linux-snps-arc.lists.infradead.org archive mirror
 help / color / mirror / Atom feed
From: Vineet.Gupta1@synopsys.com (Vineet Gupta)
To: linux-snps-arc@lists.infradead.org
Subject: [PATCH 5/6] ARC: mm: tlb flush optim: elide repeated uTLB invalidate in loop
Date: Mon, 16 Sep 2019 14:32:06 -0700	[thread overview]
Message-ID: <20190916213207.12792-6-vgupta@synopsys.com> (raw)
In-Reply-To: <20190916213207.12792-1-vgupta@synopsys.com>

The unconditional full TLB flush (on say ASID rollover) iterates over each
entry and uses TLBWrite to zero it out. TLBWrite by design also invalidates
the uTLBs thus we end up invalidating it as many times as numbe rof
entries (512 or 1k)

Optimize this by using a weaker TLBWriteNI cmd in loop, which doesn't
tinker with uTLBs and an explicit one time IVUTLB, outside the loop to
invalidate them all once.

And given the optimiztion, the IVUTLB is now needed on MMUv4 too where
the uTLBs and JTLBs are otherwise coherent given the TLBInsertEntry /
TLBDeleteEntry commands

Signed-off-by: Vineet Gupta <vgupta at synopsys.com>
---
 arch/arc/mm/tlb.c | 74 +++++++++++++++++++----------------------------
 1 file changed, 29 insertions(+), 45 deletions(-)

diff --git a/arch/arc/mm/tlb.c b/arch/arc/mm/tlb.c
index 417f05ac4397..210d807983dd 100644
--- a/arch/arc/mm/tlb.c
+++ b/arch/arc/mm/tlb.c
@@ -118,6 +118,33 @@ static inline void __tlb_entry_erase(void)
 	write_aux_reg(ARC_REG_TLBCOMMAND, TLBWrite);
 }
 
+static void utlb_invalidate(void)
+{
+#if (CONFIG_ARC_MMU_VER >= 2)
+
+#if (CONFIG_ARC_MMU_VER == 2)
+	/* MMU v2 introduced the uTLB Flush command.
+	 * There was however an obscure hardware bug, where uTLB flush would
+	 * fail when a prior probe for J-TLB (both totally unrelated) would
+	 * return lkup err - because the entry didn't exist in MMU.
+	 * The Workround was to set Index reg with some valid value, prior to
+	 * flush. This was fixed in MMU v3
+	 */
+	unsigned int idx;
+
+	/* make sure INDEX Reg is valid */
+	idx = read_aux_reg(ARC_REG_TLBINDEX);
+
+	/* If not write some dummy val */
+	if (unlikely(idx & TLB_LKUP_ERR))
+		write_aux_reg(ARC_REG_TLBINDEX, 0xa);
+#endif
+
+	write_aux_reg(ARC_REG_TLBCOMMAND, TLBIVUTLB);
+#endif
+
+}
+
 #if (CONFIG_ARC_MMU_VER < 4)
 
 static inline unsigned int tlb_entry_lkup(unsigned long vaddr_n_asid)
@@ -149,44 +176,6 @@ static void tlb_entry_erase(unsigned int vaddr_n_asid)
 	}
 }
 
-/****************************************************************************
- * ARC700 MMU caches recently used J-TLB entries (RAM) as uTLBs (FLOPs)
- *
- * New IVUTLB cmd in MMU v2 explictly invalidates the uTLB
- *
- * utlb_invalidate ( )
- *  -For v2 MMU calls Flush uTLB Cmd
- *  -For v1 MMU does nothing (except for Metal Fix v1 MMU)
- *      This is because in v1 TLBWrite itself invalidate uTLBs
- ***************************************************************************/
-
-static void utlb_invalidate(void)
-{
-#if (CONFIG_ARC_MMU_VER >= 2)
-
-#if (CONFIG_ARC_MMU_VER == 2)
-	/* MMU v2 introduced the uTLB Flush command.
-	 * There was however an obscure hardware bug, where uTLB flush would
-	 * fail when a prior probe for J-TLB (both totally unrelated) would
-	 * return lkup err - because the entry didn't exist in MMU.
-	 * The Workround was to set Index reg with some valid value, prior to
-	 * flush. This was fixed in MMU v3 hence not needed any more
-	 */
-	unsigned int idx;
-
-	/* make sure INDEX Reg is valid */
-	idx = read_aux_reg(ARC_REG_TLBINDEX);
-
-	/* If not write some dummy val */
-	if (unlikely(idx & TLB_LKUP_ERR))
-		write_aux_reg(ARC_REG_TLBINDEX, 0xa);
-#endif
-
-	write_aux_reg(ARC_REG_TLBCOMMAND, TLBIVUTLB);
-#endif
-
-}
-
 static void tlb_entry_insert(unsigned int pd0, pte_t pd1)
 {
 	unsigned int idx;
@@ -219,11 +208,6 @@ static void tlb_entry_insert(unsigned int pd0, pte_t pd1)
 
