyn.295.ca [45.58.219.4]) by smtp.infotech.no (Postfix) with ESMTPA id 9819820425C; Mon, 24 May 2021 03:02:29 +0200 (CEST) From: Douglas Gilbert To: linux-scsi@vger.kernel.org Cc: martin.petersen@oracle.com, jejb@linux.vnet.ibm.com, hare@suse.de Subject: [PATCH v19 34/45] sg: protect multiple receivers Date: Sun, 23 May 2021 21:01:36 -0400 Message-Id: <20210524010147.94845-35-dgilbert@interlog.com> X-Mailer: git-send-email 2.25.1 In-Reply-To: <20210524010147.94845-1-dgilbert@interlog.com> References: <20210524010147.94845-1-dgilbert@interlog.com> MIME-Version: 1.0 Precedence: bulk List-ID: X-Mailing-List: linux-scsi@vger.kernel.org If two threads call ioctl(SG_IORECEIVE) [or read()] on the same file descriptor there is a potential race on the same request response. Use atomic bit operations to make sure only one thread gets each request response. [The other thread will either get another request response or nothing.] Also make sfp cleanup a bit more robust and report if the number of submitted requests (which are decremented when completed) is other than the expected value of zero. Reviewed-by: Hannes Reinecke Signed-off-by: Douglas Gilbert --- drivers/scsi/sg.c | 48 ++++++++++++++++++++++++++++++++++------------- 1 file changed, 35 insertions(+), 13 deletions(-) diff --git a/drivers/scsi/sg.c b/drivers/scsi/sg.c index 9dc4946261d5..c67edc76d0e1 100644 --- a/drivers/scsi/sg.c +++ b/drivers/scsi/sg.c @@ -109,6 +109,7 @@ enum sg_rq_state { /* N.B. sg_rq_state_arr assumes SG_RS_AWAIT_RCV==2 */ #define SG_FRQ_SYNC_INVOC 2 /* synchronous (blocking) invocation */ #define SG_FRQ_NO_US_XFER 3 /* no user space transfer of data */ #define SG_FRQ_DEACT_ORPHAN 6 /* not keeping orphan so de-activate */ +#define SG_FRQ_RECEIVING 7 /* guard against multiple receivers */ /* Bit positions (flags) for sg_fd::ffd_bm bitmask follow */ #define SG_FFD_FORCE_PACKID 0 /* receive only given pack_id/tag */ @@ -1275,6 +1276,7 @@ sg_ctl_ioreceive(struct file *filp, struct sg_fd *sfp, void __user *p) SG_LOG(3, sfp, "%s: non_block(+IMMED)=%d\n", __func__, non_block); /* read in part of v3 or v4 header for pack_id or tag based find */ id = pack_id; +try_again: srp = sg_find_srp_by_id(sfp, id); if (!srp) { /* nothing available so wait on packet or */ if (unlikely(SG_IS_DETACHING(sdp))) @@ -1289,6 +1291,10 @@ sg_ctl_ioreceive(struct file *filp, struct sg_fd *sfp, void __user *p) if (res) return res; /* signal --> -ERESTARTSYS */ } /* now srp should be valid */ + if (test_and_set_bit(SG_FRQ_RECEIVING, srp->frq_bm)) { + cpu_relax(); + goto try_again; + } return sg_receive_v4(sfp, srp, p, h4p); } @@ -1325,7 +1331,7 @@ sg_ctl_ioreceive_v3(struct file *filp, struct sg_fd *sfp, void __user *p) if (test_bit(SG_FFD_FORCE_PACKID, sfp->ffd_bm)) pack_id = h3p->pack_id; - +try_again: srp = sg_find_srp_by_id(sfp, pack_id); if (!srp) { /* nothing available so wait on packet or */ if (unlikely(SG_IS_DETACHING(sdp))) @@ -1340,6 +1346,10 @@ sg_ctl_ioreceive_v3(struct file *filp, struct sg_fd *sfp, void __user *p) if (unlikely(res)) return res; /* signal --> -ERESTARTSYS */ } /* now srp should be valid */ + if (test_and_set_bit(SG_FRQ_RECEIVING, srp->frq_bm)) { + cpu_relax(); + goto try_again; + } return sg_receive_v3(sfp, srp, SZ_SG_IO_HDR, p); } @@ -1492,6 +1502,7 @@ sg_read(struct file *filp, char __user *p, size_t count, loff_t *ppos) want_id = h2p->pack_id; } } +try_again: srp = sg_find_srp_by_id(sfp, want_id); if (!srp) { /* nothing available so wait on packet to arrive or */ if (unlikely(SG_IS_DETACHING(sdp))) @@ -1507,6 +1518,10 @@ sg_read(struct file *filp, char __user *p, size_t count, loff_t *ppos) return ret; /* otherwise srp should be valid */ } + if (test_and_set_bit(SG_FRQ_RECEIVING, srp->frq_bm)) { + cpu_relax(); + goto try_again; + } if (srp->s_hdr3.interface_id == '\0') ret = sg_read_v1v2(p, (int)count, sfp, srp); else @@ -2975,28 +2990,29 @@ sg_finish_scsi_blk_rq(struct sg_request *srp) atomic_dec(&sfp->submitted); atomic_dec(&sfp->waiting); } + + /* Expect blk_put_request(rq) already called in sg_rq_end_io() */ + if (rq) { /* blk_get_request() may have failed */ + srp->rq = NULL; + if (scsi_req(rq)) + scsi_req_free_cmd(scsi_req(rq)); + blk_put_request(rq); + } if (srp->bio) { bool us_xfer = !test_bit(SG_FRQ_NO_US_XFER, srp->frq_bm); + struct bio *bio = srp->bio; - if (us_xfer) { - ret = blk_rq_unmap_user(srp->bio); + srp->bio = NULL; + if (us_xfer && bio) { + ret = blk_rq_unmap_user(bio); if (ret) { /* -EINTR (-4) can be ignored */ SG_LOG(6, sfp, "%s: blk_rq_unmap_user() --> %d\n", __func__, ret); } } - srp->bio = NULL; - } - /* In worst case READ data returned to user space by this point */ - - /* Expect blk_put_request(rq) already called in sg_rq_end_io() */ - if (rq) { /* blk_get_request() may have failed */ - if (scsi_req(rq)) - scsi_req_free_cmd(scsi_req(rq)); - srp->rq = NULL; - blk_put_request(rq); } + /* In worst case, READ data returned to user space by this point */ } static int @@ -3421,6 +3437,7 @@ sg_deact_request(struct sg_fd *sfp, struct sg_request *srp) return; sbp = srp->sense_bp; srp->sense_bp = NULL; + srp->frq_bm[0] = 0; sg_rq_state_chg(srp, 0, SG_RS_INACTIVE, true /* force */, __func__); /* maybe orphaned req, thus never read */ if (sbp) @@ -3553,6 +3570,7 @@ static void sg_remove_sfp_usercontext(struct work_struct *work) { __maybe_unused int o_count; + int subm; unsigned long idx, iflags; struct sg_device *sdp; struct sg_fd *sfp = container_of(work, struct sg_fd, ew_fd.work); @@ -3590,6 +3608,10 @@ sg_remove_sfp_usercontext(struct work_struct *work) SG_LOG(6, sfp, "%s: kfree: srp=%pK --\n", __func__, srp); kfree(srp); } + subm = atomic_read(&sfp->submitted); + if (subm != 0) + SG_LOG(1, sfp, "%s: expected submitted=0 got %d\n", + __func__, subm); xa_destroy(xafp); xa_lock_irqsave(xadp, iflags); e_sfp = __xa_erase(xadp, sfp->idx);