 #else	/* CONFIG_ARC_MMU_VER >= 4) */
 
-static void utlb_invalidate(void)
-{
-	/* No need since uTLB is always in sync with JTLB */
-}
-
 static void tlb_entry_erase(unsigned int vaddr_n_asid)
 {
 	write_aux_reg(ARC_REG_TLBPD0, vaddr_n_asid | _PAGE_PRESENT);
@@ -267,7 +251,7 @@ noinline void local_flush_tlb_all(void)
 	for (entry = 0; entry < num_tlb; entry++) {
 		/* write this entry to the TLB */
 		write_aux_reg(ARC_REG_TLBINDEX, entry);
-		write_aux_reg(ARC_REG_TLBCOMMAND, TLBWrite);
+		write_aux_reg(ARC_REG_TLBCOMMAND, TLBWriteNI);
 	}
 
 	if (IS_ENABLED(CONFIG_TRANSPARENT_HUGEPAGE)) {
@@ -278,7 +262,7 @@ noinline void local_flush_tlb_all(void)
 
 		for (entry = stlb_idx; entry < stlb_idx + 16; entry++) {
 			write_aux_reg(ARC_REG_TLBINDEX, entry);
-			write_aux_reg(ARC_REG_TLBCOMMAND, TLBWrite);
+			write_aux_reg(ARC_REG_TLBCOMMAND, TLBWriteNI);
 		}
 	}
 
-- 
2.20.1

  parent reply	other threads:[~2019-09-16 21:32 UTC|newest]

Thread overview: 9+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2019-09-16 21:32 [PATCH 0/6] ARC MMU code updates Vineet Gupta
2019-09-16 21:32 ` [PATCH 1/6] ARCv2: mm: TLB Miss optim: SMP builds can cache pgd pointer in mmu scratch reg Vineet Gupta
2019-09-16 21:32 ` [PATCH 2/6] ARCv2: mm: TLB Miss optim: Use double world load/stores LDD/STD Vineet Gupta
2019-09-16 21:32 ` [PATCH 3/6] ARC: mm: TLB Miss optim: avoid re-reading ECR Vineet Gupta
2019-09-16 23:36   ` Alexey Brodkin
2019-09-17 21:08     ` Vineet Gupta
2019-09-16 21:32 ` [PATCH 4/6] ARC: mm: tlb flush optim: Make TLBWriteNI fallback to TLBWrite if not available Vineet Gupta
2019-09-16 21:32 ` Vineet Gupta [this message]
2019-09-16 21:32 ` [PATCH 6/6] ARC: mm: tlb flush optim: elide redundant uTLB invalidates for MMUv3 Vineet Gupta

Reply instructions:

You may reply publicly to this message via plain-text email
using any one of the following methods:

* Save the following mbox file, import it into your mail client,
  and reply-to-all from there: mbox

  Avoid top-posting and favor interleaved quoting:
  https://en.wikipedia.org/wiki/Posting_style#Interleaved_style

* Reply using the --to, --cc, and --in-reply-to
  switches of git-send-email(1):

  git send-email \
    --in-reply-to=20190916213207.12792-6-vgupta@synopsys.com \
    --to=vineet.gupta1@synopsys.com \
    --cc=linux-snps-arc@lists.infradead.org \
    /path/to/YOUR_REPLY

  https://kernel.org/pub/software/scm/git/docs/git-send-email.html

* If your mail client supports setting the In-Reply-To header
  via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line before the message body. 
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).